Automated composer recognition for multi-voice piano compositions using rhythmic features, n-grams and modified cortical algorithms

Size: px
Start display at page:

Download "Automated composer recognition for multi-voice piano compositions using rhythmic features, n-grams and modified cortical algorithms"

Transcription

1 Complex Intell. Syst. (2018) 4: ORIGINAL ARTICLE Automated composer recognton for mult-voce pano compostons usng rhythmc features, n-grams and modfed cortcal algorthms Nadne Hajj 2 Maurce Flo 1 Marette Awad 2 Receved: 6 March 2016 / Accepted: 19 July 2017 / Publshed onlne: 8 August 2017 The Author(s) Ths artcle s an open access publcaton Abstract Wth the explosve growth of dgtal musc data beng stored and easly reachable on the cloud, as well as the ncreased nterest n affectve and cogntve computng, dentfyng composers based on ther muscal work s an nterestng challenge for machne learnng and artfcal ntellgence to explore. Capturng style and recognzng musc composers have always been perceved reserved for traned muscal ears. Whle there have been many researchers targetng musc genre classfcaton for mproved recommendaton systems and lstener experence, few works have addressed automatc recognton of classcal pano composers as proposed n ths paper. Ths paper dscusses the applcablty of n-grams on MIDI musc scores coupled wth rhythmc features for feature extracton specfcally of mult-voce scores. In addton, cortcal algorthms (CA) are adapted to reduce the large feature set obtaned as well as to effcently dentfy composers n a supervsed manner. When used to classfy unknown composers and capture dfferent styles, our proposed approach acheved a recognton rate of 94.4% on a home grown database of 1197 peces wth only 0.1% of the 231,542 generated features whch motvates follow-on research. The retaned most sgnfcant features, ndeed, provded nterestng conclusons on capturng musc style of pano composers. B Marette Awad Nadne Hajj Maurce Flo 1 Department of Mechancal Engneerng, Unversty of Calforna Santa Barbara, Santa Barbara, CA 93117, USA 2 Department of Electrcal and Computer Engneerng, Amercan Unversty of Berut, Berut, Lebanon Keywords Composer recognton n-grams Cortcal algorthms Introducton The exponental growth of the musc database easly accessble and stored n the cloud promotes automated classfcaton of muscal peces for mproved user experence and recommender systems. For classcal musc, the unqueness of each classcal musc composer s revealed by hs/her own dstnct style: a style that s characterzed by varous factors such as the rhythmc structure that defnes the speed and varatons of the musc pece, the ptch that establshes the melody characterzng the musc as beng joyful, passonate, dramatc, ntense, etc. Often, these varous styles are the consequences of ether the lfestyle of the composers or the tme era they lved n. Our work focuses on revealng the structures n pano compostons that defne the style of the composer. Ths allows us to move a step closer to what actually defnes musc style that dstngushes one pano composer from the other. Ideally, style dentfcaton can help musc theorsts, recommender systems, musc annotaton and generaton, as well as gude future composers n wrtng musc wth the style of classcal muscans such as Beethoven, Mozart, Bach Although classcal composer dentfcaton s often perceved as reserved to traned muscal ears, an automated system capable of recognzng composers gven raw data s desrable as t suppresses the need for a professonal to perform such task. Recent studes nvolvng self-acclamed experts show a low recognton of 48% [2]. Addtonally, gven the nherently large aspect of musc data, automatng ths task becomes a necessty rather than a luxury. Ths task s the focus of ths paper whch proposes an automated artst recognton system based on a unque framework for feature generaton, feature reducton and supervsed classfcaton.

2 56 Complex Intell. Syst. (2018) 4:55 65 In general, the muscal composton can be ether classfed as a recordng or a symbolc score. A recordng can be n the form of an audo fle or a real-tme sequenced MIDI fle. Generatng musc data as recordngs encodes two coupled sets of nformaton: (a) composer style and (b) performer nterpretaton of that partcular composton. Decouplng the two sets of nformaton s a dfferent problem by tself. Snce ths paper targets the composer recognton problem, the musc data used here are step-by-step sequenced MIDI fles, whch are bascally dgtal symbolc scores obtaned from the Humdrum project lbrary [19]. Ths avods the nterference of the performer nterpretaton durng the composer recognton process snce performance s dfferent than style. After all, a Beethoven composton for nstance can be nterpreted very dfferently by Vladmr Horowtz and the Beatles. The remander of ths paper s organzed such that secton Lterature revew revews prevous related works on composer and performer recognton. Secton Feature extracton llustrates our proposed algorthm for feature generaton usng musc n-grams. Secton Cortcal algorthm for feature reducton and classfcaton descrbes the adaptaton of cortcal algorthms (CA) for feature reducton whle secton Expermental results shows our encouragng expermental results wth a dscusson on captured musc style. Fnally, secton Concluson concludes wth potental follow-on work. Lterature revew Lterature survey shows research work that attempted to recognze composers n an automated manner. In what follows, we summarze the most related ones. One of the technques used to classfy composers s language modelng. The style of each composer s encoded n a language model usng n-grams based on the assumpton that the man features of a musc composer style are captured n some repettons and recurrences, much lke wrtng styles n language modelng. Such an approach was tested on a corpus of fve composers from the Baroque and Classcal perods wth accuracy attanng 95% [17]. N-grams representatons were also used n [35] where three types of n-grams: melodc, rhythmc and combned were constructed. The number of occurrences of each n-gram and a smlarty measure was calculated on some MIDI fles. A grd search usng n = 6 gave the best accuracy of 84% on the testng set. It should be noted that each pece of the sets was chosen to be well sequenced,.e., each track has to represent only one staff or hand. Choosng the peces that satsfy ths crteron was the only human preprocessng task carred out on ths data. However, t s more challengng to get the md fle n a sngle track and perform the analyss accordngly. Tradtonal machne learnng technques have also been appled for the task of composer recognton, among whch we cte neural network archtectures [6,18], for ther smlarty to our recognton model. The system used n [18] s a cascade of a two-layer Stacked Denosng Autoencoder (STA), a two-layer Deep Belef Network and a logstc regresson layer on the top. The STA extracts features from raw data by corruptng the clean data by addng some nose or smply randomly settng some elements to 0, then trans the autoencoder wth the corrupted data as nput and the clean data as target n the end. The DBN s a stack of two layers of Restrcted Boltzmann Machnes amng at creatng a probablstc representaton of the nput whle the regresson layer labels the nput. A recognton rate of 76% was obtaned on an n-house dataset. On the other hand Buzzanca [6], used artfcal shallow neural networks to recognze the style of musc compostons traned usng back-propagaton. Shared weghts as proposed by LeCun et al. [24] allowed for a varable number of nputs whch s a must n ths case because muscal phrases have dfferent lengths. A three-hdden layer network topology was employed and the nput layer was dvded nto sets of four n order to represent the 1/8th note (half beat) wth two nput nodes encodng the ptch. The frst two hdden layers serve as encoders of the rhythm and recognze whether a slent note (rest) or sounded note has been played. 400 peces for dfferent composers were used to tran the network whch took around 20 h to complete. Results showed an accuracy of 97.09%. The sequental aspect of musc can be captured usng Hdden Markov Models. Prevous works adopted ths approach on both audo and symbolc data such as [3,21,25,29]. A combnaton of spectral features such as Spectral Centrod, Rolloff, Flux and Mel-Frequency Cepstral Coeffcents, and content specfc harmones are fed nto a 4-node Markov chan n [3]. A custom database consstng of 400 dstnct audo samples of 30 seconds n length belongng to Bach, Beethoven, Chopn, and Lszt s used to valdate the model wth recognton rates attanng 83%. A varant of Hdden Markov Model, the weghted Markov Chan Model, has also been appled to classfy composers [21]. Structural nformaton from nter onset and ptch ntervals between pars of consecutve notes s used to buld a weghted varaton of a frst-order Markov chan traned usng a proposed evolutonary procedure that automatcally tunes the ntroduced weghts. Expermental results on a collecton of 150 peces by Haydn, Mozart and Beethoven showed recognton rates up to 88% for Beethoven Haydn classfcaton, 70% for Haydn-Mozart classfcaton and 87% for Mozart Beethoven recognton. Lu and Selfrdge-Feld [25] used Markov chans by defnng the state space as the tonalty. Hence, after defnng the repertore of a style, encodng and defnng the state space, the transton matrx can then be calculated and, thus, the Markov Chan s constructed. A dstance measure as proposed by [10] s used to compare an unlabeled nstance to prevously constructed chans

3 Complex Intell. Syst. (2018) 4: and hence dentfy the composer. Expermental results on a collecton of 100 movements of Mozart Strng quartets and 212 movements from Haydn (MIDI fles) showed recognton rates between 52.8 and 68%. Representng musc as states of relatve ptch and rhythm, Pollastr and Smoncell [29] used Hdden Markov Models (HMMs) to determne the style of a composer whch s dentfed based on a hghest probablty scheme. Expermental results based on a dataset ncludng 605 peces for varous artsts ncludng classcal artsts and pop artsts showed accuracy of 42%. Valdaton wth human subjects returned 48% accuracy for expert muscans and 24.6% for amateurs. Other methods modeled artst styles and classfy composers. Mearns et al. [26] employed counterpont as a hgh level feature from muscal theory background to classfy the style of muscal peces based on a symbolc representaton of the score (Kern Format). Counterpont s the study of musc of two or more voces whch sound smultaneously and progress accordng to a set of rules. A classfer based on a Nave Bayes and Decson Tree correctly labeled 44 of the 66 peces used for testng belongng to Renassance and Baroque composers. Geertzen and van Zaanen [9] appled grammatcal nference (GI) to automatcally recognze composer based on learnng recurrng patterns n the muscal peces. An unsupervsed machne learnng technque extensvely used n language recognton, GI reveals underlyng structure of symbolc sequental data. For composer recognton, melody and rhythm as features were used to buld typcal phrases or patterns n an unsupervsed manner and rhythm was employed only to perform supervsed classfcaton. The GI component n ther system was realzed by Algnment-Based Learnng (ABL) [33]. Ths technque generates consttuents based on syntactc smlarty by algnng muscal phrases, then selects the best consttuents that descrbes the structure of the muscal composton. ABL error rates averaged about 21 on a baroque dataset wth preludes from Bach, Chopn and about 24 on a classc dataset wth quartet peces from Beethoven, Haydn and Mozart. Feature extracton In any machne learnng applcaton, raw data are preprocessed n order to create a numercal representaton that can be processed by the recognzer. Ths phase s referred to as feature extracton and t affects tremendously the machne learnng performance [22]. For our partcular applcaton, the detals of ths phase are shown below. Data extracton Startng wth a sngle track raw md fle, the feature extracton stage ams to construct a feature vector representng the md fle. Naturally, the frst step s to comple the data from whch the features are generated. The ultmate goal s to create a representaton of audo fles for the performed musc when recorded usng mcrophones. Ths would requre performng automatc transcrpton,.e., dentfyng the sequence of notes and ther characterstcs such as the dynamcs, rhythms, ptches, etc. However, the ablty to perform automatc transcrpton s stll an mmature feld due to the fact that advanced audtory sgnal processng technques are the focus of researchers worldwde especally for polyphonc musc [1,4,23,30,31]. In addton to that, as we mentoned earler, a recordng of the same muscal composton mght be very dfferent f nterpreted by dfferent performers. Hence to classfy a recordng based solely on the composer style, one has to develop a scheme to deconvolve the nterpretaton of the performer whch becomes a completely dfferent problem. For these reasons, the data set s extracted from MIDI fles (representatons of the sequental progresson of muscal notes and ther characterstcs) usng an avalable Matlab toolbox [8]. The most mportant parameters representng a muscal pece are the ptch and duraton of each note n a md sequence. However, snce the same muscal pece mght be played n a dfferent tonalty or even tempo; therefore, a relatve representaton of the ptch and duraton s preferable compared to the absolute representaton that was proposed n [23]. For the th and ( 1)th notes havng a ptch p and p 1 and a duraton t and t 1, respectvely, we construct a representaton P for the ptch dfference and T for the tme duraton dfference as shown n Eqs. 1 and 2. P = p p ( 1 ) (1) t T = round log 2 (2) Feature generaton t 1 After data extracton, features are generated. These features should characterze the data n a compact and effcent form for our purpose. As suggested n [32], the hghest ptched notes (skylne) reveal the melody of a certan pece, whle the progresson of the base notes (groundlne) gves a good ndcaton of the style of the composer. Therefore, we start frst by extractng the melody and the bass notes and then the correspondng ptch and duraton n-grams as proposed n [23]. However, n addton, we propose another set of features whch characterzes the rhythmc style of the composer. These features quantze the number of notes that have onsets on a complete beat, half of a beat, quarter of a beat, one thrd of a beat, etc. Ths captures the style of the composer n placng notes on beats or between beats (the so called syncope) or even trplets. We ncluded n secton Illustratve examples an llustratve example of our proposed feature generaton.

4 58 Complex Intell. Syst. (2018) 4:55 65 Ptch n-gram Duraton n-gram Md Fle Extreme (Hgh) Extreme (Low) Average Flter (Hgh) Flter (Low) Quantze Quantze Extreme (Hgh) Extreme (Low) Rhythm Vector Ptch n-gram Feature Vector Duraton n-gram Rhythm Vector Fg. 1 Block dagram for feature generaton Fgure 1 shows a detaled setup for the feature generaton stage from a md fle. Frst, the note matrx (see Table 1) s constructed, and then the hghest and lowest note sequences are extracted. The note sequences are then fltered based on a computed threshold defned as the average of the centrods of the hghest and lowest note sequences. The process s repeated resultng n a representaton of the pece consstng of a robust melody and bass extracton. The repetton of the process s partcularly useful to categorze a solo note sequence as melody or bass. The fnal feature vector for the nput md fle s a vector contanng occurrences of N possble n-grams n addton to the onset beat features we propose. It s mportant to note that N vares wth each md fle. Hence to obtan a fxed length feature vector, automatc zero paddng s employed where non-exstng n-grams are added wth 0 occurrences. Illustratve examples In ths secton, we frst llustrate the performance of the melody/bass extracton procedure. Then, we show how the n-grams are constructed. To show the approach of our melody/bass extracton algorthm, we use (as an llustratve example) the frst seventy beats of Bach s Prelude No. 17 n A flat major. Fgure 2 shows the pano rolls of Bach s Prelude before and after melody/bass separaton. Next, we llustrate wth a smple example the constructon of n-grams. We use an evaluaton copy of Mozart Software [34] to wrte the man melody of the Lebanese Natonal Anthem and convert t to a MIDI fle. The man melody of the frst phrase s shown n Fg. 3 and the correspondng data Table 1 Note matrx of the sample n Fg. 3 Onset (beats) Md ptch Duraton (s) extracted also known as the note matrx are shown n Table 1. Onsets and Duratons (whether n beats or n seconds) gve rhythm characterstcs, ptches represent the name of the note (60 represents the central C note) and velocty represents the dynamcs of the note. The data are generated as follows: ungrams are constructed usng Eqs. 1 and 2. Then, ptch and duraton n-grams extracted are constructed. For example, 3- grams are formed for the sample mentoned. Ptch (duraton) ungrams and 3-grams of the sample are shown n Tables 2 and 3, respectvely. Fnally, the rhythm feature vector for the sample s constructed and shown n Table 4.

5 Complex Intell. Syst. (2018) 4: Fg. 2 Pano rolls of Bach s Prelude No. 17 n A flat major. The top fgure shows the pano roll of the full composton. The bottom fgure shows the pano roll of the separated melody (red) and bass (blue) Fg. 3 Sample of the Lebanese natonal anthem produced by an evaluaton copy of Mozart Table 2 Ptch 3-grams Absolute ptch Relatve ptch 3-grams Occurrence Table 3 Duraton 3-grams Absolute duraton Relatve duraton 3-grams Occurrence Cortcal algorthm for feature reducton and classfcaton The large number of n-grams possble combnatons leads to a very large feature vector sze. As a result, a feature reducton scheme s hghly desrable to buld a robust ML model. In ths secton, we provde a mathematcal formulaton for CA as ntroduced n lterature, as well as for our proposed feature reducton scheme.

6 60 Complex Intell. Syst. (2018) 4:55 65 Table 4 Rhythm features Rhythm features Complete beats 10 Half beats 0 Quarter beats 4 1/8th beats 0 1/3rd beats 2 Prmer on cortcal algorthms Occurrence CA are a bologcally nspred machne learnng (ML) approach whch structure (columnar organzaton) has been proposed by Mountcastle and Edelman [7]. The model mmcs aspects of the human vsual cortex that recalls sequences stored herarchcally n an nvarant. A mathematcal formulaton and a computatonal mplementaton of ths model were further developed by Hashm and Lpast [13 16]. The model conssts of an assocaton of sx-layered structures of dfferent thckness consttuted of a very large number of columns strongly connected va feed-forward and feedback connectons. The column, consdered as the basc unt n a cortcal network, s consttuted by a group of neurons that share the same nput. An assocaton of columns s referred to as hyper-column or layer. The connectons n the network occur n two drectons: vertcal connectons between columns of consecutve layers, and horzontal connectons between columns wthn the same layer. Though CA can be consdered as a type of deep neural networks (DNNs) due to ts hexa-layered archtecture, two man dfferences wth DNNs can be noted at the levels of structure and tranng: tradtonal DNNs employ a feedforward archtecture whle feedback connectons are adopted n CA, and whle DNNs rely on a backpropagaton algorthm for ther tranng, CA ams at formng unque representatons n ts supervsed learnng scheme [12]. A mathematcal representaton for CA, as we developed based on the descrpton provded n [7], s detaled below. The nomenclature we adopt n our work s as follows:, j,k represents the weght of the connecton between the jth neuron of the th column of layer r and the kth column of the prevous layer (r 1) durng the tranng epoch t.for the rest of ths paper, scalar enttes are represented n talc, vector enttes n bold, whle underlned varables represent matrces [11]. The weght matrx representng the state of a column composed of M nodes at the tranng epoch t s [ =, j,1,,2 ] r,t r,t,...,w, j,...,w,m represents the weght vector of the connectons enterng neuron j, of column n layer r whch can be wrtten as: (3) [ ] T, j =, j,1, r,t r,t, j,2,...,w, j,k,...,w, j,l r 1 (4) L r 1 s the number of columns n the layer (r 1) and the subscrpt T stands for the transpose operator. Expandng yelds: =,1,1..., j,1...,m, ,1,k..., j,k...,m,k ,1,l r 1..., j,l r 1...,M,L r 1 Defnng Z r,t as the output vector of layer r for epoch t and the output of column, we can wrte Z r,t Z r,t = [ Z r,t 1, Zr,t 2,...,Zr,t (5),...,Z r,t L r ] T (6) The output of a neuron s the result of the nonlnear actvaton functon f (.) n response to the weghted sum of the connectons enterng the neuron. The output of a mn-column s defned as the sum of the outputs of the neurons n ths column: Z r,t = Z r, j M j=1 Z r,t, j ; Zr,t, j = f ( k, j,k Zr 1,t k s the output of the jth neuron consttutng the th column of the rth layer and the actvaton functon s defned as follows: 1 f (ν) = { 1+exp [ ν (φ(ν)+τ)] 2 fν = 1 (8) φ(ν) = ν otherwse Note that Z 0 refers to the nput vector X =[X 1,...,X D ] and we denote by D the sze of the nput vector whch corresponds to the number of features or the data dmensonalty. τ s a tolerance parameter emprcally selected and t s assumed constant for all epochs and columns. Ths parameter sets the tolerance of the actvaton functon to large nput values, the larger ths value s the less senstve the actvaton functon becomes. The tranng of a cortcal network has three phases: 1. Random ntalzaton: Intally, all the synaptc weghs are ntalzed to random values that are very close to 0, so that no preference for any partcular pattern s shown. ) (7)

7 Complex Intell. Syst. (2018) 4: Unsupervsed feed-forward: The feedforward learnng has the objectve of tranng mncolumns to dentfy features not detected by others. When the random frng of a partcular column concdes wth a partcular nput pattern, ths actvaton s enforced, through strengthenng of the column weghts and nhbtng of neghborng columns. The nhbtng and strengthenng weghts update rules are shown n Eqs. 9 and 10, respectvely. ( +1, j,k = Z r 1,t k, j,k ( ) ) (9) Equaton 9 suggests that when a node s nhbted, a value that s proportonal to ts nput Z r 1,t k (whch corresponds to the output of the column connected to t) s subtracted from the correspondng weght value. +1, j,k = Zr 1,t k, j,k + Cr,t where ( ) s gven by: ( ) = M = L r 1 j=1 k=1, j,k + ρ C r,t, j,k, j,k ; Cr,t, j,k { 1 f W r,t, j,k >ɛ 0 otherwse 1 + exp (, j,k τ ) (10) where ρ s a tunng parameter determnng the amount of strengthenng to add to a partcular weght, the larger ρ, the stronger the strengthenng. Smaller values of ρ may lead to a lengther tranng as mnmal change s obtaned durng strengthenng whle large values may drve the network towards nstablty. ɛ s emprcally chosen as the frng threshold assumed constant for all epochs and columns. Smlarly, Eq. 10 shows that strengthenng a connecton occurs by addng a value proportonal to the nput weght and n the form of an exponental functon of the weght tself. Wth repeated exposure, the network extracts features of the nput data, by tranng the columns to fre for specfc patterns. 3. Supervsed feedback: The supervsed feedback has a goal to correct msclassfcatons of the same pattern. Another varaton of the same pattern that s qute dfferent from the prevous one may lead to a msclassfcaton when the top layer columns that are supposed to fre for that pattern do not fre. To reach a desred frng scheme over multple exposures also known as stable actvaton, the error occurrng at the top layer generates a feedback sgnal forcng the column frng for the orgnal pattern to fre and also nhbtng the column that s frng for the new varaton. The feedback sgnal s propagated back to the prevous layer once the top level columns start to gve a stable actvaton for all varatons of the same pattern. The feedback sgnal s sent to precedng layers only once the error n the layer concerned converges to a value below a certan pre-defned tolerance threshold. Each layer s traned untl convergence crtera expressed as an error term n functon of the actual output and a desred output (scheme of frng) s reached. Cortcal algorthm for feature reducton One of the advantages of CA s ts ablty to extract and learn the aspects of the data n a herarchcal manner. Ths makes CA a possble canddate for feature extracton or reducton, whch to the best of our knowledge has not been nvestgated yet [22]. As descrbed prevously, one goal of CA tranng s to tran cortcal columns to fre for partcular aspects of the patterns durng the feedforward unsupervsed learnng. In addton, durng the feedback supervsed learnng, nvarant representatons are created on all levels of the network n a herarchcal manner. As a result, a repettve frng scheme s observed for a partcular class of the data. The weghts of the nput layer drectly connected to the feature vector are, therefore, an ndcaton of the sgnfcance of the correspondng features. Our method employs hereby denoted by CA-FR, the nput layer weghts to rank the feature vector n terms of sgnfcance, dscardng the lowest ranked features, allowng for a hgher dscrmnatve performance of the algorthm and a robust model of the raw data. The followng algorthm llustrates our approach hereby denoted by CA-FR: 1. Tran the cortcal network usng all features provded. 2. For all nput nodes calculate average weght assocated wth each feature. 3. Rank features n a decreasng order accordng to ther assocated weghts. 4. Elmnate a pre-defned percentage of the lowest ranked features (heurstcally chosen). Expermental results Dataset In ths secton, we summarze the expermental results obtaned. We perform our experments on a database of note fles we compled from the Humdrum project lbrary [19] descrbed n Table 5. The era of each composer as well as the total number of beats for each composer are also shown n ths table.

8 62 Complex Intell. Syst. (2018) 4:55 65 Table 5 Dataset descrpton Composer Area No. of fles Total sze (MB) Total number of beats Bach Baroque ,259 Beethoven Classcal-romantc ,485 Chopn Romantc ,355 Corell Baroque ,009 Haydn Classcal ,763 Jopln Ragtme ,043 Mozart Classcal ,912 Scarlatt Baroque ,849 Vvald Baroque ,544 Experments Our experments are desgned to test the features we proposed, our feature reducton technque, as well as CA as a classfcaton technque for the recognton of artsts. Input weghts of the network were used to dscern the least sgnfcant features (correspondng to the weakest nput weghts). A fracton of these features s elmnated at each teraton. Thus, we perform four experments usng our dataset: Experment 1: Usng our proposed feature vector, and our feature reducton scheme (CA-FR). Experment 2: Usng only the n-grams features (as wdely proposed n the lterature) and our feature reducton algorthm (CA-FR). Experment 3: Usng the entre feature vector proposed and a prncpal component analyss (PCA) for feature reducton. Experment 4: Usng only the n-grams features and a PCA for feature reducton. For each of the four experments, we subsequently elmnate a fracton of the least sgnfcant features. The resultng subsets are fed to three recognzers: CA, SVM one vs. all usng Gaussan Radal Bass Functon (RBF) kernel and a one nearest neghbor classfer (1NN). SVM and 1NN were chosen for comparson purposes snce they are wdely adopted n a multtude of machne learnng applcatons [27]. We evaluate the performance of these classfers n terms of recognton rate. We used a cortcal network consttuted of 6 layers startng wth 2000 columns n the frst layer and decreasng ths number by half between consecutve layers, each column consstng of 20 nodes. Ths structure was expermentally proven to perform well on a varety of databases [11]. Smulatons were executed usng Matlab R2011a, on an Intel core 7 CPU at 2 GHz under Wndows 7 operatng systems. The CA network s deployed usng the CNS lbrary, a framework for smulatng cortcally organzed networks, developed by MIT [28]. Table 7 shows the recognton rates obtaned usng the three classfers descrbed above for dfferent percentages of the ntal dataset reduced usng PCA and CA-FR as descrbed n the expermental setup, whle Table 8 shows the requred tranng tme. The best recognton rate for each experment s shown n bold. All results are based on a fourfold cross valdaton scheme. Table 9 shows the recognton rate obtaned usng experment 1 settngs on 0.1% of the features for a 2 two-composer classfcaton task. Composers are arranged accordng to ther year of brth to reflect ther closeness n eras. In addton, Table 10 provdes the mean, standard devaton, maxmum and mnmum recognton rate obtaned for every experment (for all reducton rates) as well as the overall statstc for every classfer. Table 6 comples results from recent work on these composers. The followng can be concluded from our expermental results. Usng the entre dataset wth the proposed added features, the best recognton rate acheved s 88.4% usng CA as classfer compared to 70.3 and 60.6% usng SVM and NN, respectvely, and thus an mprovement of CA of 18.1 and 27.8% over SVM and NN, respectvely. Usng only 0.1% of the features ncludng the proposed attrbutes, the best performance acheved s 94.9% usng CA as a classfer, 72.8% usng SVM and 62.2% usng NN. Ths s a strong ndcator of the superor performance of CA s classfcaton compared to classcal pattern recognton classfers. For the same reducton rate of 99.9%, our proposed added features ntroduced an mprovement from 86.4 to 90.4% usng PCA and from 91.1 to 94.9% usng CA-FR, showng the sgnfcance of our proposed added features n dscrmnatng composers. In other terms, an mprovement of 4% s obtaned usng our proposed features and PCA for feature reducton (Exp. 3 vs. Exp. 4) and 3.8% usng our proposed features and CA-FR for feature reducton (Exp.1 vs. Exp. 2) For the same percentage of features (0.1%), the best recognton rate obtaned s 94.9% usng CA as classfer and our feature reducton algorthm whle PCA led to an accuracy of 90.4% for the same reducton rate. Ths shows CA-FR s

9 Complex Intell. Syst. (2018) 4: Table 6 State of the art results Reference Task Recognton rate [17] Bach vs. Haydn 96.8 Bach vs. Mozart 97.5 Haydn vs. Mozart 74.7 [20] Bach vs. Beethoven 81.2 Bach vs. Chopn 87.0 Bach vs. Haydn 69.3 Bach vs. Mozart 82.2 Beethoven vs. Chopn 63.8 Beethoven vs. Haydn 65.0 Beethoven vs. Mozart 69.8 Chopn vs. Haydn 77.8 Chopn vs. Mozart 77.0 Haydn vs. Mozart 55.8 [18] Bach vs. not Bach 93.1 Beethoven vs. not Beethoven 63.3 Chopn vs. not Chopn Haydn vs. not Haydn 40 Mozart vs. not Mozart 74.6 Vvald vs. not Vvald 87.0 Table 7 Recognton rates comparson Exp. Classfer % of features Exp. 1 CA SVM NN Exp. 2 CA SVM NN Exp. 3 CA SVM NN Exp. 4 CA SVM NN Bold values ndcate best values ablty n extractng the most sgnfcant features whle dscardng the less mportant ones. Usng only the n-grams features and our feature reducton technque resulted n a slghtly better performance compared wth usng PCA on the entre set, ndcatng the strong capacty of CA-FR of extractng meanngful nformaton and creatng dscrmnatve representatons of the nput data. Comparng feature reducton technques, the maxmal mprovement of CA-FR vs. PCA s of 4.9% (94.4 vs. 90.4%) observed on the entre set and 3.3% (91.1 vs. 87.8%) on Table 8 Tranng tme comparson Exp. Classfer % of features Exp. 1 CA SVM Exp. 2 CA SVM Exp 3 CA SVM Exp. 4 CA SVM the n-grams attrbutes. In average, usng CA-FR for feature reducton resulted n an accuracy of 88.4% compared to 84.2% usng PCA,.e., an mprovement of 4.2%. Evaluatng the sgnfcance of the proposed features, the maxmal accuracy attaned usng only n-gram features s 91.8% compared to 94.9% usng our proposed feature vector. The effect of the feature reducton stage s most shown n the classfcaton results: an mprovement n accuracy s overall notced n experments employng our feature reducton algorthm, a smaller effect of feature reducton can be seen for PCA whch shows the robustness of our algorthm n dscernng sgnfcant features. Whle a nearest neghbor classfer requres practcally zero tranng tme, the hgh computatonal demand of computng dstances (specally for hgh dmensonal data) renders ts use n an onlne scheme qute challengng. Conversely whle CA and SVM requre lengthy tranng, the testng complexty of both these algorthms s relatvely low (SVM only requres the computaton of the equaton of a lne whle CA requres only the propagaton of the nput through the network) and hence are more sutable for an onlne framework whle the tranng can be performed offlne. Comparng our proposed method wth the state of the art results shown n Table 6, a notable mprovement n recognton rate can be observed n all one-to-one tasks. Ths s further ntensfed n tasks nvolvng composers wthn the same era such as Haydn and Mozart where the recognton rate saw an ncrease from 74.7 to 82.9% as well as Beethoven and Chopn where our algorthm acheved 86.1% compared to 63.8% reported n the state of the art. Comparng classfers, one can see the hgher ablty of CA to robustly classfy composers by analyzng Table 10: whle CA s hgher standard devaton may be regarded as a drawback, ts lowest recognton rates rval the best performances of SVM and 1NN consstently for all experments, hence justfyng ts use as a classfer. Comparng the recognton rates observed for one-to-one composer classfcaton tasks, one can see the robustness

10 64 Complex Intell. Syst. (2018) 4:55 65 Table 9 Recognton rates for one-to-one composer classfcaton Corell Vvald Bach Scarlatt Haydn Mozart Beethoven Chopn Jopln Corell (1653) Vvald (1678) Bach (1685) Scarlatt (1685) Haydn (1732) Mozart (1756) Beethoven (1770) Chopn (1810) 98.7 Jopln (1868) Table 10 Statstcal comparson of experments Metrc Exp. 1 Exp. 2 Exp. 3 Exp. 4 CA SVM 1NN CA SVM 1NN CA SVM 1NN CA SVM 1NN Mean Std Max Mn of our proposed method: the lowest reported accuracy s at 86.1% for a Chopn vs. Beethoven task compared wth a 63.8% reported n [20], whle the best performance marked a 99.5% accuracy classfyng Vvald vs. Jopln. Overall, our proposed feature set and feature reducton system showed to be powerful for the recognton of musc composers wth a 94.4% recognton rate obtaned usng 0.1% of the entre features (n-grams and proposed attrbutes), CA for feature reducton and classfcaton. From a muscal perspectve A deeper look on our proposed framework can be taken by nvestgatng the retaned features for the best performance obtaned,.e., for Exp. 1 at 99.9% reducton. As ndcated by the nput weghts of the network, the fve most sgnfcant features are the ones correspondng to the occurrences of notes wth onset on quarter beats, occurrences of notes wth onset on complete beats, occurrence of notes wth onset on half beats, and 2 n-gram combnatons correspondng to bass notes; all of whch are proposed by ths work. From a musc perspectve, these results propose that pano composers mostly dstngush ther style va the way they deal wth bass notes on one hand, and beats/onsets on the other hand. Obvously, melodes play an mportant role n revealng the style of a partcular composer. However, accordng to our work, the spreadng of notes over beats/onsets, together wth the development of bass patterns that accompany the melody, has a lot more to say about the style of a partcular composer. It s well known that humans, n general, are based to perceve hgh ptched melodes much more than the accompaned bass notes. However, gven any melody, dfferent composers develop dfferent beats and dfferent bass notes whch harmonze wth the same melody. Ths s partcularly what dstngushes the style of a composer as the retaned most sgnfcant features propose. Furthermore, lookng at the results of one-to-one classfcaton (Table 9), confuson rates ranged from 17.1% (Mozart vs. Haydn) to 0.5% (Vvald vs. Jopln) wth Jopln beng the least confused composer, due to hs belongng to a dfferent era than all other composers. The least accurate predcton rates were found for the tasks of dstngushng Haydn vs. Mozart, and Beethoven vs. Chopn, consstent wth the smlartes n style between the mentoned composers. In fact, t s known n the musc socety that Haydn and Mozart were frends wth the latter beng nspred by the former [5]. Addtonally, whle both Beethoven and Chopn have ther own styles, some peces of these composers possess some smlar features (partcularly, Beethoven s Adago of op. 106 or the Aroso n op. 110 has a chromatc style, a feature Chopn has been known for). Concluson In ths paper, we have presented an automated dentfcaton framework for the recognton of classcal musc composers usng a novel n-gram based feature extracton technque along wth a cortcal algorthm for the feature reducton and recognton stages. Our proposed system s a power-

11 Complex Intell. Syst. (2018) 4: ful scheme nhertng the robustness of feature reducton usng CA and the dscrmnatve capacty of our proposed feature. Expermental results show that wth only 0.1% of the feature set as dentfed by CA, our proposed approach acheved a recognton rate of 94.4% on a dataset of 1197 peces belongng to 9 composers whch motvate testng on larger datasets. Acknowledgements Ths work s funded by the Unversty Board at the Amercan Unversty of Berut. Open Access Ths artcle s dstrbuted under the terms of the Creatve Commons Attrbuton 4.0 Internatonal Lcense ( ons.org/lcenses/by/4.0/), whch permts unrestrcted use, dstrbuton, and reproducton n any medum, provded you gve approprate credt to the orgnal author(s) and the source, provde a lnk to the Creatve Commons lcense, and ndcate f changes were made. References 1. Abdallah SA, Plumbley MD (2003) An ndependent component analyss approach to automatc musc transcrpton. In: Proceedngs of the 114th AES conventon, Amsterdam, The Netherlands, March Abeler N (2015) Muscal composer dentfcaton MUS Bartle A (2012) Composer dentfcaton of dgtal audo modelng content specfc features through Markov models 4. Bello JP, Mont G, Sandler MB et al (2000) Technques for automatc musc transcrpton. In: Proceedngs of the frst nternatonal symposum on musc nformaton retreval (ISMIR-00), Plymouth, Massachusetts, USA 5. Brown P (1992) Haydn and Mozart s 1773 stay n Venna: weedng a muscologcal garden. J Muscol 10: Buzzanca G (2002) A supervsed learnng approach to muscal style recognton. In: Musc and artfcal ntellgence. Addtonal proceedngs of the second nternatonal conference, ICMAI, vol 2002, p Edelman GM, Mountcastle VB (1978) The mndful bran: cortcal organzaton and the group-selectve theory of hgher bran functon. Massachusetts Insttute of Technology, Cambrdge 8. Eerola T, Tovanen P (2004) Md toolbox: Matlab tools for musc research. Unversty of Jyväskylä, Jyväskylä, Fnland, p Geertzen J, van Zaanen M (2008) Composer classfcaton usng grammatcal nference. In: Proceedngs of the MML 2008 nternatonal workshop on machne learnng and musc held n conjuncton wth ICML/COLT/UAI, pp Grachten M, Wdmer G (2009) Who s who n the end? recognzng pansts by ther fnal rtardand. In: ISMIR. Cteseer, pp Hajj N, Awad M (2013) Weghted entropy cortcal algorthms for solated Arabc speech recognton. In: The 2013 nternatonal jont conference on neural networks (IJCNN). IEEE, pp Hajj N, Rzk Y, Awad M (2015) A mapreduce cortcal algorthms mplementaton for unsupervsed learnng of bg data. Proc Comput Sc 53: Hashm A, Lpast MH (2010) Dscoverng cortcal algorthms. In: Internatonal jont conference on computatonal ntellgence (Internatonal conference on fuzzy computaton and nternatonal conference on neural computaton), pp Hashm A, Berry H, Temam O, Lpast M (2011) Automatc abstracton and fault tolerance n cortcal mcroachtectures. In: ACM SIGARCH computer archtecture news, vol 39. ACM, pp Hashm A, Nere A, Thomas JJ, Lpast M (2011) A case for neuromorphc ISAs. ACM SIGARCH Comput Arch News 39(1): Hashm AG, Lpast MH (2009) Cortcal columns: buldng blocks for ntellgent systems. In: IEEE symposum on computatonal ntellgence for multmeda sgnal and vson processng, CIMSVP 09. IEEE, pp Hontanlla M, Pérez-Sancho C, Inesta JM (2013) Modelng muscal style wth language models for composer recognton. In: Pattern recognton and mage analyss. Sprnger, Berln, pp Hu Z, Fu K, Zhang C (2013) Audo classcal composer dentfcaton by deep neural network. arxv: Huron D (1997) Humdrum and kern: selectve feature encodng. In: Beyond MIDI. MIT Press, Cambrdge, pp Kalakatsos-Papakostas MA, Eptropaks MG, Vrahats MN (2010) Muscal composer dentfcaton through probablstc and feedforward neural networks. In: European conference on the applcatons of evolutonary computaton. Sprnger, New York, pp Kalakatsos-Papakostas MA, Eptropaks MG, Vrahats MN (2011) Weghted Markov chan model for muscal composer dentfcaton. In: Applcatons of evolutonary computaton. Sprnger, New York, pp Khanna R, Awad M (2015) Effcent learnng machnes: theores, concepts, and applcatons for engneers and system desgners. Apress, New York 23. Klapur AP (2004) Automatc musc transcrpton as we know t today. J New Musc Res 33(3): LeCun Y et al (1989) Generalzaton and network desgn strateges. Connectons n perspectve. North-Holland, Amsterdam 25. Lu Y-W, Selfrdge-Feld E (2002) Modelng musc as Markov chans: composer dentfcaton. edu/~jacoblu/254report/ 26. Mearns L, Tdhar D, Dxon S (2010) Charactersaton of composer style usng hgh-level muscal features. In: Proceedngs of 3rd nternatonal workshop on machne learnng and musc. ACM, pp Mchalsk RS, Carbonell JG, Mtchell TM (2013) Machne learnng: an artfcal ntellgence approach. Sprnger Scence & Busness Meda, New York 28. Mutch J, Knoblch U, Poggo T (2010). CNS: a GPU-based framework for smulatng cortcally-organzed networks. Massachusetts Insttute of Technology, Cambrdge, Tech. Rep. MIT-CSAIL-TR /CBCL Pollastr E, Smoncell G (2001) Classfcaton of melodes by composer wth hdden Markov models. In: Proceedngs of frst nternatonal conference on web delverng of musc. IEEE, pp Raphael C (2002) Automatc transcrpton of pano musc. In: Proceedngs of the nternatonal conference on musc nformaton retreval (ISMIR), Pars, France 31. Ryynänen MP, Klapur A (2005) Polyphonc musc transcrpton usng note event modelng. In: IEEE workshop on applcatons of sgnal processng to audo and acoustcs. IEEE, pp Utdenbogerd A, Zobel J (1999) Melodc matchng technques for large musc databases. In: Proceedngs of the seventh ACM nternatonal conference on multmeda (Part 1). ACM, pp Van Zaanen MM (2002) Bootstrappng structure nto language: algnment-based learnng. arxv:cs/ Webber D (2015) Mozart musc software. uk/ 35. Wołkowcz J, Kulka Z, Kešelj V (2008) N-gram-based approach to composer recognton. Arch Acoust 33(1):43 55

Following a musical performance from a partially specified score.

Following a musical performance from a partially specified score. Followng a muscal performance from a partally specfed score. Bryan Pardo and Wllam P. Brmngham Artfcal Intellgence Laboratory Electrcal Engneerng and Computer Scence Dept. and School of Musc The Unversty

More information

LOW-COMPLEXITY VIDEO ENCODER FOR SMART EYES BASED ON UNDERDETERMINED BLIND SIGNAL SEPARATION

LOW-COMPLEXITY VIDEO ENCODER FOR SMART EYES BASED ON UNDERDETERMINED BLIND SIGNAL SEPARATION LOW-COMPLEXITY VIDEO ENCODER FOR SMART EYES BASED ON UNDERDETERMINED BLIND SIGNAL SEPARATION Jng Lu, Fe Qao *, Zhjan Ou and Huazhong Yang Department of Electronc Engneerng, Tsnghua Unversty ABSTRACT Ths

More information

Instructions for Contributors to the International Journal of Microwave and Wireless Technologies

Instructions for Contributors to the International Journal of Microwave and Wireless Technologies Instructons for Contrbutors to the Internatonal Journal of Mcrowave and Wreless Technologes Frst A. Author 1, Second Author 1,2, Thrd Author 2 1 Cambrdge Unversty Press, Ednburgh Buldng, Shaftesbury Road,

More information

A STUDY OF TRUMPET ENVELOPES

A STUDY OF TRUMPET ENVELOPES A STUDY OF TRUMPET ENVELOPES Roger B. Dannenberg, Hank Pellern, and Istvan Dereny School of Computer Scence, Carnege Mellon Unversty Pttsburgh, PA 15213 USA rbd@cs.cmu.edu, hank.pellern@andrew.cmu.edu,

More information

tj tj D... '4,... ::=~--lj c;;j _ ASPA: Automatic speech-pause analyzer* t> ,. "",. : : :::: :1'NTmAC' I

tj tj D... '4,... ::=~--lj c;;j _ ASPA: Automatic speech-pause analyzer* t> ,. ,. : : :::: :1'NTmAC' I ASPA: Automatc speech-pause analyzer* D. GERVERt and G. DNELEY Unversty of Durham, Durham, England ASPA: The Programs Snce the actual detals of nterface samplng, dsk storage routnes, etc., wll depend upon

More information

Hybrid Transcoding for QoS Adaptive Video-on-Demand Services

Hybrid Transcoding for QoS Adaptive Video-on-Demand Services 732 IEEE Transactons on Consumer Electroncs, Vol. 50, No. 2, MAY 2004 Hybrd Transcodng for QoS Adaptve Vdeo-on-Demand Servces Ilhoon Shn and Kern Koh Abstract Transcodng s a core technque that s used n

More information

System of Automatic Chinese Webpage Summarization Based on The Random Walk Algorithm of Dynamic Programming

System of Automatic Chinese Webpage Summarization Based on The Random Walk Algorithm of Dynamic Programming Send Orders for Reprnts to reprnts@benthamscence.ae The Open Cybernetcs & Systemcs Journal, 205, 9, 35-322 35 Open Access System of Automatc Chnese Webpage Summarzaton Based on The Random Walk Algorthm

More information

RIAM Local Centre Woodwind, Brass & Percussion Syllabus

RIAM Local Centre Woodwind, Brass & Percussion Syllabus 8 RIAM Local Centre Woodwnd, Brass & Percusson Syllabus 2015-2018 AURAL REQUIREMENTS AND THEORETICAL QUESTIONS REVISED FOR ALL PRACTICAL SUBJECTS AURAL TESTS From Elementary to Grade V ths area s worth

More information

Decision Support by Interval SMART/SWING Incorporating. Imprecision into SMART and SWING Methods

Decision Support by Interval SMART/SWING Incorporating. Imprecision into SMART and SWING Methods Decson Support by Interval SMART/SWING Incorporatng Imprecson nto SMART and SWING Methods Abstract: Interval judgments are a way of handlng preferental and nformatonal mprecson n multcrtera decson analyss.

More information

Modeling Form for On-line Following of Musical Performances

Modeling Form for On-line Following of Musical Performances Modelng Form for On-lne Followng of Muscal Performances Bryan Pardo 1 and Wllam Brmngham 2 1 Computer Scence Department, Northwestern Unversty, 1890 Maple Ave, Evanston, IL 60201 2 Department of Math and

More information

SONG STRUCTURE IDENTIFICATION OF JAVANESE GAMELAN MUSIC BASED ON ANALYSIS OF PERIODICITY DISTRIBUTION

SONG STRUCTURE IDENTIFICATION OF JAVANESE GAMELAN MUSIC BASED ON ANALYSIS OF PERIODICITY DISTRIBUTION SOG STRUCTURE IDETIFICATIO OF JAVAESE GAMELA MUSIC BASED O AALYSIS OF PERIODICITY DISTRIBUTIO D. P. WULADARI, Y. K. SUPRAPTO, 3 M. H. PUROMO,,3 Insttut Teknolog Sepuluh opember, Department of Electrcal

More information

arxiv: v1 [cs.cl] 12 Sep 2018

arxiv: v1 [cs.cl] 12 Sep 2018 Powered by TCPDF (www.tcpdf.org) Neural Melody Composton from Lyrcs Hangbo Bao, Shaohan Huang 2, Furu We 2, Le Cu 2, Yu Wu 3, Chuanq Tan 3, Songhao Pao, Mng Zhou 2 School of Computer Scence, Harbn Insttute

More information

The UCD community has made this article openly available. Please share how this access benefits you. Your story matters!

The UCD community has made this article openly available. Please share how this access benefits you. Your story matters! Provded by the author(s) and Unversty College Dubln Lbrary n accordance wth publsher polces., Please cte the publshed verson when avalable. tle Dynamc Complexty Scalng for Real-me H.264/AVC Vdeo Encodng

More information

Optimized PMU placement by combining topological approach and system dynamics aspects

Optimized PMU placement by combining topological approach and system dynamics aspects Optmzed PU placement by combnng topologcal approach and system dynamcs aspects Jonas Prommetta, Jakob Schndler, Johann Jaeger Insttute of Electrcal Energy Systems Fredrch-Alexander-Unversty Erlangen-Nuremberg

More information

Accepted Manuscript. An improved artificial bee colony algorithm for flexible job-shop scheduling problem with fuzzy processing time

Accepted Manuscript. An improved artificial bee colony algorithm for flexible job-shop scheduling problem with fuzzy processing time Accepted Manuscrpt An mproved artfcal bee colony algorthm for flexble ob-shop schedulng problem wth fuzzy processng tme Ka Zhou Gao, Ponnuthura Nagaratnam Suganthan, Quan Ke Pan, Tay Jn Chua, Chn Soon

More information

Error Concealment Aware Rate Shaping for Wireless Video Transport 1

Error Concealment Aware Rate Shaping for Wireless Video Transport 1 Error Concealment Aware Rate Shapng for Wreless Vdeo Transport 1 Trsta Pe-chun Chen and Tsuhan Chen 2 Abstract Streamng of vdeo, whch s both source- and channel- coded, over wreless networks faces the

More information

A Quantization-Friendly Separable Convolution for MobileNets

A Quantization-Friendly Separable Convolution for MobileNets arxv:1803.08607v1 [cs.cv] 22 Mar 2018 A Quantzaton-Frendly Separable for MobleNets Abstract Tao Sheng tsheng@qt.qualcomm.com Xaopeng Zhang parker.zhang@gmal.com As deep learnng (DL) s beng rapdly pushed

More information

Technical Information

Technical Information CHEMCUT Techncal Informaton CORPORATION Introducton The Chemcut CC8000 etcher has many new features desgned to reduce the cost of manufacturng and, just as mportantly, the cost of ownershp. Keepng the

More information

Why Take Notes? Use the Whiteboard Capture System

Why Take Notes? Use the Whiteboard Capture System Why Take Notes? Use the Whteboard Capture System L-we He Zhengyou Zhang and Zcheng Lu September, 2002 Techncal Report MSR-TR-2002-89 Mcrosoft Research Mcrosoft Corporaton One Mcrosoft Way Redmond, WA 98052

More information

Novel Quantization Strategies for Linear Prediction with Guarantees

Novel Quantization Strategies for Linear Prediction with Guarantees Smon S. Du* Ychong Xu* Yuan L Hongyang Zhang Aart Sngh Pulkt Grover Carnege Mellon Unversty, Pttsburgh, PA 15213, USA *: Contrbute equally. SSDU@CS.CMU.EDU YICHONGX@CS.CMU.EDU LIYUANCHRISTY@GMAIL.COM HONGYANZ@CS.CMU.EDU

More information

A Comparative Analysis of Disk Scheduling Policies

A Comparative Analysis of Disk Scheduling Policies A Comparatve Analyss of Dsk Schedulng Polces Toby J. Teorey and Tad B. Pnkerton Unversty of Wsconsn* Fve well-known schedulng polces for movable head dsks are compared usng the performance crtera of expected

More information

QUICK START GUIDE v0.98

QUICK START GUIDE v0.98 QUICK START GUIDE v0.98 QUICK HELP Q A 1 STEP BY STEP 3 GLOSSARY 2 A B C 1 INSTALLATION 1. Make sure that the hardware nstallaton s performed by a certfed vendor 2. Install OTOTRAK app from Apple s App

More information

Anchor Box Optimization for Object Detection

Anchor Box Optimization for Object Detection Anchor Box Optmzaton for Object Detecton Yuany Zhong 1, Janfeng Wang 2, Jan Peng 1, and Le Zhang 2 1 Unversty of Illnos at Urbana-Champagn 2 Mcrosoft Research 1 {yuanyz2, janpeng}@llnos.edu, 2 {janfw,

More information

current activity shows on the top right corner in green. The steps appear in yellow

current activity shows on the top right corner in green. The steps appear in yellow Browzwear Tutorals Tutoral ntroducton Ths tutoral leads you through the best practces of color ways operatons usng an llustrated step by step approach. Each slde shows the actual applcaton at the stage

More information

AN INTERACTIVE APPROACH FOR MULTI-CRITERIA SORTING PROBLEMS

AN INTERACTIVE APPROACH FOR MULTI-CRITERIA SORTING PROBLEMS AN INTERACTIVE APPROACH FOR MULTI-CRITERIA SORTING PROBLEMS A THESIS SUBMITTED TO THE GRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCES OF MIDDLE EAST TECHNICAL UNIVERSITY BY BURAK KESER IN PARTIAL FULFILLMENT

More information

Statistics AGAIN? Descriptives

Statistics AGAIN? Descriptives Cal State Northrdge Ψ427 Andrew Answorth PhD Statstcs AGAIN? What do we want to do wth statstcs? Organze and Descrbe patterns n data Takng ncomprehensble data and convertng t to: Tables that summarze the

More information

AMP-LATCH* Ultra Novo mm [.025 in.] Ribbon Cable 02 MAR 12 Rev C

AMP-LATCH* Ultra Novo mm [.025 in.] Ribbon Cable 02 MAR 12 Rev C AMP-LATCH* Ultra Novo Applcaton Specfcaton Receptacle Connectors for 114-40056 0.64 mm [.025 n.] Rbbon Cable 02 MAR 12 All numercal values are n metrc unts [wth U.S. customary unts n brackets]. Dmensons

More information

THE IMPORTANCE OF ARM-SWING DURING FORWARD DIVE AND REVERSE DIVE ON SPRINGBOARD

THE IMPORTANCE OF ARM-SWING DURING FORWARD DIVE AND REVERSE DIVE ON SPRINGBOARD THE MPORTANCE OF ARM-SWNG DURNG FORWARD DVE AND REVERSE DVE ON SPRNGBOARD Ken Yokoyama Laboratory of Bomechancs Faculty ofeducaton Kanazawa Unversty Kanazawa, Japan J unjro Nagano Department of Physcal

More information

Simple VBR Harmonic Broadcasting (SVHB)

Simple VBR Harmonic Broadcasting (SVHB) mple VBR Harmonc Broadcastng (VHB) Hsang-Fu Yu ab, Hung-hang Yang a, Y-Mng hen c, -Mng Tseng a, and hen-y Kuo a a Dep. of omputer cence & Informaton Engneerng, atonal entral Unversty, Tawan b omputer enter,

More information

Simon Sheu Computer Science National Tsing Hua Universtity Taiwan, ROC

Simon Sheu Computer Science National Tsing Hua Universtity Taiwan, ROC Mounr A. Tantaou School of Electrcal Engneerng and Computer Scence Unversty of Central Florda Orlando, FL 3286-407-823-393 tantaou@cs.ucf.edu Interacton wth Broadcast Vdeo Ken A. Hua School of Electrcal

More information

Failure Rate Analysis of Power Circuit Breaker in High Voltage Substation

Failure Rate Analysis of Power Circuit Breaker in High Voltage Substation T. Suwanasr, M. T. Hlang and C. Suwanasr / GMSAR Internatonal Journal 8 (2014) 1-6 Falure Rate Analyss of Power Crcut Breaker n Hgh Voltage Substaton Thanapong Suwanasr, May Thandar Hlang and Cattareeya

More information

MODELING AND ANALYZING THE VOCAL TRACT UNDER NORMAL AND STRESSFUL TALKING CONDITIONS

MODELING AND ANALYZING THE VOCAL TRACT UNDER NORMAL AND STRESSFUL TALKING CONDITIONS MODELING AND ANALYZING THE VOCAL TRACT UNDER NORMAL AND STRESSFUL TALING CONDITIONS Ismal Shahn and Naeh Botros 2 Electrcal/Electroncs and Comuter Engneerng Deartment Unversty of Sharjah, P. O. Box 27272,

More information

TRADE-OFF ANALYSIS TOOL FOR INTERACTIVE NONLINEAR MULTIOBJECTIVE OPTIMIZATION Petri Eskelinen 1, Kaisa Miettinen 2

TRADE-OFF ANALYSIS TOOL FOR INTERACTIVE NONLINEAR MULTIOBJECTIVE OPTIMIZATION Petri Eskelinen 1, Kaisa Miettinen 2 Internatonal Conference 20th EURO Mn Conference Contnuous Optmaton and Knowledge-Based Technologes (EurOPT-2008) May 20 23, 2008, Nernga, LITHUANIA ISBN 978-9955-28-283-9 L. Saalausas, G.W. Weber and E.

More information

Study on the location of building evacuation indicators based on eye tracking

Study on the location of building evacuation indicators based on eye tracking Study on the locaton of buldng evacuaton ndcators based on eye trackng Yue L Tsnghua Unversty yue-l5@malstsnghuaeducn Png hang Tsnghua Unversty zhangp@malstsnghuaeducn Hu hang Tsnghua Unversty, zhhu@tsnghuaeducn

More information

Quantization of Three-Bit Logic for LDPC Decoding

Quantization of Three-Bit Logic for LDPC Decoding Proceedngs of the World Congress on Engneerng and Computer Scence 2011 Vol II, October 19-21, 2011, San Francsco, USA Quantzaton of Three-Bt Logc for LDPC Decodng Raymond Moberly and Mchael E. O'Sullvan

More information

Integration of Internet of Thing Technology in Digital Energy Network with Dispersed Generation

Integration of Internet of Thing Technology in Digital Energy Network with Dispersed Generation Amercan Scentfc Research Journal for Engneerng, Technology, and Scences (ASRJETS) ISS (Prnt) 2313-4410, ISS (Onlne) 2313-4402 Global Socety of Scentfc Research and Researchers http://asrjetsjournal.org/

More information

Small Area Co-Modeling of Point Estimates and Their Variances for Domains in the Current Employment Statistics Survey

Small Area Co-Modeling of Point Estimates and Their Variances for Domains in the Current Employment Statistics Survey Small Area Co-Modelng of Pont Estmates and Ther Varances for Domans n the Current Employment Statstcs Survey Jule Gershunskaya, Terrance D. Savtsky U.S. Bureau of Labor Statstcs FCSM, March 2018 Dsclamer:

More information

Correcting Image Placement Errors Using Registration Control (RegC ) Technology In The Photomask Periphery

Correcting Image Placement Errors Using Registration Control (RegC ) Technology In The Photomask Periphery Correctng Image Placement Errors Usng Regstraton Control (RegC ) Technology In The Photomask Perphery Av Cohen 1, Falk Lange 2 Guy Ben-Zv 1, Erez Gratzer 1, Dmtrev Vladmr 1 1. Carl Zess SMS Ltd., Karmel

More information

A Scalable HDD Video Recording Solution Using A Real-time File System

A Scalable HDD Video Recording Solution Using A Real-time File System H. L et al.: A Scalable HDD Vdeo Recordng Soluton Usng A Real-tme Fle System A Scalable HDD Vdeo Recordng Soluton Usng A Real-tme Fle System Hong L, Stephen R. Cumpson Member, IEEE, Robert Jochemsen, Jan

More information

Cost-Aware Fronthaul Rate Allocation to Maximize Benefit of Multi-User Reception in C-RAN

Cost-Aware Fronthaul Rate Allocation to Maximize Benefit of Multi-User Reception in C-RAN Cost-Aware Fronthaul Rate Allocaton to Maxmze Beneft of Mult-User Recepton n C-RAN Dora Bovz, Chung Shue Chen, Sheng Yang To cte ths verson: Dora Bovz, Chung Shue Chen, Sheng Yang. Cost-Aware Fronthaul

More information

Critical Path Reduction of Distributed Arithmetic Based FIR Filter

Critical Path Reduction of Distributed Arithmetic Based FIR Filter Crtcal Path Reducton of strbuted rthmetc Based FIR Flter Sunta Badave epartment of Electrcal and Electroncs Engneerng.I.T, urangabad aharashtra, Inda njal Bhalchandra epartment of Electroncs and Telecommuncaton

More information

Reduce Distillation Column Cost by Hybrid Particle Swarm and Ant

Reduce Distillation Column Cost by Hybrid Particle Swarm and Ant From the SelectedWorks of Dr. Sandp Kumar Lahr Summer July 20, 2016 Reduce Dstllaton Column Cost by Hybrd Partcle Swarm and Ant Dr. Sandp k lahr chnmaya lenka Avalable at: https://works.bepress.com/sandp_lahr/33/

More information

Craig Webre, Sheriff Personnel Division/Law Enforcement Complex 1300 Lynn Street Thibodaux, Louisiana 70301

Craig Webre, Sheriff Personnel Division/Law Enforcement Complex 1300 Lynn Street Thibodaux, Louisiana 70301 DATE OF APPLCATON: Craig Webre, Sheriff Personnel Division/Law Enforcement Complex 1300 Lynn Street Thibodaux, Louisiana 70301 N GENERAL EMAL ADDRESS: For Local Calls - (985) 532-4380 (985) 446-2255 (985)

More information

The Traffic Image Is Dehazed Based on the Multi Scale Retinex Algorithm and Implementation in FPGA Cui Zhe1, a, Chao Li2, b *, Jiaqi Meng3, c

The Traffic Image Is Dehazed Based on the Multi Scale Retinex Algorithm and Implementation in FPGA Cui Zhe1, a, Chao Li2, b *, Jiaqi Meng3, c 3rd Internatonal Conference on Mechatroncs and Industral Informatcs (ICMII 2015) The Traffc Image Is Dehazed Based on the Mult Scale Retnex Algorthm and Implementaton n FPGA Cu Zhe1, a, Chao L2, b *, Jaq

More information

Analysis of Subscription Demand for Pay-TV

Analysis of Subscription Demand for Pay-TV Analyss of Subscrpton Demand for Pay-TV Manabu Shshkura Researcher Insttute for Informaton and Communcatons Polcy 2-1-2 Kasumgasek, Chyoda-ku Tokyo 110-8926 Japan m-shshkura@soumu.go.jp Tel: 03-5253-5496

More information

Simple Solution for Designing the Piecewise Linear Scalar Companding Quantizer for Gaussian Source

Simple Solution for Designing the Piecewise Linear Scalar Companding Quantizer for Gaussian Source 94 J. NIKOIĆ, Z. PERIĆ,. VEIMIROVIĆ, SIMPE SOUTION FOR DESIGNING THE PIECEWISE INEAR SCAAR Smle Soluton for Desgnng the Pecewse near Scalar Comandng Quantzer for Gaussan Source Jelena NIKOIĆ, Zoran PERIĆ,

More information

Scalable QoS-Aware Disk-Scheduling

Scalable QoS-Aware Disk-Scheduling Scalable QoS-Aware Dsk-Schedulng Wald G. Aref Khaled El-Bassyoun Ibrahm Kamel Mohamed F. Mokbel Department of Computer Scences, urdue Unversty, West Lafayette, IN 47907-1398 anasonc Informaton and Networkng

More information

AIAA Optimal Sampling Techniques for Zone- Based Probabilistic Fatigue Life Prediction

AIAA Optimal Sampling Techniques for Zone- Based Probabilistic Fatigue Life Prediction AIAA 2002-383 Optmal Samplng Technques or Zone- Based Probablstc Fatgue Le Predcton M. P. Enrght Southwest Research Insttute San Antono, TX H. R. Mllwater Unversty o Texas at San Antono San Antono, TX

More information

Production of Natural Penicillins by Strains of Penicillium chrysogenutn

Production of Natural Penicillins by Strains of Penicillium chrysogenutn Producton of Natural Pencllns by Strans of Pencllum chrysogenutn a J. FUSK and ЬЕ. WELWRDOVÁ ^Department of Mcrobology and Bochemstry, Slovak Techncal Unversty, Bratslava b Botka, Slovenská Ľupča Receved

More information

Expressive Musical Timing

Expressive Musical Timing Axel Berndt, Tlo Hähnel Department of Smulaton and Graphcs Otto-von-Guercke Unversty of Magdeburg {aberndt tlo}@sg.cs.un-magdeburg.de Abstract. Tmng s crucal for the qualty of expressve musc performances.

More information

FPGA Implementation of Cellular Automata Based Stream Cipher: YUGAM-128

FPGA Implementation of Cellular Automata Based Stream Cipher: YUGAM-128 ISSN (Prnt) : 2320 3765 ISSN (Onlne): 2278 8875 Internatonal Journal of Advanced Research n Electrcal, Electroncs and Instrumentaton Engneerng An ISO 3297: 2007 Certfed Organzaton Vol. 3, Specal Issue

More information

INSTRUCTION MANUAL FOR THE INSTALLATION, USE AND MAINTENANCE OF THE REGULATOR GENIUS POWER COMBI

INSTRUCTION MANUAL FOR THE INSTALLATION, USE AND MAINTENANCE OF THE REGULATOR GENIUS POWER COMBI NSTRUCTON MANUAL FOR THE NSTALLATON, USE AND MANTENANCE OF THE REGULATOR GENUS POWER COMB (TRANSLATON OF THE ORGNAL NSTRUCTON MANUAL N TALAN) PRELMNARY VERSON WARRANTY The devce s guaranteed 24 months

More information

Detecting Errors in Blood-Gas Measurement by Analysiswith Two Instruments

Detecting Errors in Blood-Gas Measurement by Analysiswith Two Instruments CLIN. CHEM. 33/4, 512-517 (1987) Detectng Errors n Blood-Gas Measurement by Analysswth Two Instruments LouIs F. Metzger, Wllam B. Stauffer, Ann V. Kruplnskl, Rchard P. MIIlman,3 George S. Cembrowskl,2

More information

Product Information. Manual change system HWS

Product Information. Manual change system HWS Product Informaton HWS HWS Flexble. Compact. Productve. HWS manual change system Manual tool change system wth ntegrated ar feed-through and optonal electrc feed-through Feld of applcaton Excellently sutable

More information

Lost on the Web: Does Web Distribution Stimulate or Depress Television Viewing?

Lost on the Web: Does Web Distribution Stimulate or Depress Television Viewing? Lost on the Web: Does Web Dstrbuton Stmulate or Depress Televson Vewng? Joel Waldfogel The Wharton School Unversty of Pennsylvana August 10, 2007 Prelmnary comments welcome Abstract In the past few years,

More information

SKEW DETECTION AND COMPENSATION FOR INTERNET AUDIO APPLICATIONS. Orion Hodson, Colin Perkins, and Vicky Hardman

SKEW DETECTION AND COMPENSATION FOR INTERNET AUDIO APPLICATIONS. Orion Hodson, Colin Perkins, and Vicky Hardman SKEW DETECTION AND COMPENSATION FOR INTERNET AUDIO APPLICATIONS Oron Hodson, Coln Perkns, and Vcky Hardman Department of Computer Scence Unversty College London Gower Street, London, WC1E 6BT, UK. ABSTRACT

More information

Clock Synchronization in Satellite, Terrestrial and IP Set-top Box for Digital Television

Clock Synchronization in Satellite, Terrestrial and IP Set-top Box for Digital Television Clock Synchronzaton n Satellte, Terrestral and IP Set-top Box for Dgtal Televson THESIS Submtted n partal fulflment of the requrements for the degree of DOCTOR OF PHILOSOPHY by MONIKA JAIN Under the Supervson

More information

User s manual. Digital control relay SVA

User s manual. Digital control relay SVA User s manual Dgtal control relay DISIBEINT ELECTRONIC S.L, has been present n the feld of the manufacture of components for the ndustral automaton for more than years, and mantans n constant evoluton

More information

Conettix D6600/D6100IPv6 Communications Receiver/Gateway Quick Start

Conettix D6600/D6100IPv6 Communications Receiver/Gateway Quick Start Conettx / Communcatons Recever/Gateway Quck Start.0 Parts Lst able : Conettx System Components Qty. Descrpton Conettx Communcatons Recever/Gateway AC power cord Battery cable P660 I/O cable P660 Rack mount

More information

Product Information. Manual change system HWS

Product Information. Manual change system HWS Product Informaton HWS HWS Flexble. Compact. Productve. HWS manual change system Manual tool change system wth ntegrated ar feed-through and optonal electrc feed-through Feld of applcaton Excellently sutable

More information

3 Part differentiation, 20 parameters, 3 histograms Up to patient results (including histograms) can be stored

3 Part differentiation, 20 parameters, 3 histograms Up to patient results (including histograms) can be stored st Techncal Specfcatons Desgned n France Wth a rch past and a professonal experence bult-up over 35 years, SFRI s a French nvatve company commtted to developng preon In Vtro Dst solutons. SFRI has bult

More information

Multi-Line Acquisition With Minimum Variance Beamforming in Medical Ultrasound Imaging

Multi-Line Acquisition With Minimum Variance Beamforming in Medical Ultrasound Imaging IEEE Transactons on Ultrasoncs, Ferroelectrcs, and Frequency Control, vol. 60, no. 12, Decemer 2013 2521 Mult-Lne Acquston Wth Mnmum Varance Beamformng n Medcal Ultrasound Imagng Ad Ranovch, Zv Fredman,

More information

Improving Reliability and Energy Efficiency of Disk Systems via Utilization Control

Improving Reliability and Energy Efficiency of Disk Systems via Utilization Control Ths paper appeared n the Proceedngs of the 2th IEEE Symposum on Computers and Communcatons (ISCC'08, Marrakech, Morocco, July 2008. Improvng Relablty and Energy Effcency of Dsk Systems va Utlzaton Control

More information

Environmental Reviews. Cause-effect analysis for sustainable development policy

Environmental Reviews. Cause-effect analysis for sustainable development policy Envronmental Revews Cause-effect analyss for sustanable development polcy Journal: Envronmental Revews Manuscrpt ID er-2016-0109.r2 Manuscrpt Type: Revew Date Submtted by the Author: 24-Feb-2017 Complete

More information

in Partial For the Degree of

in Partial For the Degree of 37q h8( sta. co AN ANALYSS OF ROBERT NATHANEL DETT'S N THE BOTTOMS THESS Presented to the Graduate ouncl of the North Texas State Unversty n Partal Fulfllment of the Requrements For the Degree of MASTER

More information

T541 Flat Panel Monitor User Guide ENGLISH

T541 Flat Panel Monitor User Guide ENGLISH T541 Flat Panel Montor User Gude ENGLISH Frst Edton (June / 2002) Note : For mportant nformaton, refer to the Montor Safety and Warranty manual that comes wth ths montor. Ths publcaton could contan techncal

More information

IN DESCRIBING the tape transport of

IN DESCRIBING the tape transport of Apparatus For Magnetc Storage on Three-Inch Wde Tapes R. B. LAWRANCE R. E. WILKINS R. A. PENDLETON IN DESCRIBING the tape transport of the DATAmatc 1, t s perhaps well to begn by revewng the nfluental

More information

Modular Plug Connectors (Standard and Small Conductor)

Modular Plug Connectors (Standard and Small Conductor) Modular Plug Connectors (Standard and Small Conductor) Applcaton Specfcaton 114-6016 04 APR 11 All numercal values are n metrc unts [wth U.S. customary unts n brackets]. Dmensons are n mllmeters [and nches].

More information

Color Monitor. L200p. English. User s Guide

Color Monitor. L200p. English. User s Guide Color Montor L200p User s Gude Englsh Frst Edton (February / 2003) Note : For mportant nformaton, refer to the Montor Safety and Warranty manual that comes wth ths montor. Contents ENGLISH Safety (Read

More information

Product Information. Miniature rotary unit ERD

Product Information. Miniature rotary unit ERD Product Informaton ERD ERD Fast. Compact. Flexble. ERD torque motor Powerful torque motor wth absolute encoder and electrc and pneumatc rotary feed-through Feld of applcaton For all applcatons wth exceptonal

More information

Five Rounds. by Peter Billam. Peter J Billam, 1986

Five Rounds. by Peter Billam. Peter J Billam, 1986 Fve Rounds by Per Bllam Per J Bllam, 986 Ths score s offered under the Creatve Commons Attrbuton.0 Inrnatonal lcence; see creatvecommons.org The copyrght owner remans the composer, Per Bllam. Ths edton

More information

Product Information. Universal swivel units SRU-plus

Product Information. Universal swivel units SRU-plus Product Informaton Unversal swvel unts SRU-plus SRU-plus Unversal swvel unts Robust. Fast. Hgh Performance. SRU-plus unversal rotary actuator Unversal unt for pneumatc swvel and turnng movements. Feld

More information

Discussion Paper Series

Discussion Paper Series Doshsha Unversty Center for the Study of the Creatve Economy Dscusson Paper Seres No. 2013-04 Nonlnear Effects of Superstar Collaboraton: Why the Beatles Succeeded but Broke Up Tadash Yag Dscusson Paper

More information

(12) Ulllted States Patent (10) Patent N0.: US 8,269,970 B2 P0lid0r et a]. (45) Date of Patent: Sep. 18, 2012

(12) Ulllted States Patent (10) Patent N0.: US 8,269,970 B2 P0lid0r et a]. (45) Date of Patent: Sep. 18, 2012 US008269970B2 (12) Ulllted States Patent (10) Patent N0.: P0ld0r et a]. (45) Date of Patent: Sep. 18, 12 (54) OPTICAL COMPARATOR WITH DIGITAL 6,945,652 B2 9/05 sakqta et a1 GAGE 7,058,109 B2* 6/06 Davs.....

More information

Loewe bild 5.55 oled. Modular Design Flexible configuration with individual components. Set-up options. TV Monitor

Loewe bild 5.55 oled. Modular Design Flexible configuration with individual components. Set-up options. TV Monitor Product nformaton Loewe bld 5.55 oled Page of 3 Loewe bld 5.55 oled EU energy effcency class: B Screen dagonal (n cm) / Screen dagonal (n nch): 39 / 55 Power consumpton ON (n W): 50 Annual energy consumpton

More information

Fast Intra-Prediction Mode Decision in H.264/AVC Based on Macroblock Properties

Fast Intra-Prediction Mode Decision in H.264/AVC Based on Macroblock Properties Fast Intra-Predcton Mode Decson n H.264/AVC Based on Macroblock Propertes Abstract Intra-predcton s a wdely used tecnque n ntra codng. H.264/AVC adopts rate-dstorton optmzaton (RDO) tecnque to obtan te

More information

User Manual. AV Router. High quality VGA RGBHV matrix that distributes signals directly. Controlled via computer.

User Manual. AV Router. High quality VGA RGBHV matrix that distributes signals directly. Controlled via computer. User Manual AV Router Hgh qualty VGA RGBHV matrx that dstrbutes sgnals drectly. Controlled va computer. Notce: : The nmaton contaned n ths document s subject to change wthout notce. SmartAVI makes no warranty

More information

Loewe bild 7.65 OLED. Set-up options. Loewe bild 7 cover Incl. Back cover. Loewe bild 7 cover kit Incl. Back cover and Speaker cover

Loewe bild 7.65 OLED. Set-up options. Loewe bild 7 cover Incl. Back cover. Loewe bild 7 cover kit Incl. Back cover and Speaker cover Product nformaton Loewe bld 7.65 Page of March 07 Loewe bld 7.65 OLED EU energy effcency class: B Screen dagonal (n cm) / Screen dagonal (n nch): 64 / 65 Power consumpton ON (n W): 80 Annual energy consumpton

More information

User guide. Receiver-In-The-Ear hearing aids, rechargeable Hearing aid charger. resound.com

User guide. Receiver-In-The-Ear hearing aids, rechargeable Hearing aid charger. resound.com User gude Recever-In-The-Ear hearng ads, rechargeable Hearng ad charger resound.com 400973011US-18.08-Rev.A.ndd 1 01-08-2018 14:18:12 Left Hearng Ad Rght Hearng Ad Seral number Seral number Model number

More information

User guide. Receiver-In-The-Ear hearing aids, rechargeable Hearing aid charger. resound.com

User guide. Receiver-In-The-Ear hearing aids, rechargeable Hearing aid charger. resound.com User gude Recever-In-The-Ear hearng ads, rechargeable Hearng ad charger resound.com Seral number Model number Recever type Left Hearng Ad Low Power Medum Power Hgh Power Ultra Power Seral number Model

More information

JTAG / Boundary Scan. Multidimensional JTAG / Boundary Scan Instrumentation

JTAG / Boundary Scan. Multidimensional JTAG / Boundary Scan Instrumentation JTAG / Boundary Scan Multdmensonal JTAG / Boundary Scan Instrumentaton 2 GOEPEL electronc & JTAG / Boundary Scan COMPANY GOEPEL electronc GmbH GOEPEL electronc s a global company that has been developng

More information

Product Bulletin 40C 40C-10R 40C-20R 40C-114R. Product Description For Solvent, Eco-Solvent, UV and Latex Inkjet and Screen Printing 3-mil vinyl films

Product Bulletin 40C 40C-10R 40C-20R 40C-114R. Product Description For Solvent, Eco-Solvent, UV and Latex Inkjet and Screen Printing 3-mil vinyl films Product Bulletn 40C Revson D, Effectve February 2016 (Replaces C, Apr. 15) 40C-10R 40C-20R 40C-114R Product Descrpton For Solvent, Eco-Solvent, UV and Latex Inkjet and Screen Prntng 3-ml vnyl flms Quck

More information

Step 3: Select a Class

Step 3: Select a Class R Step 1: Roll Ablty Scores a. ndcate dce-rollng method (p. 13):. Roll 3d6 sx tmes, n order. 11. Roll 3d6 twce per ablty, select ether. 111. Roll 3d6 sx tmes and assgn to abltes as desred. V. Roll 3d6

More information

Sealed Circular LC Connector System Plug

Sealed Circular LC Connector System Plug Sealed Crcular LC Connector System Plug Instructon Sheet Kt 1828618- [ ], Receptacle Kt 1828619- [ ], 408-10079 and EMI Receptacle Kt 1985193- [ ] 07 APR 11 Plug Kt 1828618 -[ ] Cable Fttng Receptacle

More information

JTAG / Boundary Scan. Multidimensional JTAG / Boundary Scan Instrumentation. Get the total Coverage!

JTAG / Boundary Scan. Multidimensional JTAG / Boundary Scan Instrumentation. Get the total Coverage! JTAG / Boundary Scan Multdmensonal JTAG / Boundary Scan Instrumentaton IEEE 1149.6 IEEE 1149.1 IEEE 1149.7 Multdmensonal JTAG / Boundary Scan Instrumentaton IEEE 1149.4 IEEE 1532 Get the total Coverage!

More information

Automatic Piano Music Transcription

Automatic Piano Music Transcription Automatic Piano Music Transcription Jianyu Fan Qiuhan Wang Xin Li Jianyu.Fan.Gr@dartmouth.edu Qiuhan.Wang.Gr@dartmouth.edu Xi.Li.Gr@dartmouth.edu 1. Introduction Writing down the score while listening

More information

Product Information. Universal swivel units SRU-plus 25

Product Information. Universal swivel units SRU-plus 25 Product Informaton SRU-plus Robust. Fast. Hgh Performance. SRU-plus unversal rotary actuator Unversal unt for pneumatc swvel and turnng movements. Feld of applcaton Can be used n ether clean or contamnated

More information

User guide. Receiver-In-Ear hearing aids. resound.com

User guide. Receiver-In-Ear hearing aids. resound.com User gude Recever-In-Ear hearng ads resound.com 400786011US-17.07-Rev.A.ndd 1 20-07-2017 12:52:40 Left Hearng Ad Rght Hearng Ad Seral number Seral number Model number Model number Recever type Recever

More information

CONNECTIONS GUIDE. To Find Your Hook.up Turn To Page 1

CONNECTIONS GUIDE. To Find Your Hook.up Turn To Page 1 CONNECTIONS GUIDE To Fnd Your Hook.up Turn To Page 1 Connectng TV to Antenna (or Cable Wthout Cable Box) and No VCR (Hook-up 1A)... 2 Monaural VCR (Hook-up 1B)... 3 StereoVCR (Hook-up 1C)... 4 Cable Wth

More information

Printer Specifications

Printer Specifications : Characterfonts:! Fort Pont 7P 0.5 pl Ptch 5cpl, Ocpl, 2cpl Proptlonel Epson Draf!o 0 lo j Epson Cower 0 0 0 Epson Roman O O O 0 Epson San6 Sent 0 O O O Epson Presllge j0 0 ~Epson Scnpt O 0 Epson Sormt

More information

Emotional Metaphors for Emotion Recognition in Chinese Text

Emotional Metaphors for Emotion Recognition in Chinese Text Emotonal Metaphors for Emoton Recognton n Chnese Text Xaox Huang 1, Yun Yang 2 and Changle Zhou 1,2 1 College of Computer Scence, Zhejang Unversty, 310027, P.R. Chna tshere@zju.edu.cn 2 Insttute of Artfcal

More information

9! VERY LARGE IN THEIR CONCERNS. AND THEREFORE, UH, i

9! VERY LARGE IN THEIR CONCERNS. AND THEREFORE, UH, i 340 WELL, alack PAJAMAS WAS A SOMEWHAT METAPHORCAL 2 TERM. MANY VETNAMESE PEASANTS TENDED TO WEAR 3 BLACK PAJAMAS, BUT WHAT AM REFERRNG TO S THAT 4 OUTSDE OF THE NORTH VETNAMESE UNTS AND ~OME OF 5 THE

More information

V (D) i (gm) Except for 56-7,63-8 Flute and Oboe are the same. Orchestration will only list Fl for space purposes

V (D) i (gm) Except for 56-7,63-8 Flute and Oboe are the same. Orchestration will only list Fl for space purposes Measure # 1 2 3 5 6 7 8 9 10 11 12 13 1 15 16 17 18 Intro Motve 1 A Motve 1 B - "Wakarathe" Dynamcs mp terraced dymancs to mm.7 p Moderately fast and lght (quarter=96) 3 G Mnor motve 1 con't mp melody

More information

CONNECTIONS GUIDE. To Find Your Hook.up Turn To Page 1

CONNECTIONS GUIDE. To Find Your Hook.up Turn To Page 1 CONNECTIONS GUIDE To Fnd Your Hook.up Turn To Page 1 Connectng TV to Antenna (or Cable Wthout Cable Box) and No VCR (Hook-up 1A)...2 Monaural VCR (Hook-up 1B)...3 Stereo VCR (Hook-up 1C)... 4 Cable Wth

More information

User Manual ANALOG/DIGITAL, POSTIONER RECEIVER WITH EMBEDDED VIACCESS AND COMMON INTERFACE

User Manual ANALOG/DIGITAL, POSTIONER RECEIVER WITH EMBEDDED VIACCESS AND COMMON INTERFACE User Manual ANALOG/DIGITAL, POSTIONER RECEIVER WITH EMBEDDED VIACCESS AND COMMON INTERACE CONTENTS. Safety nstructons -------------------------------------------------------------------. eatures -------------------------------------------------------------------------------.

More information

A question of character. Loewe Connect ID.

A question of character. Loewe Connect ID. A queston of character. Loewe Connect ID. Modern. Etquette you can learn, character s nnate. You make a clear dstncton between good manners and genune style. Your s s lke yourself: unfussy, busnesslke

More information

S Micro--Strip Tool in. S Combination Strip Tool ( ) S Cable Holder Assembly (Used only

S Micro--Strip Tool in. S Combination Strip Tool ( ) S Cable Holder Assembly (Used only Instructon Sheet LghtCrmp* Plus LC 408-10103 (for Jacketed Cable) Connectors 18 AUG 09 Rear Protectve Cap Termnaton CoverG Boot Connector Assembly Crmp Eyelet Duplex Clp G Connector kt s shpped wth these

More information

INTERCOM SMART VIDEO DOORBELL. Installation & Configuration Guide

INTERCOM SMART VIDEO DOORBELL. Installation & Configuration Guide INTERCOM SMART VIDEO DOORBELL Installaton & Confguraton Gude ! Important safety nformaton Read ths manual before attemptng to nstall the devce! Falure to observe recommendatons ncluded n ths manual may

More information

INIHODU~IION AND NOI[~ OJ KJUN~ HO rahk

INIHODU~IION AND NOI[~ OJ KJUN~ HO rahk HAROLD COURLANO[R, C[N[RAL [OITOR INIHODU~IION AND NOI[~ OJ KJUN~ HO rahk Covel desgon by Ronald Clyne , ~ C» ==-== :' C ::::J FOLKWA YS RECORDS FE 4424 ~, C C.., -- s;::: :- C I:) : C I:) ~..,.,---, a,

More information

CASH TRANSFER PROGRAMS WITH INCOME MULTIPLIERS: PROCAMPO IN MEXICO

CASH TRANSFER PROGRAMS WITH INCOME MULTIPLIERS: PROCAMPO IN MEXICO FCND DP No. 99 FCND DISCUSSION PAPER NO. 99 CASH TRANSFER PROGRAMS WITH INCOME MULTIPLIERS: PROCAMPO IN MEXICO Elsabeth Sadoulet, Alan de Janvry, and Benjamn Davs Food Consumpton and Nutrton Dvson Internatonal

More information