36 IJCSNS Iteratioal Joural of Computer Sciece ad Network Security, VO.6 No.7A, July 006 The Commuicatio ethod of Distace Educatio System ad Soud Cotrol Characteristics aabu Ishihara, Departmet of Electrical ad Computer Egieerig, Oyama Natioal College of Techology, Oyama-City, Tochigi, 33-0806 JAPAN Summary The preset i used distace educatio system is based o audiovisual commuicatios. So whe a teacher gives lessos to may studets who are dispersed i distat places, the teacher obtais iput from studets through image ad soud iformatio. The studets ca be visually distiguished by separately displayig a active picture ad a o-active backgroud. But it is hard to distiguish the studets aurally whe multiple souds come from the same loudspeaker simultaeously. I a ormal class, a teacher ca usually obtai feedback from studets ad ca respod to a questio immediately. It is equally ecessary to get studet feedback through soud iformatio i distace educatio. We examied how backgroud soud affects the active soud i a practical distace educatio system, ad studied how to mai image iterferece. d experimet, Next, we defie the soud loudess as the soud pressure, perform AHP (Aalytic Hierarchy Process) aalysis for relatio betwee the soud pressure ad the soud image i this case, ad elucidate the audibility characteristics. We defie the time delay as the soud pressure, perform AHP aalysis for relatio betwee the time-differece ad the soud image i this case, ad elucidate the audibility characteristics. Key words: Auditory sese, Iteret, Classroom desig, Distace educatio, ultipoit commuicatio, AHP(Aalytic Hierarchy Process), Soud pressure, Time-differece. Itroductio Recetly, commuicatio techology is rapidly developig. Therefore, it becomes possible for schools ad campuses which are physically separate to be coected by a digital trasmissio etwork ad commuicatio satellite. A joit class o real-time basis ca be carried out []. We improve distace educatio system usig a video telecoferece system, ad i this way joit class i the uiversity ad distace educatio for studets who refuse to go to school were carried out [][3][4]. Although, the distace educatio system has already etered the practical stage, but there seems to a problem with respect to the soud. The preset distace educatio system is based o audiovisual commuicatios. So whe a teacher gives lessos to may studets distat places, the teacher gets feedback from studets through image ad soud iformatio. The studets ca be visually distiguished by separately displayig a active picture ad a o-active backgroud. But it is hard to distiguish the studets aurally whe multiple souds come from the same loudspeaker simultaeously. I a ormal class, a teacher ca usually get feedback from studets ad respod to a questio immediately. It is equally ecessary to get studet feedback through soud i distace educatio. The teacher always eeds feedback from the aspect the pupil. We produced such a situatio i the actual distace educatio system. We tested to what degree the backgroud soud had o effect o the active soud [5]. As a result of this experimet, recogitio rate was determied to be about 70% eve with 70dB soud pressure backgroud i case of the moaural presetatio. oreover, it was show that eve whe the soud pressure of the backgroud was 60dB, the presetatio could ot be easily uderstood. I stereo presetatio however, soud pressure of the backgroud does ot affect the much active soud, eve if it is 90dB[6]. Next, we defie the soud loudess as the soud pressure, perform AHP (Aalytic Hierarchy Process) aalysis for relatio betwee the soud pressure ad the soud image i this case, ad elucidate the audibility characteristics. We defie the time delay as the soud pressure, perform AHP aalysis for relatio betwee the time-differece ad the soud image i this case, ad elucidate the audibility characteristics[7].. The Cofiguratio of The Distace Educatio System. The Distace Educatio System I this study, the video telecoferece system used was "PictureTel System 000 odel 30". Two systems were placed i adjoiig classrooms, to create the distace educatio system were[5]. The mai body of video auscript received July 8, 006. auscript revised July30, 006.
IJCSNS Iteratioal Joural of Computer Sciece ad Network Security, VO.6 No.7A, July 006 37 telecoferece systems placed i each the computer practice room by the Digital PBX at TBRI (Triple Basic Rate Iterface). This allowed a maximum data trasfer of 384kbps, ad movig image of 30fps. The microphoe for the collectig soic reflectio collects soud of a wider scope usig "Poweric (collectig soic reflectio rage of 360 degrees, 00Hz-7kHz frequecy rage)".. Outlie of AHP The AHP is a tool for judgmet or decisio that has ot yet bee structuralized, ad was proposed by T..Saaty[6]. It has also show that this method ca express huma ad physical sesatios with a comparative correctess. We calculated the importace ratio(weight) of alterative I ad J based o a assessmet criterio, ad made a matrix of paired comparisos. Eigevector compoets (the sum of compoets ormalized as ) belogig to the maximum eigevector of this matrix were cosidered to be the weight (priority) of each alterative. For example, assume that there are assessmet items of I, ad I, ad their origial weights are W, ad W. The paired compariso value a ij of the importace of items I i ad I j is give as A = [ a ij ] of paired comparisos becomes. W W A = W a =. atrix ultiplyig the vector of weight o the right side of A it will chage to the followig: W W W ij W / W W / W W / W i j () W / W W W W / W W W = W / W W W ( ) ν ν ν = ν () 3 From the relatio of eigevalue ad eigevector, a formula ca be derived that ( ) A ν = λ maxν 4 I additio, the cosistecy idex (C.I.) is used i the assessmet of AHP results. ( max ) /( ) ( 5) C. I. = λ C.I. is zero i matrix A with perfect cosistecy. The larger C.I., the higher the icosistecy. This meas that the smaller C.I. is the better. The cosistecy scale is used as the cosistecy ratio (C.R.) ad is judged to have cosistecy whe C.R. is below 0.. The experimet was as follows. et us determie a priority scale i the followig experimet. et A, B, C, ad D stad for listeers, arraged i a straight lie, leadig away from acoustic sigal (soud pressure source). We develop a priority scale of relative loudess for the acoustic sigal. Judgmets will be obtaied from a idividual who stads by the acoustic sigal source ad is asked, for example, How strogly louder is acoustic sigal B tha acoustic sigal C? The subject will the give oe umber for compariso i the table ad this will be etered i the matrix i positio (B, C). By covetio, the compariso of stregth is always of a activity appearig i the colum i the left agaist a activity appearig i the row o top. We the have the pairwise compariso matrix with four rows ad four colums (a 4 4 matrix). (Table ). Table. atrix of AHP(Example) From this formula, it is clear that the weight vector is the eigevector, is the eigevalue ad the maximum oe of matrix A. After fidig the maximum eigevalue ad eigevector of A, the eigevector turs out to be the weight of each assessmet item. If takig λ max as the maximum eigevalue ad ν as the eigevector of a geeral matrix A of paired compariso, it becomes. A B C D A B C D
38 IJCSNS Iteratioal Joural of Computer Sciece ad Network Security, VO.6 No.7A, July 006 The agreed upo umbers are the followig, give elemets A ad B; if: A ad B are equally importat, isert A is weakly more importat tha B, isert 3 A is strogly more importat tha B, isert 5 A is demostrably or very strogly more importat tha B, isert 7 A is absolutely more importat tha B, isert 9 i the positio (A,B) where row A meets colum B. A elemet is equally importat whe compared with itself, so where row A ad colum A meet (A,A), isert. Thus, the mai diagoal of a matrix must cosist of s. A meets row B, i.e., positio (B,A) for the reverse compariso of B with A. The umbers, 4, 6, 8 ad their reciprocals are used to facilitate compromise betwee slightly differig judgmets. We also use ratioal umbers to from ratios from the above values whe it is desired to force cosistecy o the etire matrix from a few judgmets, i.e., a miimum of -. 3. Experimetal 3. The presetatio method of image ad voice Next cosider methods used for presetatio of image ad voice from actual live class to the preset remote educatio system. ()The presetatio by actual live class (Voice: ulti-chael, Image: ulti-chael). ()The presetatio of the stereo space (simulatio space) usig equipmet (Voice: chaels, Image: chaels). (3)The presetatio by the preset distace educatio system (simulatio space) (Voice: Oe chael, Image: Oe chael). We measured () ad (3) as presetatio methods for utilizatio i the distace educatio system. 3. The setece itelligibility experimet usig the preset distace educatio system This experimet used the above-metioed distace educatio system. I the ocal room, the preseter vocalized soud from the teacher s desk, ad it was collected by microphoe. Simultaeously, backgroud soud from the loudspeaker was also caught by the microphoe (Figure.). At this time, the preseter vocalized soud at a costat soud pressure, while the backgroud soud was varied. Ad, two types of oise as were used backgroud soud: readig aloud ad white oise. Each was made to vary over the followig rage: 60dB, 70dB, 80dB ad 85dB. Active soud ad backgroud soud were preseted from the four loudspeakers i the ceilig of the Remote room as part of the distace educatio system. 5 examiees were statioed i Remote (A,B,C,D, ad E), ad they heard the presetatio from the loudspeakers(figure.). icrophoe Camera PC E C ocal Vol 40 m 3.35 Remote Vol 56. m 3 B 4.0 Figure The classroom of ocal ad Remote 0.86.35 Ad, examiees observed the figure of preseter was captured by cameras i the ocal room o sub-moitors istalled i the Remote room. The setece itelligibility the preseter s questio ad subject s aswer was measured. The questio was a short setece which required oly a geeral aswer. 0 questios were used for each set of coditios. 3.3 The setece itelligibility experimet with stereo equipmet space (simulatio space) Separate voices (active soud ad backgroud soud) were preseted by the right ad left chaels of stereo headphoe. That is to say, the chael of oe side preseted the voice for the setece itelligibility test, ad the other chael preseted the oise (readig aloud). The cofiguratio of the experimet is show i Figure.. The backgroud soud was made to chage i the same rage as the above-metioed experimet: 60dB, 70dB, 80dB ad 90dB. Simultaeously, the face of the preseter of the setece itelligibility test ad the face presetig the A D PC
IJCSNS Iteratioal Joural of Computer Sciece ad Network Security, VO.6 No.7A, July 006 39 oise (readig aloud) were give o separate screes i order to sychroize the visual sese. The voice of the scree i the right would be preseted from the right hael of the headphoe. It is also the same o the left side. Backgroud Camera oitor Active years old to 0 years old. () We make use of 0,, 4, 6, 8 ad 0[dB] soud pressure level differeces. () We make use of 0, 0., 0.5, 0.5, 0.75 ad.0 [ms] soud timediffereces. The subjects liste to white-oise sigal usig headphoes. A perso uder test heard the white-oise through audio amplifier with a headphoe. The output level of the perso uder test could hear the souds most easily. VCR Amp Headphoe Backgroud 60dB 70dB 80dB 90dB Figure Active Examiees were the 5 persos metioed above. The setece itelligibility test was carried out uder the coditios exactly the same as the above-metioed experimet. The, sychroizatio was reversed. The importace of sychroous auditory sese ad visual sese was evaluated by reversig voice ad image betwee right ad left chaels. 3.4 Experimetal of soud pressure level ad timediffereces A schematic diagram of this experimet is illustrated i Figure 3. As show i the figure, we take a method to geerate white-oise with a certai soud pressure or timediffereces from a speaker at a poit ad to hear the soud at a positio determied i advace. Because of a limitatio of the measuremet for idividuals, we placed a oise meter at every positio of idividuals ad measured the soud loudess with each oise meter. The we recorded the soud with the same loudess as of this measured soud o a Computer ad made a CD-RO for soud pressure or time-differeces stimulatio. The examiee listes to the soud pressure level or timediffereces stimulatio CD-RO prepared i advace by usig a headphoe (DR-CD350, SONY). The listeers were five healthy me. The age of the examiee is from 9 R 60dB Correspodece No- Correspodece Experimet cofiguratio of the stereo space by the equipmet 4. Results ad Discussio 4. I case of the presetatio method of image ad voice First we will preset the results of the experimet i the moaural experimet. Figure4 shows setece itelligibility i the case whe white oise plays the role of backgroud oise ad Figure 5, i the case whe readig aloud plays the role of backgroud oise. Accordig to both Figure 4 ad Figure 5, whe the oise soud pressure is 70dB or less, the setece itelligibility is more tha 70 percet. The the data ca be partitioed grouped ito two groups: 60-70dB ad 80-85dB. Cosiderig these results, there is a kid of threshold betwee 70dB ad 80dB.
40 IJCSNS Iteratioal Joural of Computer Sciece ad Network Security, VO.6 No.7A, July 006 I the experimet usig white oise, as i Figure 4, setece itelligibility is almost 00 percet o the coditio that the soud pressure is 60dB, while i the experimet usig readig aloud, as i Figure 5, setece itelligibility is low compared with that i Figure 4. I the stereo experimet we used a method called dichotic listeig, i which the subject listes to differet souds i each of his/her ears. setece itelligibility is almost 00 percet. Thus there is o differece i setece itelligibility betwee CC ad No-CC. Ulike the moaural experimet, the subject ca aurally distiguish the target from the oise i the stereo experimet. Whe two images are used, there is o differece betwee CC ad No-CC. Therefore the stereo effect is cosidered to be crucial. O this basis we coducted two types of experimet. Oe (Correspodece Coditio, heceforth CC) is whe two images o the left correspods to the soud i the left, ad similarly for the right. But this does ot bold for the other coditio (No-Correspodece Coditio, heceforth No-CC) is ot. Figure 6 is setece itelligibility i CC ad Figure 7 is oe i No-CC. I cotract to the moaural experimet, i the stereo experimet, eve whe the oise soud pressure is 90dB 4. I case of soud pressure level ad timediffereces We made a paired comparisos amog these soud pressure level or time-differece respectively ad obtaied the relatio betwee the soud pressure level ad the timedifferece. I this experimet, the soud pressure level ad the weight(ratio) appear to be i logarithmic relatio. This is said almost applicable to the case of soud like a geeral relatio betwee itesities of the physical stimulatio ad the resulted sesatio, kow as Weber-Fecher's law. It is show i figure8. We made paired comparisos amog these soud image respectively ad obtaied the relatio betwee the soud image ad the time-differece. As show i the figure9, this ca be regarded as almost liear characteristics. We show a calculatio result of cosistecy idex (C.I.) for table. Besides, cosistecy idex calculated for every examiee was below 0.5 throughout the experimet. I the AHP, criterio of acceptace (good) is defied as the cosistecy idex beig less tha 0.5, so each examiee is regarded to have made cosistet decisios i the preset experimet.
IJCSNS Iteratioal Joural of Computer Sciece ad Network Security, VO.6 No.7A, July 006 4 Soud pressure level Cosistecy Idex Time-differece eft Right eft Right A 0.3 0.9 0.4 0.068 B 0. 0.065 0.060 0.074 C 0.077 0.09 0.00 0.087 D 0.0 0.03 0.053 0.058 E 0.090 0.4 0.090 0.093 Avg. 0.04 0.088 0.083 0.076 5. Coclusios Whe soud pressure was poor i the moaural presetatio, the recogitio rate was much lower for recitatio backgroud soud tha for white oise. It may be that the questios were igored because attetio was paid to the backgroud recitatio. I the moaural presetatio, the experimet used ceilig loudspeaker at the 4 corers i the classroom. It will be possible to raise the degree of recogitio of the active soud, if the stereo loudspeaker are istalled i idividual machies. The presece of 3D soud allows for more accurate recogitio of the active soud. However, the effect of reflected soud of the classroom ad voice of other pupil earby, etc. were sigificat factors i the test case. Therefore, though i additio, we eed research i classroom desig, o could also use stereo systems, 5. style stereo ad virtual reality systems i combiatio with PC i the persoal relatios. This time, we coducted the experimet usig headphoes i order to preset the ideal eviromet. The recogitio rate of the stereo presetatio was far higher tha that of the moaural presetatio. Therefore, this may be a effective way to achieve as a biaural system. Geerally, i most of the curret systems the images presetatio a big scree, ad the voice is moaural. Such distace educatio system has bee itroduced i great umbers. However, such eviromet should be chaged allow for the voice to be stereo rather tha moaural. The teacher could raise the recogitio rate of the pupil, if this were doe. I the case of multisite commuicatio, the face of the pupils from the sites are displayed o the scree. I this case, the arragemet o this scree meas beig a simulated locatio. Therefore, it is ecessary to preset a soud image similar to the visual image. If the positio of the widow o the display is made to accord with the stereo image, the by usig the coordiatio of the visual sese ad the auditory sese, the degree of recogitio of the active soud will rise. I short, a system which also moves stereo soud image whe the widow is moved o the display would be a effective tool for multisite commuicatio. Next experimet, the soud image ad the weight(ratio) appear to be i logarithmic relatio. Ad the timedifferece ad the weight(ratio) appear to be i liear relatio. This is said almost applicable to the case of soud like a geeral relatio betwee itesities of the physical stimulatio ad the resulted sesatio, kow as Weber- Fecher's law. Besides, cosistecy idex calculated for every examiee was below 0.5 throughout the experimet. I the AHP, criterio of acceptace (good) is defied as the cosistecy idex beig less tha 0.5, so each examiee is regarded to have made cosistet decisios i the preset experimet. It was foud that there was a possibility of estimatig the relatio betwee the soud pressure level ad the time-differece despite persoal characteristics of each examiee.
4 IJCSNS Iteratioal Joural of Computer Sciece ad Network Security, VO.6 No.7A, July 006 Refereces [] asaori SHINOHARA, Hiroyuki SAITO, Satoru WADA, Yasuhisa SAGAWA ad Fumiori HIGASHINO: "Distace learig at schools from museums usig telecoferece system", IEICE Techical Report, ET99-37, pp.33-39, (999), (i Japaese) [] asahiko SUGIOTO, Kohtatsu SAKUOTO, Hiroyuki SAKURAI, aabu ISHIHARA ad Yutaka SHIKATA: "Costructio of remote class system bu usig PC coferece system", The Japa Soc. for Sci. Educatio workshop research report, -5, pp.33-36, (998), (i Japaese) [3] Kohtatsu SAKUOTO, Hiroyuki SAKURAI, asahiko SUGIOTO, Hiroshi KURAOCHI, aabu ISHIHARA ad Yutaka SHIKATA: "Evaluatio of simultaeous operatio betwee facig class ad remote class by usig video coferece", The Japa Soc. for Sci. Educatio workshop research report, 3-6, pp.33-44, (999), (i Japaese) [4] asahiko SUGIOTO, Kohtatsu SAKUOTO, Hiroyuki SAKURAI, aabu ISHIHARA, Kazutaka SUGIOTO ad Yutaka SHIKATA: "Practice of Computer Sciece Educatio for Studets who refuse to go to School by usig PC Coferece System", The Japa Soc. for Sci. Educatio workshop research report, 4-, pp.7-, (999), (i Japaese) [5] asahiko SUGIOTO, Kohtatsu SAKUOTO, Hiroyuki SAKURAI, aabu ISHIHARA, Keigo NOFUJI, ad Yutaka SHIKATA: " Commuicatio - ethod of Distace Educatio System for ay Studets ad Oe Teacher", Iteratioal Ergoomics Symposium 000, pp.85-88, (000) [6] asahiko SUGIOTO, Kohtatsu SAKUOTO, Hiroyuki SAKURAI, aabu ISHIHARA, Keigo NOFUJI, ad Yutaka SHIKATA: A Distace Educatio System Commuicatio ethod for Poit-to-ultipoit, Proceedig of Seveth Wester Pacific Regioal Acoustics Coferece, pp.405-408(oct.000) [7] aabu ISHIHARA, Ju SHIRATAKI, Keji OHSHIA, Yukio KOBAYASHI ad Shi-osuke SUZUKI: Soud Effect by Time-Differece ad Soud Pressure evel, Proceedig of Nith Wester Pacific Regioal Acoustics Coferece, ID440 (pp.-7)(jue.006) aabu Ishihara received the Ph.D. degrees from eisei Uiversity, Tokyo, Japa i 987, respectively. Now, He was a associate professor i the Departmet of Electrical ad Computer Egieerig, Oyama Natioal College of Techology. His research iterests iclude multimedia, computer etwork, huma egieerig. He is a member of the IEEE, the IEICE, the IEE of Jp. ad ASJ.