Acoustic annoyance inside aircraft cabins A listening test approach Lena SCHELL-MAJOOR ; Robert MORES Fraunhofer IDMT, Hör-, Sprach- und Audiotechnologie & Cluster of Excellence Hearing4All, Oldenburg University of Applied Sciences Hamburg, Hamburg ABSTRACT The Speech Interference Level 3H (L SIL ) is widely used for the evaluation of the acoustic comfort aboard jet-engine aircrafts. The aim of this study is to explore the potential of L SIL to represent the acoustic annoyance of sounds inside aircraft cabins and to compare its performance to L Aeq and loudness. In a listening test, untrained listeners had to rate the annoyance of 5 real noise samples. The ratings are mapped to a relative annoyance scale and compared to the corresponding L SIL, L Aeq and loudness values. The results show that L SIL does not represent annoyance. L SIL yields a correlation of only 0.58, while L Aeq and loudness yield correlations of 0.8 and 0.93, respectively. Therefore, loudness seems to be a useful parameter for evaluating the acoustic comfort in aircraft cabins. Keywords: Annoyance, Loudness, Comfort I-INCE Classification of Subjects Number(s): 63., 63., 63.7. INTRODUCTION Comfort issues are increasingly important in aircraft engineering especially for VIP-aircrafts. Major factors for comfort and well-being aboard an aircraft are seat comfort, motion, air quality and noise (, ). In the aircraft industry, the established parameter for the evaluation of acoustic comfort is the Speech Interference Level 3H (L SIL ). It was originally developed as a measure for the impact of noise on speech communication and is defined as the arithmetic average of the sound pressure levels in the khz-, khz- and 4 khz-octave bands (3, 4). Due to this definition, numerical values for L SIL are always lower than those for other commonly used measures such as the sound pressure level (SPL) or the A-weighted sound pressure level (L Aeq ), when applied to the same signal. Therefore, its use has been appreciated from a marketing point of view until now. However, for the purpose of increasing cabin comfort in the future, applied noise measures have to proof their potential of representing human perception. Pennig et al. (5), e.g., already pointed out, that comfort and well-being was significantly decreased with rising L Aeq. One potential alternative measure could also be the psychoacoustic measure loudness. Studies on the sound quality of other objects revealed that this measure strongly influences the perceived quality of a sound (6, 7). An influence of loudness on annoyance ratings was also found in a study on exterior aircraft noise (8). As one affective reaction to noise is annoyance and the term annoyance is commonly used in investigations on auditory perception and sound quality (9, 0), the concept of annoyance is also applied for the study presented in this paper. In the few existing studies on human perception of aircraft interior noise and on effects of such noise on humans the relation between perceived acoustic comfort or annoyance and neither L SIL nor loudness has been addressed so far (, 9, ). Therefore, the aim of this study is to explore the potential of L SIL to represent the acoustic annoyance of sounds inside aircraft cabins and to compare its performance to L Aeq and loudness according to ISO 53-975 (method B). Lena.Schell-Majoor@idmt.fraunhofer.de Robert.Mores@haw-hamburg.de 4633
. METHOD. Subjects non-expert subjects participated voluntarily in the test. There were twelve male and nine female subjects with an average age of 9.7 years (standard deviation σ = 7.5 years) and all of them had travelled by airplane before. Two subjects reported hearing impairments (Tinnitus, slight hearing impairment). Their data was included in the analysis because no peculiarities could be found. The other subjects reported normal hearing abilities.. Stimuli T The main test included 5 stimuli which were recorded during in-flight measurements on different positions aboard seven different jet-engine aircrafts. The samples were chosen to map a total range of ΔL SIL = 8 db and cover the full range from very silent VIP areas to standard areas in aircraft cabins without any interior installations. All samples were monaurally recorded using the same handheld measurement device (B&K 50 Handheld Analyzer) with a sample rate of 48 khz and had a length of ten seconds. Out of the 05 potentially possible pairs from 5 noise samples, two sets of 30 pairs have been defined, following a fractional factorial design paradigm (Table). Four pairs were used in both sets which led to a total of 56 pairs evaluated in the test. Pair Table Pairs of sound samples used in this study Group Group Group Group Pair # x x x x7 #6 x6 x7 x4 x3 # x x5 x x0 #7 x6 x0 x5 x0 #3 x x x x #8 x7 x8 x5 x #4 x x5 x x #9 x7 x x5 x3 #5 x x3 x x7 #0 x8 x9 x5 x4 #6 x x6 x x8 # x8 x x6 x7 #7 x x3 x x9 # x9 x0 x6 x #8 x3 x4 x x #3 x9 x3 x6 x5 #9 x3 x7 x3 x8 #4 x0 x x7 x3 #0 x3 x4 x3 x9 #5 x0 x4 x8 x3 # x4 x5 x3 x0 #6 x x x8 x4 # x4 x8 x3 x #7 x x5 x9 x5 #3 x4 x5 x4 x6 #8 x x3 x x5 #4 x5 x6 x4 x9 #9 x3 x4 x x4 #5 x5 x9 x4 x0 #30 x4 x5 x4 x5.3 Test apparatus and procedure Stimuli were presented in pairs via one loudspeaker (Mackie HR84) placed visibly in a distance of.9 m in front of the subject. The subjects were randomly assigned to one of the two sets of noise samples described above. The order of the pairs and the assignment of sample and sample were randomized as well. Before the test each subject was given written instructions for the test procedure and a questionnaire retrieving some statistical data. During the test subjects were alone in the test room and had to complete the sentence is than by making a choice on a 7-point scale (Table ). The original scale from () was adapted and translated into German. A 4634
graphical user interface was applied for this test which allowed the subject to start and to repeat the samples as necessary for the assessment. By clicking a next-button, the assessment of the next pair of noise samples was loaded without a way back to previous assessments. After 5 pairs each subject had to take a break for one minute followed by 5 more pairs. The sound level in the listening room was adjusted to match the L SIL values of the sound samples measured in flight with a predefined tolerance of ±0.5 db. In order to relate the results of the test to the actual physical parameters, L SIL, ns and L Aeq were measured in the test situation and these values are used for all analyses. Table 7-point scale used for the rating of the pairs of noise samples -3 - - 0 3 much less less slightly less just as slightly more more much more 3. RESULTS AND DISCUSSION 3. Relative annoyance a i The relative annoyance a i of each noise sample i was calculated via the arithmetic average of all ratings a ij which include sample i for each subject as shown in equation (): a i N N j a ij () N is the number of pairs that include noise sample i, j is a running number for the samples which are compared to sample i. It is giving a value between +3 and -3 inclusive. From these values the most noise sample is x3 with a i =.76 and x4 is the least with a i = -.. The standard deviation varies between σ min = 0.49 (x) and σ max =.4 (x) and there is a tendency towards larger deviation for the noise samples ranked in the middle range. This uncertainty was expected because subjects tend to agree on extreme sounds (very high or very low annoyance) whereas they might follow individual guidelines while assessing sounds of comparable annoyance. 3. Annoyance vs. L SIL Figure displays relative annoyance over L SIL for each noise sample. The results suggest a slight tendency that a i increases with increasing L SIL. However, there are pairs of noise samples where the L SIL values have a large difference and the relative annoyance is barely different, such as x5, x9 and x. On the other hand, there are pairs, where similar L SIL values belong to noise samples where the annoyance ratings have rather large discrepancies, such as x4 vs. x5, or x8 vs. x3. The correlation coefficient yields only a medium value of r = 0.58 (p = 0.04). These results indicate that L SIL is not suitable for the evaluation of the acoustic annoyance inside aircraft cabins. L SIL does not account for any perceptual effects and neglects parts of the audible frequency range. Therefore, it was not expected to be good annoyance measure. 4635
Figure - Mean relative annoyance and standard deviations versus L SIL for noise samples x to x5 3.3 Annoyance vs. n s Figure shows the relative annoyance a i plotted over loudness n s. In general it shows that increasing loudness leads to increasing annoyance. The correlation coefficient, r = 0.93 (p < 0.0), confirms this finding. For some noise samples this relation is not true, e.g., x0 has a loudness of n s = 40.3 sone and is more (a i = 0.8) than x (a i = 0.45) which has a larger loudness of n s = 47.6 sone. This effect is found mainly for noise samples with medium annoyance and large standard deviations. Loudness seems to be able to represent the annoyance of the noise samples reasonably well. This is in line with findings from other studies: Fastl and Zwicker (3) found that loudness is contributing significantly to the perception of pleasantness, but there are more psychoacoustic parameters influencing this perception, for example sharpness and roughness. Sharpness and roughness but also tonality are contained in the samples in different ways and might explain some of the deviations from a perfect linear relation between the annoyance ratings and ns. Figure - Mean relative annoyance and standard deviations versus loudness for noise samples x to x5 3.4 Annoyance vs. L Aeq In Figure 3 the relative annoyance is related to L Aeq. The diagram and the correlation coefficient 4636
(r = 0.8, p < 0.0) lead to the assumption that L Aeq represents the annoyance better than L SIL but the correlation is not as strong as for loudness. Due to the fact, that the L Aeq includes characteristics of the human auditory system it was expected to represent the perception better than L SIL. Still, L Aeq is a technical parameter and by definition valid only for low levels (< 0 sone). Applied to relatively loud cabin noise samples low frequencies are systematically underestimated (4). Furthermore, the noise samples are not distributed equally over the range of L Aeq. The noise samples cover a range of ΔL Aeq =. db between the lowest and highest values (L Aeq,min = 70.9 db and L Aeq,max = 83. db), but six of the noise samples have an L Aeq between 76.9 db and 78.9 db. Figure 3 - Mean relative annoyance and standard deviations versus L Aeq for noise samples x to x5 4. CONCLUSION The main result of this study is that L SIL is not the adequate choice for the evaluation of the annoyance of noise inside jet-engine aircraft cabins. With only a medium correlation (correlation coefficient r = 0.58, p = 0.04) it is not able to represent the annoyance of potential passengers in the listening test. Additionally, it was shown that loudness represents the perceived annoyance much better than the L SIL (r = 0.93, p < 0.0) whereas the L Aeq is ranked in between (r = 0.8, p < 0.0). Therefore it is suggested to use ns for the evaluation of the noise aboard aircrafts within the aircraft industry. Generally, the results match the findings from earlier research that loudness and annoyance are strongly correlated. However, there are still influences on the perceived relative annoyance which are not covered by ns. Further investigation should deal with the influence of other psychoacoustic parameters especially sharpness and roughness, because those are also known to contribute to the perceived annoyance of sounds. They might also be helpful to investigate the different internal criteria which presumably led to large standard deviations in relative annoyance for noise samples of comparably annoyance. REFERENCES. Richards LG, Jacobson ID. Ride Quality Assessment III: Questionnaire Results of a Second Flight Programme. Ergonomics. 977;0(5):499-59.. Mellert V, Baumann I, Freese N, Weber R. Investigation of noise and vibration impact on aircraft crew, studied in an aircraft simulator. Proc INTER-NOISE 04, -5 August 004; Prague, Czech Republic 004. 3. Webster J. Updating and Interpreting the Speech Interference Level (SIL). Journal of the Audio Engineering Society. 970;8():4-8. 4. Webster J. Psychoacoustic problems with noise control. 5st Convention of the Audio Engineering Society 975, Preprint No. 08 (G-3); 975. 4637
5. Pennig S, Quehl J, Rolny V. Effects of aircraft cabin noise on passenger comfort. Ergonomics. 0;55:5-65. 6. Fastl H. quality of electric razors effects of loudness. Proc. of INTER-NOISE 00; 7-30 August 000; Nice, France 000. p. 445-3. 7. Altinsoy E, Kanca G, Belek HT. A comparative study on the sound quality of wet-and-dry type vacuum cleaners. Proc ICSV 999; 5-8 July 999; Copenhagen, Denmark 999. p. 3079-86. 8. More SR. Aircraft Noise Characteristics and Metrics. PhD Thesis, Purdue University, 00. 9. Västfjäll D. Affective Reactions to Interior Aircraft s. Acta Acustica United with Acustica. 003;89:693-70. 0. Guski R. Psychological Methods for Evaluating Quality and Assessing Acoustic Information. Acta Acustica. 997;83:765-774.. Quehl J. Comfort studies on aircraft interior sound and vibration. PhD Thesis, University of Oldenburg 0.. Bech S, Zacharov N. Perceptual Audio Evaluation Theory, Method and Application. Chichester, England: John Wiley & Sons; 006. 3. Fastl H, Zwicker E. Psychoacoustics - Facts and Models. 3rd ed. Heidelberg, Germany: Springer; 007. 4. Fastl H. Psychoakustische Methoden. In: Kalivoda MT, Steiner JW, editors. Taschenbuch der angewandten Psychoakustik, Wien, Austria:Springer, Wien, 998. 4638