Intensity Discrimination and Binaural Interaction

Transcription

1 Technical University of Denmark Intensity Discrimination and Binaural Interaction 2 nd semester project DTU Electrical Engineering Acoustic Technology Spring semester 2008 Group 5 Troels Schmidt Lindgreen s David Pelegrin Garcia s Eleftheria Georganti s Instructor Sébastien Santurette

2

3 Technical University of Denmark DTU Electrical Engineering Ørsteds Plads, Building Lyngby Denmark Telephone Title: Intensity Discrimination and Binaural Interaction Course: Auditory Signal Processing and Perception Spring semester 2008 Project group: 5 Participants: Troels Schmidt Lindgreen David Pelegrin Garcia Eleftheria Georganti Supervisor: Torsten Dau Instructor: Sébastien Santurette Date of exercise: April 10 th Synopsis: This report deals with intensity discrimination and binaural interaction of the auditory system. Measurements concerning difference limen took place. The existence of a binaural processor in the human auditory system is investigated and the effects of the interaural level differences (ILD) and the interaural time differences (ITD) are discussed. Test results show that humans have a kind of binaural processor and that it is possible to change the perception of the position of the source image by modifying the values of ITD and ILD. Pages: 23 Copies: 4 No part of this report may be published in any form without the consent of the writers.

4

5 Introduction This exercise deals with the intensity discrimination and binaural interaction of the auditory system. Various measurements, namely the intensity discrimination, the timeintensity trading and the binaural masking level difference experiments, are carried out. Monaural intensity discrimination ability is investigated. The intensity Just Noticeable Difference (JND) is determined and corresponds to the smallest intensity difference that the human auditory system can detect. The results are compared to the ones predicted by Weber s law for broadband noise stimuli. The existence of a binaural processor in the auditory system is examined by carrying out an intensity discrimination experiment. The results obtained empirically are compared to the ones using only monaural detection cues. In addition, an experiment that examines the effect of the interaural level differences, the interaural time differences, and the connection between them is performed. The different cues are investigated and their role in sound localization is discussed. Finally, the phenomenon of Binaural Masking Level Difference (BMLD) is investigated. The BMLD effect implying that the masking thresholds may be decreased if a specific binaural presentation is applied to the test subject is observed and discussed. Technical University of Denmark, April 21, 2008 Troels Schmidt Lindgreen David Pelegrin Garcia Eleftheria Georganti

6

7 Contents 1 Theory Intensity discrimination Just noticeable interaural level difference Interaural time difference Time-intensity trading Binaural masking level difference (BMLD) Results Intensity discrimination Time-intensity trading Binaural Masking Level Difference Discussion Interaural level discrimination Time-intensity trading Binaural masking level difference Conclusion 21 Bibliography 23 A Matlab code A1

8

9 April 2008 Chapter 1. Theory Chapter 1 Theory In general terms, Just Noticeable Difference (JND) is the smallest difference in a specified modality of sensory input that is detectable by a human being. It is also known as the difference limen or the differential threshold. 1.1 Intensity discrimination The smallest intensity difference a listener can detect is called intensity JND. Weber s law states that intensity JND ( I) is proportional to the intensity (I) of a sound. This is called the Weber fraction: I I = c [Dau et al., 2008, p.3] (1.1) Where: I is intensity JND. I is the intensity. c is a constant. Intensity JND can be expressed as a change in level at threshold ( L): ( ) I + I L = 10 log 10 I [Dau et al., 2008, p. 3] (1.2) Page 1

10 Chapter 1. Theory Technical University of Denmark Assuming the Weber fraction, (1.1), the threshold is equal for all levels: Where: k is a constant. ( L = 10 log I ) I = 10 log 10 (1 + c) (1.3) = k (1.4) It is obvious that (1.4) cannot hold for all levels. Low levels [Sone] and the level of pain sets limits for (1.4). It should be noted that Weber s law holds for wide-band noise above 20 db SL, but not for a 1 khz tone. The observation that Weber s law does not hold for tones is often called the near-miss to Weber s law. [Moore, 2004, p. 145] 1.2 Just noticeable interaural level difference Interaural Level Difference (ILD) is the difference in the sound pressure level of the signals arriving at both ears, which can provide sound localization cues. The effect of head diffraction and shadow on the ILD is sketched in figure 1.1 and is most pronounced for frequencies above 1500 Hz. Head and pinna diffraction attenuates sound at the far ear, while boosting the sound at near ear. Figure 1.1: Effect of head shadow on the ILD. Sound waves are emitted from a loudspeaker. Monaural intensity JND can be measured as described in section Assuming that intensity JND is identical for both ears and in the hypothetic case that humans have no gain in intensity JND from binaural hearing, the value of intensity JND is the same for monaural intensity JND and binaural intensity JND. If there is any gain by using both ears, the binaural intensity JND is less than for monaural intensity JND. Page 2 Intensity Discrimination and Binaural Interaction

11 April 2008 Chapter 1. Theory Whether humans benefit from binaural hearing regarding intensity JND can be investigated by presenting a stimuli in both ears where the sound level is lowered by half ILD in one ear and raised with half ILD in the other ear, as shown in figure 1.2. Assuming no benefit from binaural hearing regarding intensity JND, the intensity JND will be twice as high as the monaural intensity discrimination. This is because the ILD effect is divided to both ears and therefore the level change has to be twice as high in order to reach the same level change for each ear compared to receiving all the energy monaurally. x x? x 65 db SPL 65 db SPL + ILD/2 65 db SPL - ILD/2 Intensity JND = x If there is no improvement from binaural hearing: ILD = 2 x If there is an improvement from binaural hearing: ILD > 2 x Figure 1.2: Just noticeable monaural intensity discrimination and just noticeable interaural level difference. Taking the value of intensity JND for monaural intensity discrimination as 1, the value of intensity JND for binaural intensity discrimination should be 2 if there is no benefit of binaural hearing regarding intensity JND. On the other hand, if humans have some kind of binaural processor, there would benefits from binaural and the intensity JND will be lower. 1.3 Interaural time difference As commented in section 1.2, the level difference between the two ears (or interaural level difference, ILD) provides cues for localization of sounds in the horizontal plane. Another cue for localization is the interaural time difference (ITD), i.e. the delay existing between the signals arriving at the two ears, produced by a path difference between them, as can be observed in figure 1.3. Modeling the human head as a sphere of radius r, this delay can be described as a function of the angle of incidence θ using equation (1.5). ITD = rθ + r sin θ c (1.5) Where: c is the propagation speed of sound in the air. Page 3

12 Chapter 1. Theory Technical University of Denmark Figure 1.3: Sound arriving from a direction θ. Modeling the human head as a sphere of radius r, the path difference between the sound arriving at both ears can be described as d = rθ + r sin θ [Dau, 2008, p. 23]. 1.4 Time-intensity trading In the previous sections, it has been stated that small differences in level or delay of the signals at both ears make it possible to detect the incidence direction of the sound, i.e. the sound image is shifted from the median plane. This effect is also known as lateralization. However, if the same signal is presented to both ears (diotic stimulus), the sound appears to come from the center of the head. If the level of the signal at one ear is rised, the sound image shifts towards that side. It is also possible to shift the sound image by making one of the signals lead the other. In this case, the sound image shifts to the side where the signal is leading. Thus, it is possible to design an experiment where the sound image shift produced by level cues is compensated by interaural time delays in the opposite direction, so the sound appears to come from the median plane. This kind of experiment is called a trading experiment. The result of this experiment is the ratio between ITD and ILD, called compensation factor or trading ratio, i.e. the amount of delay that it is necessary to compensate (bring the sound image back to the center) for a level difference of 1 db. The trading ratio depends on the presentation level, and usually less ITD is necessary to compensate ILD at higher presentation levels [David et al., 1959], as shown in figure 1.4. There is one hypothesis (called Latency theory ) which states that intensity differences might be transformed into time differences at a certain stage of neural processing. However, some of the effects observed in this experiment, such as increase of spaciousness or perception of separate images, prove that both cues, ITD and ILD, are not equivalent at all [Dau et al., 2008]. Page 4 Intensity Discrimination and Binaural Interaction

13 April 2008 Chapter 1. Theory Figure 1.4: Trading curves of a typical subject for broadband clicks at various levels. The curves drawn through the plots were obtained using third-degree polynomial fits [David et al., 1959]. 1.5 Binaural masking level difference (BMLD) The changes in interaural level and time differences that were examined in the previous section offer valuable cues to improve sound localization. Another interesting aspect of binaural process is that of the Binaural Masking Level Difference (BMLD). This process implies that a test subject will have remarkably lower masking threshold when he/she makes use of binaural information, than when making use of monaural information cues, at certain presentation conditions. An interesting example showing this is that of a listener detecting sounds in a noisy environment (low SNR) by making use of binaural interactions. In order to derive a definition for the BMLD an example is given. One can consider the situation that is shown in figure 1.5, where a masker and a tone signal are presented to a test subject via earphones. In figure 1.5 (b) white noise from the same sound generator is fed to both ears. Additionally, a pure tone, from the same signal generator is fed separately to each ear and mixed with the noise. In this point, it is assumed that the level of the tone is adjusted so that it can be just masked by the noise, and presenting a level of L 0 db. In this case, the total signals at the two ears are identical and the tone is inaudible. In figure 1.5 (c) the situation is slightly different. The tone signal at one of the Page 5

14 Chapter 1. Theory Technical University of Denmark Figure 1.5: Illustration of two situations in which binaural masking level differences (BMLD s) occur. In situation (a) a tone signal is being presented to the test subject (reference). In condition (b) the detectability of the tone is poor, while in condition (c), where the interaural relations of the signal and masker are different, detectability is good [Moore, 2004, p. 258]. ears is inverted, by adding a phase shift of π radians. The tone now becomes audible again. The tone can be adjusted to a new level L π, so that it is once again at its masked threshold. In this point the definition of BMLD can be introduced, which is the difference between the two levels, L 0 L π [db]. Its value may be as large as 15 db at low frequencies (around 500 Hz), decreasing to 2-3 db for frequencies above 1500 Hz. [Moore, 2004, p.258] Figure 1.6: Illustration of situations in which binaural masking level differences (BMLD s) occur. In condition (a) the detectability of the tone is poor when noise and signal are fed to one ear only. In condition (b) the same noise is added at the other ear, and the tone becomes audible again. [Moore, 2004, p. 258]. Another interesting example of the effects of BMLD can be seen in figure 1.6, where at first noise and signal are fed to one ear only (figure 1.6 (a)), and the signal is adjusted Page 6 Intensity Discrimination and Binaural Interaction

15 April 2008 Chapter 1. Theory to be at its masked threshold, thus inaudible. Then, if the same noise is added at the other ear (figure 1.6 (b)), the tone becomes audible again. This implies that by adding noise at the ear where no signal is present the tone becomes more detectable. This noise, however should be correlated with the masker at the other ear. Uncorrelated maskers will not result in BMLD. It is important to note that BMLD phenomenon is not limited to pure tone signals but it is also observed for complex signals too (complex tones, clicks and speech sounds). Over the years, many different models have been proposed to account for the various aspects of binaural processing [Moore, 2004, p.263]. Durlach presented a rather simple model that was able to describe the BMLD phenomenon. This model was based on the Equalization - Cancellation (EC) concept. That is based on the idea that the auditory system tries to eliminate the masker by manipulating the stimuli between the two ears, until the masker components are equal (equalization) and then subtracts them so that they cancel each other (cancellation) [Dau et al., 2008]. The first step of this model (equalization process) involves amplification or attenuation of the stimuli and shifting in time. This is because the model needs to compensate for the interaural level differences (ILD) and interaural time differences (ITD). In order to take into consideration the finite limit of the BMLD phenomenon the model assumes an imperfect equalization process. This assumption is realized by introducing two sources of error: jitter in the amplification (ɛ) and shifting in time (δ). ɛ and δ are asummed to be statistically independent random variables, with zero mean and variances σ 2 ɛ and σ 2 δ respectively. Under those assumptions, Durlach [Dau et al., 2008] was able to show that the difference between the N 0 S 0 and N 0 S π condition can be described by the following equation: where: 2e ω2 0 σ2 δ BMLD(N 0 S 0 N 0 S π ) = 10 log (1.6) 1 + σ 2 ɛ e ω2 0 σ2 δ ω 0 is the center frequency (in radians) of the auditory filter with the stimulus. equation (1.6) it is possible to note that for very low frequencies: ( lim ω0 0BMLD(N 0 S 0 N 0 S π ) = 10 log ) σ 2 ɛ From (1.7) Page 7

16 Chapter 1. Theory Technical University of Denmark Figure 1.7: BMLD (N 0 S 0 N 0 S π ) as a function of ω 0. Data points are from 18 different studies. The curve corresponds to equation (1.6) with σ δ = µs and σ ɛ = [Dau et al., 2008] and that at very high frequencies: lim ω0 BMLD(N 0 S 0 N 0 S π ) = 10 log 10 (1) = 0 (1.8) Equations (1.7) and (1.8) show that at very low frequencies BMLD depends on σ ɛ. For a value of σ ɛ = 0.25, equation (1.7) results in 15 db of BMLD. Equation (1.8), on the other hand, implies that at high frequencies the BMLD phenomenon is not observed. The modeled BMLD curve is better illustrated in figure 1.7, where data points from 18 different studies are plotted together with the Durlach s model estimation. In figure 1.7 it can be seen that the curve indicating the BMLD level seems to fall to zero for frequencies around 3 khz. On the other hand at the same frequency range the empirical data seem to converge asymptotically to a value around 3 db. Durlach suggested that a possible error source for this discrepancy might be due to the various mathematical operations of the model. For example a calculation that corresponds to the averaging over the equalization errors before forming the SNR at the output of the EC mechanism might cause this discrepancy. [Durlach, 1963]. Page 8 Intensity Discrimination and Binaural Interaction

17 April 2008 Chapter 2. Results Chapter 2 Results 2.1 Intensity discrimination Monaural intensity discrimination Monaural intensity JND is measured for three test subjects using bandpass noise from 250 Hz to 4 khz, in a period of 500 ms at two presentation levels: 65 db SPL and 75 db SPL. A Matlab script with an adaptive three-interval, three-alternative, forced-choice procedure with one-up, two-down tracking rule is used for this experiment. Results are shown in figure 2.1. intjnd 3 dp eg JND [db] 2 1 tl level [db SPL] Figure 2.1: Monaural intensity discrimination. Page 9

18 Chapter 2. Results Technical University of Denmark The results in figure 2.1 are shown as L as a function of frequency. In figure 2.2 the same results are shown as the Weber fraction (see (1.1) on page 1) as a function of level. The results in figure 2.2 are calculated by rearranging equation (1.2) to: I I = 10 L/10 1 (2.1) dp eg tl I/I Level [db SPL] Figure 2.2: Weber fraction calculated from results shown in figure Binaural intensity discrimination Binaural intensity JND is measured for three test subjects using bandpass noise from 250 Hz to 4 khz, in a period of 500 ms at 65 db SPL. The stimulus is presented in both ears where the sound level is lowered by half ILD in the right ear and raised with half ILD in the left ear, as shown in figure 1.2. A Matlab script with an adaptive three-interval, three-alternative, forced-choice procedure with one-up, two-down tracking rule is used for this experiment. Results are shown in figure 2.3. Page 10 Intensity Discrimination and Binaural Interaction

19 April 2008 Chapter 2. Results JND [db] 4 2 dp eg tl reference level [db SPL] Figure 2.3: Binaural intensity discrimination. 2.2 Time-intensity trading A trading experiment was performed in order to measure the amount of delay (ITD) that is necessary to bring the sound image to the center when a level difference (ILD) was applied to one ear. The method used was an adaptive one-up, one-down rule in combination with a two-interval, two-alternative, forced choice procedure implemented in a Matlab script. The signals in the two intervals were the same, but the channels were reversed in the second one (the left channel in the first interval was presented as the right channel in the second one, and vice versa). The stimulus signal was a binaural click train comprised of ten five-sample clicks separated 2,200 samples at a sampling frequency of 44.1 khz. At one ear, the signal was amplified ILD/2 whereas at the other ear, an attenuation of ILD/2 was applied. Moreover, one channel was advanced half ITD (which was the tracking variable) and the other one was delayed half ITD. At each presentation the test subject had to say which side the sound image had moved towards. Three test subjects (DP, EG and TL) took the experiment. The ITD was measured for seven different ILDs (-10, -7, -4, 0, 4, 7 and 10 db) at two different presentation levels (L p ): 45 db SPL and 60 db SPL. For each test subject and presentation level, a thirdorder polynomial was used to fit the data. The measured data and the fitted polynomials for test subjects DP, EG and TL are shown in figures 2.4, 2.5 and 2.6, respectively. The values of the trading ratio and the bias calculated as the first and zero order coefficients of the fitted polynomials fitted from each measurement are shown in table 2.1. Page 11

20 Chapter 2. Results Technical University of Denmark Measurements Fitted curve Measurements Fitted curve 1 1 ITD [ms] 0 ITD [ms] interaural level difference [db] interaural level difference [db] (a) L p = 45 db SPL (b) L p = 60 db SPL Figure 2.4: Results of the time-intensity trade experiment for test subject DP at presentation levels of 45 db SPL (a) and 60 db SPL (b), from measured data and fitted third-order polynomials Measurements Fitted curve Measurements Fitted curve 1 1 ITD [ms] 0 ITD [ms] interaural level difference [db] interaural level difference [db] (a) L p = 45 db SPL (b) L p = 60 db SPL Figure 2.5: Results of the time-intensity trade experiment for test subject EG at presentation levels of 45 db SPL (a) and 60 db SPL (b), from measured data and fitted third-order polynomials Measurements Fitted curve Measurements Fitted curve 1 1 ITD [ms] 0 ITD [ms] interaural level difference [db] interaural level difference [db] (a) L p = 45 db SPL (b) L p = 60 db SPL Figure 2.6: Results of the time-intensity trade experiment for test subject TL at presentation levels of 45 db SPL (a) and 60 db SPL (b), from measured data and fitted third-order polynomials. Page 12 Intensity Discrimination and Binaural Interaction

21 April 2008 Chapter 2. Results 45 db SPL 60 db SPL Test subject Trading ratio [µs/db] Bias [µs] Trading ratio [µs/db] Bias [µs] DP EG TL Table 2.1: Values of the trading ratio and the bias, given by the first and zero order terms of the polynomials calculated fitting the data from the time-intensity trade experiment. 2.3 Binaural Masking Level Difference An experiment concerning the BMLD took place for three test subjects. In this experiment the masked threshold is measured with noise and signal presented in the same way at both ears. More precisely for this experiment the signals were 480 ms sinusoids with an interaural phase difference of 0 and 180. Signal frequencies of 250, 500, 1000 and 2000 Hz were used. The masker was a bandpass 500 ms noise (2 octaves wide, 65 db SPL) geometrically centered around the signal frequency. The presentation of the masker starts 10 ms before and ends 10 ms after the presentation of the signal. The masked thresholds in both the N 0 S 0 and N 0 S π conditions were measured using an adaptive three-interval, three-alternative, forced-choice procedure with an one-up, two-down tracking rule. In figures 2.7, 2.8 and 2.9 the results obtained from the experiment for three test subjects (EG, DP, TL) are presented. A matlab script called ECmodel was composed in order to implement equation 1.6 (see A). In figures 2.7, 2.8 and 2.9 apart from the empirical results, the theoretical BMLD curve by Durlach, that was calculated using equation (1.6), is also present (σ δ = 105µs and σ ɛ = 0.25). Finally, a curve showing the best fitting according to the results and the model is also shown. Page 13

22 Chapter 2. Results Technical University of Denmark 15 Measured data ECmodel σ δ = 105µs and σ ɛ = 0.25 ECmodel σ δ = 40.4µs and σ ɛ = 0.55 BMLD [db] Frequency [Hz] Figure 2.7: Data obtained from the BMLD experiment for test subject DP. Measured data only at 250, 500, 1000 and 2000 Hz. 15 BMLD [db] 10 5 Measured data ECmodel σ δ = 105µs and σ ɛ = 0.25 ECmodel σ δ = 61.5µs and σ ɛ = Frequency [Hz] Figure 2.8: Data obtained from the BMLD experiment for test subject TL. Measured data only at 250, 500, 1000 and 2000 Hz. Page 14 Intensity Discrimination and Binaural Interaction

23 April 2008 Chapter 2. Results 15 Measured data ECmodel σ δ = 105µs and σ ɛ = 0.25 ECmodel σ δ = 59.3µs and σ ɛ = 0.79 BMLD [db] Frequency [Hz] Figure 2.9: Data obtained from the BMLD experiment for test subject EG. Measured data only at 250, 500, 1000 and 2000 Hz. Page 15

24 Chapter 2. Results Technical University of Denmark Page 16 Intensity Discrimination and Binaural Interaction

25 April 2008 Chapter 3. Discussion Chapter 3 Discussion 3.1 Interaural level discrimination Weber s law states that the intensity JND should be the same for any levels as seen in equation (1.4). In figure 2.1 the results for monaural intensity JND as a function of two levels can be seen. It can be noted that equation (1.4) only holds for test subject TL. For the other two test subjects the JND became lower for 75 db SPL. This can perhaps be explained by the presence of fluctuating background noise and the reduced set of data (average values calculated from 2 measurements). In figure 3.1 the results from monaural and binaural intensity discrimination are compared. It can be seen that the binaural intensity JND is less than twice as high as the monaural intensity JND for each test subject. This proves that all three test subjects benefit from binaural hearing and have some kind of binaural processor. During the binaural intensity discrimination experiment the test subjects described the noise as one fused object. They also mentioned that the method by itself was not good. It is surprisingly difficult to press the right button (to indicate that the change was in the last interval) when the sound was moving to the left. A perhaps better way to perform the experiment is to have an instructor that will note in which interval the noise moves to the left, so the test subject only has to say interval one, two or three to indicate the interval and thereby preventing any confusions between left and right. 3.2 Time-intensity trading The results of the trading experiment are presented in figures 2.4, 2.5 and 2.6 in section 2.2. The variability of the data and the range of ITD values (between -2 ms and 2 Page 17

26 Chapter 3. Discussion Technical University of Denmark dp eg tl JND [db] Monaural intensity JND (x2) Binaural intensity JND Figure 3.1: Comparison of the JND obtained from monaural (multiplied by 2) and binaural intensity discrimination experiments at a presentation level of 65 db SPL. ms) seem to be in agreement with the results from [David et al., 1959], shown in figure 1.4. The curves are not symmetrical with respect to the origin, because the zero-order term of the polynomial used to fit the data has a value different from 0. This term indicates the ITD that is applied to perceive the source image at the median plane when the signals presented at both ears have the same level. The bias is negative at the different presentation levels, meaning that the perception of a sound image as centered does not correspond to sound events coming from the median plane, and this behavior is observed on the three test subjects. According to [David et al., 1959, p.177], the bias should be consistent among presentation levels (as it has been measured in this report) and the average of the bias among test subjects should be ideally 0. This last statement is not fulfilled in these measurements, likely due to the few test subjects used for the measurements (only 3). [David et al., 1959] suggested that the reason for the bias could be errors in the listener s centering judgment, imbalance in interaural sensitivity, or unspecific physiological sidedness imbalance. In addition, the authors thought about the correlation with the different development of the brain at both sides (as happens with the ability to write with one of the hands), but it was refused as there was no correlation between the sign of the bias and the hand that the test subjects use for writing (TL is the only left-handed). The first-order coefficient of the fitted polynomial gives the trading ratio. The values of this term for every test subject at all presentation levels are presented in table 2.1. For test subjects DP and TL, there is a slight decrease of the trading value when increasing the presentation level from 45 db SPL to 60 db SPL (-0.08 µs/db for DP and µs/db Page 18 Intensity Discrimination and Binaural Interaction

27 April 2008 Chapter 3. Discussion for TL). In contrast, the trading value increases for EG (0.25 µs/db) under the same conditions, differing from the predictions by [David et al., 1959], who found that the trading ratio decreased with increasing level. However, the obtained results should not lead to the conclusion that ITD and ILD are coded and interpreted by the brain in the same way, because the stimuli were not perceived as a single sound event coming from the center of the head. The test subjects described the effects as perception of two different sound events at both ears or sound image broadening instead of sound image displacement between the two presentations, i.e. having a clear lateralization. 3.3 Binaural masking level difference In figures 2.7, 2.8 and 2.9 the BMLD results for the three test subjects are plotted as a function of the stimulus frequency. In addition, the theoretical BMLD predicted by [David et al., 1959] with values σ δ = 105µs and σ ɛ = 0.25 is plotted. The curves with the parameters σ δ and σ ɛ that offered the best fitting for the actual measurements are also plotted in each figure. In figure 1.7 on page 8 the data obtained for 18 different test subjects deviate a lot from each other, resulting in a cloud of data points. Therefore, the expected variability among test subjects is high, and the equation (1.6) (also plotted in the same figure) just provides an approximation to the population average. The measurements for this report were obtained with just one run for each frequency of the three test subjects in a noisy environment. With these sets of data, it is not possible to get statistical values that give precise information about the average BMLD at a certain frequency and its deviation. However, test subject EG seems to have less benefit from BMLD as compared to the test subjects DP and TL. In figure 2.8 the test subject TL seems to have a dip at 1 khz. Prior to this report test subject TL has measured his audiogram which presents a dip in hearing threshold at 1 khz which is different for each ear. This could explain the lack of benefit from binaural hearing at this frequency and thereby the dip in BMLD. Nevertheless, more measurements are necessary in order to reduce the uncertainty and prove the latest hypothesis. Page 19

28 Chapter 3. Discussion Technical University of Denmark Page 20 Intensity Discrimination and Binaural Interaction

29 April 2008 Chapter 4. Conclusion Chapter 4 Conclusion In this report, the intensity discrimination and binaural interaction of the human auditory system have been discussed theoretically and from measurements. Intensity JND was measured both monaurally and binaurally. Time-intensity trading was investigated and additionally the effects of BMLD were examined. The reduction in the measured binaural JND with respect to the equivalent monaural JND show that humans have a kind of binaural processor, which takes advantage from binaural hearing detection cues. It is possible to change the perception of the position of the source image by modifying the values of ITD and ILD. ITD is not processed in the same way as ILD, as the stimuli in the experiment are not perceived as single sound events at a certain location, but as separated events. Results show that there is an improvement in the ability of detecting a signal in background noise when the noise signals presented at both ears are correlated and the signal to detect is presented to just one ear or to both in antiphase. This improvement, known as BMLD, presents values up to 15 db in these measurements. Page 21

30 Chapter 4. Conclusion Technical University of Denmark Page 22 Intensity Discrimination and Binaural Interaction

31 April 2008 Bibliography Bibliography [Dau, 2008] Dau, T. (2008). Intensity coding and binaural processing in the auditory system. Slide show. [Dau et al., 2008] Dau, T., Christiansen, T. U., and Santurette, S. (2008). Intensity Discrimination and Binaural Interaction edition. [David et al., 1959] David, E.E., J., Guttman, N., and van Bergeijk, W. (1959). Binaural interaction of high-frequency complex stimuli. Journal of the Acoustical Society of America, 31(6): [Durlach, 1963] Durlach, N. (1963). Equalization and cancellation theory of binaural masking-level differences. Acoustical Society of America Journal, 35(8): [Moore, 2004] Moore, B. C. J. (2004). An introduction to the psychology of hearing. Page 23

32 Bibliography Technical University of Denmark Page 24 Intensity Discrimination and Binaural Interaction

33 Project group 5 April 2008 Appendix A Matlab code 1 function vbmld = ECmodel ( vfreq, sigmad, sigmae ) 2 % ECmodel c a l c u l a t e s t h e BMLD (N0S0 N0Spi ) according to t h e EC model 3 % 4 % syntax : vbmld = ECmodel ( vfreq, sigmad, sigmae ) 5 % 6 % C a l c u l a t e s t h e BMLD o f a N0S0 and and N0Spi c o n f i g u r a t i o n according 7 % to t h e EC model ( see Eqn. 5.1 o f t h e e x e r c i s e guide ). 8 % 9 % i n p u t : vfreq c e n t e r f r e q u e n c y o f t h e a u d i t o r y f i l t e r with t h e 10 % t a r g e t s i g n a l ; can be g i v e n as a v e c t o r ( in Hz! ) 11 % sigmad standard d e v i a t i o n o f t h e random j i t t e r in time ( in s ) 12 % sigmae standard d e v i a t i o n o f t h e random j i t t e r in amplitude 13 % 14 % output : vbmld t h e b i n a u r a l masking l e v e l d i f f e r e n c e in db ; i f vfreq 15 % i s a vector, t h i s i s a l s o a v e c t o r 16 vbmld =10* log10 (1+2*exp( -((2* pi * vfreq ).^2) * sigmad ^2)./(1+ sigmae ^2 -exp( -((2* pi * vfreq ).^2) * sigmad ^2) )); Page A1