Proceedings of Meetings on Acoustics

Transcription

1 Proceedings of Meetings on Acoustics Volume 19, ICA 2013 Montreal Montreal, Canada 2-7 June 2013 Psychological and Physiological Acoustics Session 4aPPb: Binaural Hearing (Poster Session) 4aPPb18. Release from masking through spatial separation in distance in hearing impaired listeners Adam Westermann* and Jorg M. Buchholz *Corresponding author's address: National Acoustic Laboratories (NAL), Macquarie University, 2109, NSW, Australia, It is widely accepted that speech intelligibility improves as a speech signal and interfering masker are separated spatially in azimuth. In a previous study [Westermann et al. (2012), IHCON] a similarly strong improvement was found for normal hearing (NH) listeners when target and masker are separated in distance. In this study speech reception thresholds (SRTs) were measured for 16 hearing impaired (HI) listeners using the Listening in Spatialized Noise - Sentences Test (LiSN-S) and the Coordinate Response Measure (CRM). Acoustic scenarios were auralized via headphones using binaural room impulse responses recorded in an auditorium. In the first scenario the target was presented at a distance of 0.5 m from the center of the listeners head and the interferer at a distance of 0.5 m or 10 m. In a second setup the interferer s location was fixed and the target s location was varied. HI listeners showed a substantial release from masking as target and interferer were separated in distance. This effect was consistent for both LiSN-S and CRM, but less pronounced than for NH listeners. This study suggests that distance related cues play a significant role when listening in complex environments and are also to some extent available to HI listeners. Published by the Acoustical Society of America through the American Institute of Physics 2013 Acoustical Society of America [DOI: / ] Received 22 Jan 2013; published 2 Jun 2013 Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 1

2 INTRODUCTION Listening in complex real-life environments can be challenging, especially for hearing impaired (HI) listeners. The auditory system has various mechanisms to cope with these challenges, but their effectiveness can be reduced by a hearing impairment. One of these mechanisms is the ability to use the difference in horizontal angle of a target and a distractor to better understand the target. This is known as spatial release from masking (SRM; Kidd et al., 1998; Freyman et al., 1999; Best et al., 2006). SRM is typically measured by assessing the difference in speech reception thresholds (SRTs) for a frontal talker when the masker is positioned either in front (i.e., co-located) or to the sides of the listener (i.e., spatially separated). Initially, this effect was measured with normal hearing (NH) listeners in noise or speech babble (e.g. Bronkhorst, 2000), which mainly involves energetic masking. Later studies found this effect to be much more pronounced in the presence of speech maskers, which has generally been explained by informational masking in addition to energetic masking (Ihlefeld and Shinn-Cunningham, 2008). Several studies have investigated SRM in HI listeners (Marrone et al., 2008; Best et al., 2012; Glyde et al., 2012). Generally, HI listeners show decreased spatial benefit while co-located thresholds are similar between HI and NH listeners. There are three common explanations for the reduced spatial benefit. Firstly, it has been debated that HI listeners suffer from reduced spatial processing, i.e., the processing of interaural tume and level differences. However, several studies have shown that this can not account for the entire difference (e.g. Noble et al., 1997). Secondly, it has been claimed that the lack of spatial benefit is due to reduced release from informational masking (IM). However, studies aiming to isolate informational masking find that HI listeners are affected equally as NH listeners (Helfer and Freyman, 2008). Finally, studies by Best et al., 2012 point to a threshold floor caused solely by energetic masking (EM). In these studies, the hypothesis is that HI listeners have a higher threshold floor if only energetic masking is present (due to reduced audibility) and this limits their performance in scenarios which combine IM and EM (such as the SRM task). Distance perception is an aspect of localization which has been given considerably less attention than horizontal localization, in particular with respect to speech intelligibility measures. Where human horizontal localization is sensitive down to 1 degree (Blauert, 1996), distance appears to be a less salient measure and is often dominated by vision. For auditory distance perception several cues are available (for an overview see Zahorik et al., 2005). The most predominant cue is signal intensity. As distance increases, signal levels for omni-directional sound sources decrease according to the 1/r law (Kuttruff, 2000). This is especially relevant for speech signals, as listeners are able to estimate the source level from the applied vocal effort. When sounds are presented in reverberant environments, the auditory system can additionally use the signal s direct-to-reverberant ratio (DRR) to determine distance. When a signal is presented in a room the reflections give rise to reverberation. This reverberation energy is almost independent of position, and as the direct sound energy will decrease with distance the DRR decreases accordingly. Zahorik, 2002 measured DRR just-noticeable differences (JNDs) in NH listeners and found that this cue only provided a coarse estimate of distance as the lowest JNDs required a doubling of distance. In Akeroyd et al., 2007 similar DRR JNDs were measured for HI listeners. However, they found that HI listeners were not able to use DRR as a cue for distance perception. Therefore, they concluded that HI listeners mainly use intensity cues. In addition to intensity and DRR, other distance related cues include binaural interaural time and level differences (ITDs and ILDs, respectively) at very close distance and spectral cues from air absorption at very far distances. The effect of distance cues on SRM was measured in Brungart and Simpson, 2002, where a significant release was observed especially when the masker was similar to the target. However, they only measured for distances up to 1 m and thereby the related distance cues were mainly ILDs and ITDs. A previous study by the authors (Westermann et al., 2012) showed that NH Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 2

3 listeners are able to utilize cues arising from stimuli presented in rooms at further distances to achieve a significant improvement in speech intelligibility. Measuring speech reception thresholds, they found advantages of up to 10 db when the distance of the masker was changed from 0.5 m to 10 m with the target speech held at 0.5 m distance, and compensating for distance-related level changes. The present study applies similar methods to investigate if this (spatial) advantage is also available to HI listeners. The results may allow further understanding of why HI listeners have severe difficulties communicating in reverberant (multi-source) environments, and may also impact on modern hearing aid technologies, which often modify (or even remove) room reverberation and thus, potentially remove cues that are important for understanding speech in complex environments. METHOD Stimuli Two different speech corpuses were used in this experiment. Firstly, raw speech material was taken from the coordinate response measure (CRM) (Bolia et al., 2000), a corpus often used when measuring SRM. This corpus consists of sentences spoken by four male and four female speakers, but here only the four male speakers were used. Each sentence has the structure: Ready [call sign] go to [color] [number] now, with eight call-signs, four colors (red, green, blue and white) and eight numbers (1 through 8) resulting in 256 sentences for each speaker. Subjects were assigned a specific call-sign (here only Baron was used) and are asked to report the color and number corresponding to the speaker of their call-sign. As subjects are trying to distinguish only the color and number related to their assigned call-sign the additional speakers act as maskers. In addition to the speech maskers, a noise masker was used. This masker was realized by applying the Hilbert envelope of the speech maskers to noise with the same long-term spectrum of the speech maskers. The recordings are made so that the call-sign, color and number are approximately time aligned. Therefore, the CRM is rather sensitive to temporal changes between target and masker, which may be altered by the applied spatialization. Hence, in order to increase the general validity of the measured results the Listening in Spatialized Noise-Sentence Test (LiSN-S) was additionally used. In the LiSN-S a sentence recall task is used to measure the SRT in a continuous 2-talker background (Cameron and Dillon, 2007). Both the target and masker talkers are female. The background can be chosen to be either the same or a different speakers as the target. Here, only the different speakers were used. In order to minimize any effect that may result from differences in the signal s long-term spectra the long-term spectrum of the maskers were equalized to match the long-term spectra of the target for each of the measured spatial configurations. The long-term spectra of target and masker were determined using the Welch method and a 512 tap equalization FIR filter was designed and applied using MATLAB. This was done both for the CRM and LiSN-S. All (anechoic) speech signals were convolved with binaural room impulse responses (BRIRs) recorded in an auditorium using a B&K Head and Torso Simulator (HATS). The auditorium had a reverberation time of T 60 = 1.9s at 2 khz and a volume of approximately 1150m 3. A DynAudio BM6P 2-way loudspeaker was used as sound source to roughly approximate the directivity of a human speaker. The BRIRs were measured using logarithmic sweeps (see Muller and Massarani, 2001). The position of the HATS was kept constant and BRIRs were measured with the speaker at 0.5 m and 10 m distance directly in front of the HATS. Three different spatial configurations were tested here as shown in Fig. 1. Note that the labels of the spatial conditions show the applied distances as well as the masker type, i.e. M s10 isaspeechmaskerat10m distance and M n0.5 is a noise masker at 0.5 m distance. The delay introduced by the sound Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 3

4 propagating further was removed by time-aligning the direct sound component. M T T + M T M M 0.5 T 0.5 M 10 T 0.5 M 0.5 T 10 FIGURE 1: Schematic showing the spatial conditions depicting the target (T), masker (M), distance and spatial condition label. Procedures Experiments were carried out in a double-walled booth, using a MATLAB GUI, Sennheiser HD-215 circumaural headphones and a computer with a MATLAB GUI. For both the CRM and LiSN-S the masker level was kept at a root mean square (RMS) level of 55 db SPL for NH subjects and 60 db SPL for HI subjects, measured in a B&K type 4153 artificial ear, and before compensation for hearing loss. The target level was initially set to 67 db SPL and varied relative to the masker following a 1-up 1-down rule to thereby adaptively estimating the SRT. In order to (partly) compensate for audibility, linear amplification was applied according to the NAL-RP scheme (Dillon, 2001). First the LISN-S test was measured and then the CRM. Within each test the order of the conditions was randomized and all thresholds were repeated once. Testing each subject required a single session of 1.5 hours. After initial hearing screening and audiometry, the subjects were given verbal instruction read by the experimenter. Before the CRM experiment started training was performed to ensure familiarity with the GUI and understanding of the task. No training was applied for the LiSN-S test. Subjects Nine HI subjects (3 females and 6 males) participated (aged years, mean 67). All subjects had sloping, mild-to-moderate, sensorineural, symmetrical hearing losses, and all were experienced hearing aid users with English as native language. The individual audiograms as well as their mean value are shown in Fig. 1. All subjects were active participants from the NAL database and had significant experience with speech intelligibility tests. In order to allow comparison between the derived HI data with NH data, results from 13 normal hearing (NH) listeners (< 15 HL) were taken from Westermann et al., RESULTS The upper panel of Fig. 3a (circles) shows the SRTs measured for the HI subjects using the CRM corpus. The corresponding spatial advantage is shown in the lower panel. For reference purposes the NH data taken from Westermann et al., 2012 is additionally shown (diamonds). For a close target of 0.5 m, shifting the speech masker from 0.5 m (i.e., co-located condition M s0.5 ) to 10 m (i.e., spatially separated condition M s10 ) the SRT decreased by about 6 Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 4

5 0 10 Left ear Right ear Hearing threshold (db HL) k 2k 3k 4k 6k 8k k 2k 3k 4k 6k 8k Frequency (khz) FIGURE 2: Hearing thresholds of the 9 subjects shown together with the mean and standard deviation. db, i.e. the listener s performance is strongly improved. However, this improvement is significantly smaller than the average improvement of 10 db observed for the NH subjects. In contrast, when the masker is kept at a distance of 0.5 m and the target is moved to a distance of 10 m (i.e., considering the spatially separated condition M s0.5 T s10 )thesrt increased by about 5 db, i.e., the performance is significantly decreased. A decrease in performance is in qualitative agreement with some of the NH subjects, but a significant number of NH subjects still showed a clear improvement. This variation across NH subjects is indicated by the large standard deviation shown in Fig. 3a. When a noise masker is considered instead of a speech masker the SRTs for both spatially separated conditions are equally improved by 9 db. Additionally, the SRT improvement of the noise maskers is similar to that found with the speech masker in the M s10 condition. A two-way analysis of variance (ANOVA) shows both statistical significance for condition (F = 302, p < 0.001) and subject (F = 5.9, p < 0.001), but not for interaction. The SRTs of the LiSN-S are shown in Fig. 3b with the spatial advantage given in the lower panel. Similar to the CRM data the SRT for a target at 0.5 m decreases as the speech masker s distance is increased from 0.5 m to 10 m. With this corpus the spatial advantage is approximately 5 db. Again similar to the CRM data, when keeping the masker at 0.5 m and moving the target to 10 m large differences between subjects can be observed and the SRT for the HI subjects is increased (no data for NH subjects was measured in Westermann et al., 2012, for the M s0.5 T s10 condition). A two-way ANOVA show only statistical significance for condition (F = 42.2, p < 0.001). Overall, the HI data shows the same tendencies as the NH data in both the CRM and the LiSN-S, but the NH perform better in all conditions, especially in the M s10 condition. DISCUSSION The present study investigated SRM from spatial separation in distance in HI listeners. Generally, both the CRM and LiSN-S results show that increasing the distance of a masker results in a release from masking. The experimental data revealed that the SRT change produced by moving the speech masker further away (i.e., M s10 ) was similar to the effect of changing from a speech (i.e., M s0.5 ) to a noise masker (i.e., M n0.5 ). This observation may be explained by applying the concept of EM and IM. In the case of a noise masker only EM is involved, and very low thresholds are observed. When target and speech maskers are co-located (M s0.5 )higher Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 5

6 10 CRM LiSN S Speech masker Noise masker Speech masker Target to masker ratio (db) HI listeners NH listeners 10 Spatial Advantage (db) r 10n Distance (m) 10 M s0.5 M s10 M s0.5 T s10 M n0.5 M n0.5 T s10 M s0.5 M s10 M s0.5 T s10 Condition FIGURE 3: Top panels: Mean and across-subject standard deviation TMR at SRT (expressed per masker) for HI and NH listeners (circles and diamonds, respectively). Bottom panels: Mean and across-subject standard deviation of the spatial advantage (i.e. the difference between the related result and the M s0.5 condition). thresholds are caused by a combination of IM and EM (Freyman et al., 2001). Ihlefeld and Shinn-Cunningham, 2008 propose two main sources of IM: (1) difficulty segregating the target from the masker and/or (2) difficulty selecting which sound features are related to the target in presence of a mix of similar maskers. For both causes the common factor is target/masker confusions. If distance related cues diminish IM, it is likely because the changes to the masker aids the listener in resolving confusions. The similar results between the far masker (M s10 ) and noise masker (M n0.5 ) thereby suggest that the spatial separation in distance fully removes the IM and the SRT is determined from the EM. This in principle agrees with the concept proposed by Best et al., 2012, suggesting that maximum SRM is dictated by an EM "floor". Westermann et al., 2012 also measured CRM thresholds for 2 m and 5 m distances for NH listeners. These thresholds fall between M s0.5 and M s10, suggesting that for these distances residual IM is still present. Now considering the differences between HI and NH listeners, following the lines of Best et al., 2012 the reduced spatial benefit for HI listeners in not caused by reduced SRM, but by an higher EM floor which restricts the spatial benefit. In the case when the target was moved to 10 m and the speech masker was kept at 0.5 m (M s0.5 T s10 ), thresholds were strongly increased and a large variability across subjects could be observed which is not present in the other conditions. This behavior could be caused by several Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 6

7 factors. It is commonly acknowledged that strong reverberation leads to decreased intelligibility (Nábělek, 1993). Hence, the increased DRR of the target at 10 m could have caused the decreasing SRT. However, if this would be the case then the SRT for the M n0.5 T s10 condition should also be higher than for the M n0.5 condition, which is not the case. Alternatively, the increased SRT for the M s0.5 T s10 could have a more cognitive origin. Here the target is distorted by the reverberation, whereas the masker is clear and close to the listener. Some of the NH and HI subjects reported that in this condition it was especially hard to ignore the masker and to focus on the target, whereas other subjects reported that they were able to use the decreased DDR as a cue for selecting the target. Comparable large spreads in M s0.5 T s10 SRTs can be observed for the NH subjects, where some subjects obtained SRTs similar to the M s10 condition and other SRTs are similar to the co-located condition. Following this point of view, the increased thresholds together with their large spread may be caused by auditory attentive abilities or applied listening strategies. SUMMARY AND CONCLUSION Two speech tests, the CRM and LISN-S, were applied to measure the spatial release from masking in HI listeners introduced by separating a target source and a speech masker in distance. The main cues available to the listeners were reverberation-cues, which were introduced by applying measured binaural room impulse responses. The results revealed that HI listeners can take advantage of reverberation-based distance cues to achieve a significant release from masking, when the masker is moved further away from the target. However, the spatial benefit for HI subjects was found to be smaller than reported for NH subjects. It was proposed that this loss of performance could be caused by a higher energetic masking "floor". In the case that the target was moved further away the subjects performance was much poorer than for NH listeners and the thresholds varied largely between listeners. It was suggested that this behavior might be linked to cognitive abilities of the HI subjects or their applied listening strategies. In addition, this condition resulted in much higher thresholds than for NH listeners. Generally, there was a good agreement between the results of the CRM and the LiSN-S. Slightly higher SRTs were observed in the LiSN-S, which may be explained by the lower amount of informational masking associated with the LiSN-S corpus than with the CRM corpus. ACKNOWLEDGMENTS This work is funded by an International Macquarie University Research Excellence Scholarship (imqres) and Widex A/S. In addition, the conference participation is supported by the Australian Acoustical Society. REFERENCES Akeroyd, M., Gatehouse, S., and Blaschke, J. (2007). The detection of differences in the cues to distance by elderly hearing-impaired listeners, The Journal of the Acoustical Society of America 121, Best, V., Gallun, F., Ihlefeld, A., and Shinn-Cunningham, B. (2006). The influence of spatial separation on divided listening, The Journal of the Acoustical Society of America 120, Best, V., Marrone, N., Mason, C. R., and Kidd, Jr, G. (2012). The influence of non-spatial factors on measures of spatial release from masking., The Journal of the Acoustical Society of America 131, Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 7

8 Blauert, J. (1996). Spatial Hearing-Revised Edition: The Psychophysics of Human Sound Localization (MIT press). Bolia, R., Nelson, W., Ericson, M., and Simpson, B. (2000). A speech corpus for multitalker communications research, The Journal of the Acoustical Society of America 107, Bronkhorst, A. (2000). The cocktail party phenomenon: A review of research on speech intelligibility in multiple-talker conditions, Acta Acustica united with Acustica 86, Brungart, D. and Simpson, B. (2002). The effects of spatial separation in distance on the informational and energetic masking of a nearby speech signal, The Journal of the Acoustical Society of America 112, Cameron, S. and Dillon, H. (2007). Development of the listening in spatialized noise-sentences test (lisn-s), Ear and hearing 28, Dillon, H. (2001). Hearing aids (Thieme). Freyman, R., Balakrishnan, U., and Helfer, K. (2001). Spatial release from informational masking in speech recognition, The Journal of the Acoustical Society of America 109, Freyman, R., Helfer, K., McCall, D., and Clifton, R. (1999). The role of perceived spatial separation in the unmasking of speech, The Journal of the Acoustical Society of America 106, Glyde, H., Cameron, S., Dillon, H., Hickson, L., and Seeto, M. (2012). The effects of hearing impairment and aging on spatial processing, Ear and Hearing 34, Helfer, K. and Freyman, R. (2008). Aging and speech-on-speech masking, Ear and hearing 29, Ihlefeld, A. and Shinn-Cunningham, B. (2008). Spatial release from energetic and informational masking in a divided speech identification task, The Journal of the Acoustical Society of America 123, Kidd, G., Mason, C., Rohtla, T., and Deliwala, P. (1998). Release from masking due to spatial separation of sources in the identification of nonspeech auditory patterns, The Journal of the Acoustical Society of America 104, Kuttruff, H. (2000). Room acoustics (Taylor & Francis). Marrone, N., Mason, C., and Kidd Jr, G. (2008). The effects of hearing loss and age on the benefit of spatial separation between multiple talkers in reverberant rooms, The Journal of the Acoustical Society of America 124, Muller, S. and Massarani, P. (2001). Transfer-function measurement with sweeps, Journal of the Audio Engineering Society 49, Nábělek, A. (1993). Acoustical factors affecting hearing air performance, chapter Communication in noisy and reverberant environments, 3 14 (Allyn). Noble, W., Byrne, D., and Ter-Horst, K. (1997). Auditory localization, detection of spatial separateness, and speech hearing in noise by hearing impaired listeners., The Journal of the Acoustical Society of America 102, Westermann, A., Buchholz, J., and Dillon, H. (2012). Release from masking through spatial separation in distance, International Hearing Aid Conference (IHCON). Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 8

9 Zahorik, P. (2002). Direct-to-reverberant energy ratio sensitivity, The Journal of the Acoustical Society of America 112, Zahorik, P., Brungart, D., and Bronkhorst, A. (2005). Auditory distance perception in humans: A summary of past and present research, Acta Acustica united with Acustica 91, Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 9