1 / 20

Spatial unmasking of nearby pure-tone sources in a simulated anechoic environment

Spatial unmasking of nearby pure-tone sources in a simulated anechoic environment. Norbert Kopco Hearing Research Center and Cognitive and Neural Systems Dept Boston University. Objectives.

bluma
Télécharger la présentation

Spatial unmasking of nearby pure-tone sources in a simulated anechoic environment

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Spatial unmasking of nearby pure-tone sources in a simulated anechoic environment Norbert Kopco Hearing Research Center and Cognitive and Neural Systems Dept Boston University

  2. Objectives • How do unique acoustical properties of sources near the head influence listener’s ability to detect masked sounds? (monaural vs. binaural cues, informational masking, distance-related cues) • Can available models of auditory processing explain observed performance? • Baseline for future studies of unmasking of nearby complex/speech sources and reverberation. Spatial unmasking of nearby pure-tone sources

  3. Overview • Background • Methods • Data - Spatial Unmasking • Monaural and Binaural Factors • Binaural Unmasking - Analysis/Modeling • Independence of Responses in Adaptive Procedures Spatial unmasking of nearby pure-tone sources

  4. Background • Spatial separation of target from noise improves target detectability and intelligibility. • Typically studied under two experimental conditions: • headphone studies: • address question: “What acoustic (monaural or binaural) cues are used to detect masked signal and how (by what mechanisms) are the cues extracted?” • plenty of data (summarized in Durlach & Colburn, 1978), many conditions • mostly done under fairly artificial conditions (flat noise spectrum, no noise ILD) • data predicted well from models of binaural auditory processing • studies in real environment: • address questions:“Are non-acoustic (informational) cues (e.g., perceived direction) important for detection of masked signals?”“What is relative importance of different cues (and the mechanisms that extract them) in real listening situations?” • relatively few, studying unmasking of:- pure tones (Ebata et al., 1968; Santon, 1987; Doll et al., 1992),- noise bursts (Saberi et al., 1991; Good et al., 1997), - auditory patterns (Kidd et al., 1998), and - speech (Freyman et al., 1999; Hawley et al., 1999) Spatial unmasking of nearby pure-tone sources

  5. Background • Acoustic cues for nearby sources: • large ILDs at all frequencies • small positional changes cause large cue changes • ILDs vary with both direction and distance • analyzed by Brungart et al. (1998), Shinn-Cunningham et al. (2000) • Current state in understanding of spatial unmasking: • past studies results can be explained by a combination of acoustic (monaural, binaural) and informational (un)masking • no studies of spatial unmasking for nearby sources • no quantitative modeling of spatial unmasking • no studies of unmasking as a function of source distance • available headphone data • cannot be used to predict unmasking at all spatial combinations of tone and masker (few data with masker ILD) • no complete set of data for single subjects available Spatial unmasking of nearby pure-tone sources

  6. Present study • Goals: • Measure spatial unmasking of pure tones for nearby sources in simulated anechoic auditory space • Compare relative importance of monaural and binaural processing for unmasking • Study significance of the distance dimension for spatial unmasking (only monaural effects as in far field, or also binaural effects?) • Compare behavioral data with predictions of available models • Relation to previous studies • of spatial unmasking: • virtual auditory environment  better control of parameters  possible modeling • weaker percept of spatial position • headphone studies of binaural unmasking (and modeling): • masker shaped by HRTF  better spatial percept • masker spectrum not flat  possible challenge for modeling Spatial unmasking of nearby pure-tone sources

  7. Methods • Simulated anechoic auditory spacenear the listener • 60 target/masker configurations • individually measured HRTFs • T: 500 or 1000 Hz, 165 ms • M: 250 ms white noise, lowpass at 5000 Hz • TDT equipment • 3-down-1-up adaptive procedure (79.4%) • 3 repetitions for each measurement (+ additional as needed) • 3 blocks of 6 sessions, each session has 10 runs Spatial unmasking of nearby pure-tone sources

  8. Spatial Unmasking Spatial unmasking of nearby pure-tone sources

  9. Monaural Factors Spatial unmasking of nearby pure-tone sources

  10. Binaural Factors Spatial unmasking of nearby pure-tone sources

  11. Binaural Unmasking - Data Spatial unmasking of nearby pure-tone sources

  12. Binaural Unmasking at 500 Hz Spatial unmasking of nearby pure-tone sources

  13. Binaural Unmasking - Stern & Shear Spatial unmasking of nearby pure-tone sources

  14. p and R function at 500 Hz Spatial unmasking of nearby pure-tone sources

  15. Binaural Unmasking - Stern&Shear fit Spatial unmasking of nearby pure-tone sources

  16. Independence of Responses • One of assumptions in adaptive procedures is that there is no dependence between responses in consecutive trials • During the experiment, subjects react differently depending on feedback: • negative feedback worsens performance • Possible reasons: • change in subject’s concentration depending on feedback • subject’s learning of the adaptive procedure • Is the effect significant? • How does it influence the measured thresholds? Spatial unmasking of nearby pure-tone sources

  17. t-test of Significance • 1. For every spatial configuration estimate psychometric function at points with sufficient number of measurements • 2. Compute difference between psychometric functions given correct vs. incorrect response for all spatial configs and levels Spatial unmasking of nearby pure-tone sources

  18. t-test of Significance • H0: E{ p(c | previous c) - p(c | previous not c)} = 0 • N M1-M2 tobs p • 500 Hz S1 405 3.66 2.185 0.015 REJECT • 500 Hz S2 383 0.86 0.533 0.297 • 500 Hz S3 370 0.80 0.480 0.316 • 1000 Hz S1 407 3.88 2.259 0.012 REJECT • 1000 Hz S2 388 0.17 0.108 0.457 • 1000 Hz S3 343 -4.74 -3.098 0.001 REJECT • 500 Hz x-S 1158 1.82 1.907 0.028 REJECT • 1000 Hz x-S 1138 0.02 0.018 0.493 • x-freq x-subj 2296 0.93 1.381 0.084 • Test how does the largest observed difference influence the measured thresholds. Spatial unmasking of nearby pure-tone sources

  19. Simulation of response dependence • The value of swas estimated to be 10.1. • Simulation of subj S3 on 1000 Hz: M1-M2=-4.74, s=10.1, 10000 trials. • Feedback-determined difference of 5% leads to error in determined threshold smaller • than 0.5 dB. Spatial unmasking of nearby pure-tone sources

  20. Conclusions • from spatial unmasking point of view: • large spatial unmasking for nearby anechoic sources (-15 to 40 dB) • can be modeled very accurately using available models of auditory processing • source distance influences amount of binaural unmasking of nearby sources • monaural (better-ear) effects prevail (25 dB) • binaural processing important at lower frequencies and for masker near midline (10 dB) • binaural unmasking comparable for near and far sources • from point of view of modeling binaural auditory processing: • original Colburn (1977) p(ITD,fc) function better than Stern & Shear (1996) version at 500 Hz • to fit the 1000 Hz data Stern & Shear (1996) version of p(ITD,fc) function, or R(ILD) function, needs to be refined • if masker spectrum doesn’t deviate dramatically from flat shape, it can be sufficiently approximated by f0-centered ERB (error within 1dB) • transformed adaptive procedures: • assumption of independence between responses can be often broken, but this has small influence on the measured thresholds (up to 0.5 dB in this study) • future work: • unmasking of complex sounds/speech, effect of reverberation Spatial unmasking of nearby pure-tone sources

More Related