1 / 17

A Binaural Model of Monotic Level Discrimination

A Binaural Model of Monotic Level Discrimination. Daniel E. Shub and H. Steven Colburn Boston University, Hearing Research Center Harvard-MIT, Health Science and Technology. Introduction.

raven
Download Presentation

A Binaural Model of Monotic Level Discrimination

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. A Binaural Model of Monotic Level Discrimination Daniel E. Shub and H. Steven Colburn Boston University, Hearing Research Center Harvard-MIT, Health Science and Technology

  2. Introduction • Monaural level discrimination can be degraded by the addition of a second ear [Rowland and Tobias JSHR 1967, Bernstein JASA 2005, and Shub and Colburn ARO 2004] • Traditional models cannot predict this degradation • Understanding this degradation might be important for bilateral hearing aids and cochlear implants

  3. Outline • Psychophysical experiment • One-interval left-ear level discrimination task with a roving contralateral distractor • Extremely limited amount of published data on degradation from the other ear • Previous studies used multi-interval adaptive paradigms which increases model complexity • Predict results with a detection theoretic model based on binaural information

  4. Psychophysical Experiment Distractor tone Roving level: 50-80 dB Roving phase: ± 90° Target tone Level: 50 or 58 dB Fixed phase: 0° • Task is to detect a level increment of a monaural target in the presence of a simultaneous but contra-aural distractor • Both target and distractor are 600 Hz tones with 300 ms duration • 1-interval, 2-alternative-forced-choice with feedback paradigm • Without distractor: traditional monotic level discrimination task

  5. Stimulus Perception • Dominant perception: Single image with a salient loudness and position • Target level affects loudness and position • Distractor level affects loudness and position • Distractor phase affects position • Additional “fragile” images • Time image, image shape/width

  6. Psychophysical Results Distractor Lags Leads Pf Pd • Overall Performance: • With distractor: 73% correct • No distractor: 97% correct (not shown) • Responded “Incremented” more with intense and lagging distractors

  7. Detection Theoretic Model • Observe: • Analysis is currently limited to zero-mean Gaussian noise which is independent across the dimensions • Variances of the internal noise and value of trading ratio k are fit to previous level discrimination and lateralization experiments

  8. Ideal Observer • Ideal (maximum likelihood) observer: • Achieves 99% correct discrimination performance • Decision rule is defined by a complex surface • Divides 3-D space into regions of “Incremented” and “Un-Incremented” • Small  and large  fall into the “Un-Incremented” region • For some Q and T, there are no values of  which fall into the “Incremented” region • ,  and  carry too much information •  and  do not carry sufficient information • Consider non-ideal observer of ,  and 

  9. Non-Ideal Observer • Non-ideal observer modifies the ideal rule: • Responds “Incremented” for large , independent of  and  • Assumes subjects always respond “Incremented” whenever a “loud” stimulus is heard • Decision rule (criterion) is jittered (zero-mean Gaussian noise) to further decrease discrimination performance • Assumes subjects have difficulties implementing the multidimensional decision space; imposes cost for complex decisions • Non-ideal observer has four free parameters • Variances of the criterion jitter (sL, sQ, sT) • The  for which the response is always “Incremented” (threshold)

  10. Minimum RMS Error Predications RMS error of 11% 60% of the variance was accounted for

  11. Maximum Variance Accounted For RMS error was 17% 73% of the variance was accounted for

  12. Comparison Psychophysical Data RMS error: 11% 60% of variance accounted for Minimum RMS error predictions Captures mean, but not shape RMS error: 17% 73% of variance accounted for Maximum variance accounted for predictions Captures shape, but not mean

  13. Monotic Level Discrimination • Under monotic conditions our model is a monaural energy detector • Level discrimination models are often more complex than simple energy detectors • Our model could be modified such that under monotic conditions it reduces to these “better” models of level discrimination • Modified models have not been evaluated • Current model is run on a 54-processor supercomputer

  14. Conclusions • Monaural level discrimination can be degraded by the “other” ear • Ideal observer of two dimensions is degraded • 2D model does not predict the data accurately • Ideal observer of “loudness”, “position” and “time-image” is NOT degraded by the “other” ear • Non-ideal observer of the three dimensions predicts a large proportion of the variance of the data

  15. Acknowledgments NIH NIDCD DC00100 and DC004663 Binaural Gang at Boston University Nat Durlach

  16. Why Three Dimensions? • Observe: • The ideal observer of two dimensions has an RMS error of 37%

  17. 2-D Model Predictions RMS error was 37% and variance was added Visually a completely wrong fit

More Related