Related Experiment Videos
Effect of peak clipping on speech recognition threshold
1Department of Communication Disorders, University of Minnesota, Minneapolis.
Ear and Hearing
|December 1, 1994
Summary
Peak clipping significantly increased speech recognition thresholds (SRTs) in both hearing-impaired and normal-hearing individuals, while compression did not. Distortion, not waveform changes, caused these shifts, suggesting coherence analysis can predict intelligibility impacts.
Area of Science:
- Audiology
- Speech-Language Pathology
- Signal Processing
Background:
- Speech processing techniques like peak clipping and compression are used to alter audio signals.
- Understanding their impact on speech intelligibility is crucial for hearing aid development and audiological assessments.
- Previous research has explored various signal modifications, but the specific effects of distortion products versus envelope alterations require further clarification.
Purpose of the Study:
- To investigate the effects of peak clipping and compression on speech recognition thresholds (SRTs).
- To determine whether distortion products or waveform envelope changes are primarily responsible for altered SRTs.
- To evaluate the utility of magnitude-squared coherence analysis in predicting speech intelligibility under signal distortion.
Main Methods:
- Speech recognition thresholds (SRTs) were measured for spondee words processed with varying levels of peak clipping and compression.
- Magnitude-squared coherence analysis was performed on the processed speech stimuli to quantify distortion.
- Subjective sound quality judgments were collected for the processed stimuli.
Main Results:
- Both hearing-impaired and normal-hearing listeners exhibited significantly elevated SRTs with increasing peak clipping levels (above 18-24 dB).
- Compression processing did not result in significantly elevated SRTs for either listener group.
- Magnitude-squared coherence analysis revealed high distortion from peak clipping and low distortion from compression, correlating with SRT shifts.
Conclusions:
- Peak clipping introduces significant distortion that degrades speech recognition, whereas compression does not.
- The observed SRT shifts are primarily attributable to the addition of distortion products, not alterations in the speech envelope.
- Magnitude-squared coherence analysis shows promise as a tool for predicting the impact of distortion on speech intelligibility.