Related Experiment Videos
The effect of noise spectrum on speech recognition performance-intensity functions
G A Studebaker1, R Taylor, R L Sherbecoe
1Memphis State University, Memphis Speech and Hearing Center, TN 38105.
Journal of Speech and Hearing Research
|April 1, 1994
Summary
This study shows how different noise maskers affect speech recognition. Masking noise significantly alters the performance-intensity function slope, impacting speech intelligibility predictions.
Area of Science:
- Auditory perception
- Speech processing
- Psychoacoustics
Background:
- Articulation theory posits that auditory thresholds influence speech recognition.
- Understanding the performance-intensity (P-I) function slope is crucial for audiological models.
Purpose of the Study:
- To investigate the impact of various masking noise types on the speech recognition P-I function slope.
- To validate an articulation index model's predictive accuracy under different masking conditions.
Main Methods:
- Obtained P-I functions for 12 normal-hearing subjects using Technisonic Studios W-22 recordings.
- Employed four continuous thermal noise maskers: high-pass (HP), white, ANSI, and talker-spectrum-matched (TSM) noise.
- Analyzed P-I function slopes and compared them against an articulation index model.
Main Results:
- P-I function slopes varied significantly, from 1.6%/dB in HP noise to 6.7%/dB in TSM noise.
- An articulation index model accurately estimated P-I function position and slope at low-to-moderate intensities.
- The model overestimated performance at higher speech intensity levels.
Conclusions:
- Masking noise characteristics demonstrably alter the slope of the speech recognition P-I function.
- The articulation index model provides a useful, though not perfect, predictor of speech recognition performance across different masking conditions.
- Further refinement of audiological models is needed to account for performance at higher speech intensities.