Related Experiment Videos
The effects of digital quantization error on speech intelligibility and perceived speech quality
R W Harris1, R H Brey, Y S Chang
1Communication Sciences and Disorders, Brigham Young University.
Journal of Speech and Hearing Research
|February 1, 1991
Summary
Digital quantization error impacts speech quality. Higher bit integers (8-16) and floating-point conversions (2-7 bits) maintained speech intelligibility, but perceived quality varied.
Area of Science:
- Digital Signal Processing
- Acoustics
- Speech Perception
Background:
- Digital audio systems rely on quantization to represent analog signals.
- Quantization error can potentially degrade audio fidelity, affecting speech intelligibility and quality.
Purpose of the Study:
- To investigate the impact of digital quantization error on speech intelligibility and perceived quality.
- To determine the minimum bit depth for integer and floating-point representations to maintain speech fidelity.
Main Methods:
- Digitized speech was processed using various bit depths for integer (6-16 bits) and floating-point (2-7 bits) conversions.
- Speech intelligibility was assessed using standard tests.
- Perceived speech quality was evaluated through subjective listening experiments.
Main Results:
- Speech intelligibility remained largely unaffected for 8-16 bit integer and 2-7 bit floating-point conversions.
- Only 6-bit integer conversion at 55 dB SPL significantly degraded intelligibility.
- Perceived quality was more sensitive; higher bit integers (12-16) and floating-point (5-7 bits) were judged superior.
Conclusions:
- Speech intelligibility is robust to quantization error within typical digital audio bit depths.
- Perceived speech quality is more susceptible to quantization artifacts, with higher bit depths yielding better subjective results.