The Correlation Between Subjective and Objective Measures of Coded Speech Quality and Intelligibility Following Noise Corruption
Abstract
A scoring metric of speech intelligibility based on linear predictive coding (LPC) was developed and evaluated. The data base used for evaluating the metric consisted of a list of 50 words from the Modified Rhyme Test. The list was transmitted over a LPC-10 Vocoder with no background noise. The list was scored subjectively for intelligibility by a trained listener panel. The subjective scores were used to judge the effectiveness of the objective metric. The LPC scoring metric was calculated for the list of words and compared to the subjective scoring. The intelligibility score for the objective scoring metric was 82.99% with a standard deviation of 14.41%. The score for the subjective listener testing was 84.91% with a standard deviation of 7.47%. This shows a possible correlation between the objective LPC scoring metric and standard subjective listener scoring methods.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1981
- Accession Number
- ADA115515
Entities
People
- Jeffrey A. Kayser
Organizations
- Air Force Institute of Technology