We describe the design, implementation, and analysis of a speech intelligibility test. The test included five codec modes, four frame-erasure rates, and two background noise environments, for a total of 40 conditions. The test protocol required twenty listeners to repeat all words that they heard in short messages with median length of seven words. Each condition was tested using approximately 1100 words total. Listeners’ responses were scored against the original message transcripts to produce a count of words correctly repeated and thus a measure of speech intelligibility. We present results that show exactly how this measure of speech intelligibility drops as frame-erasure rate increases for three of the five codec modes. The remaining two codec modes did not produce valid results due to defects in the reference software provided to us.
Keywords: background noise; speech coding; packet loss; speech intelligibility; audio coding; frame erasures; acoustic noise