Subjective Performance of Selected Speech Coders in the Presence of Channel Errors

Abstract

Speech intelligibility data of several voice coding techniques suitable for use in an air-to-air RF digital communication link is presented. The voice encoding techniques analyzed and their respective bit rates were as follows: adaptive differential pulse code modulation (ADPCM) at 24 kbps, continuously variable slope delta modulation (CVSD) at 16 kbps, sub-band coding (SBC) at 16 kbps, time domain harmonic scaling with sub-band coding (TDHS/SBC) at 9.6 kbps, and linear predictive coding (LPC-10) at 2.4 kbps. The RF digital communication link was simulated by random bit error insertion between the voice encoder and voice decoder. The resulting speech intelligibility performance was measured subjectively using the Modified Rhyme Test (MRT). Data graphs depicting speech intelligibility versus bit error rate (BER) are presented. This data shows that ADPCM and CVSD are most tolerant of bit errors, followed by SBC, TDHS/SBC, and LPC-10 in that order.