Voice Over Internet Protocol: Speech Intelligibility Assessment.

摘要

In this study, we examined the speech intelligibility of five codecs and parameter settings to determine their suitability for Air Traffic Control use. These codecs (g711r64, g726r32, g729r8, g723r53, and g726r16) convert analog communications to digital communications. We used standard intelligibility tests - the Modified Rhyme Test (MRT) and the Message Completion Task (MCT) - to examine intelligibility differences. For the MRT, codec type had an effect on intelligibility. Performance in an uncompressed audio condition was the best, followed by codecs g711r64 and g726r32, then g729r8, and then g723r53 and g726r16, respectively. Reaction times were consistent with the performance data, with the better performing codecs having shorter reaction times. Subjective intelligibility and acceptability ratings matched the performance and reaction time data. For the MCT, the intelligibility and acceptability ratings for all codecs were high, but we found no clear pattern in the performance data. We recommend that the National Airspace System Voice Switch (NVS) program further evaluate the three best performing codecs, g711r64, g726r32, and g729r8. These codecs delivered the best performance, the shortest reaction times, and the most positive feedback. We also recommend that the NVS Program Office investigate the effect of channel noise, ambient noise, and level of network utilization on the intelligibility of different codecs.