A Subjective Comparison of Selected Digital Codecs for Speech

Abstract

Technological advances are continually increasing the economic viability of efficient codecs in telephone networks. A subjective evaluation is described here of the μ255 Pulse Code Modulation (PCM) algorithm and three more efficient techniques, Nearly Instantaneous Companding PCM (NIC PCM), Cummiskey-Jayant-Flanagan Adaptive Differential PCM (ADPCM), and Subscriber Loop Carrier Adaptive Delta Modulation (SLC∗ ADM). These codecs are compared under the conditions of: (i) single encodings as a function of line bit rate, input level, received volume, and error rate, (ii) tandem encodings with intermediate baseband conversion, and (iii) local, exchange, and toll network reference connections where mixed tandem encodings might be found along with typical analog impairments such as loss and random noise. The simulation of the codec algorithms on a minicomputer facility enabled the production of subjective test tapes containing speech processed under these conditions. These tapes were then evaluated in listening-type subjective tests. It is shown that (i) NIC PCM, ADPCM, and SLC ADM have approximately a 12- to 16-kb/s advantage over μ255 PCM for equivalent subjective ratings, (ii) NIC PCM, ADPCM, and SLC ADM perform comparably over the range of conditions tested; and (iii) 64-kb/s μ255 PCM can be deployed in a multiple encoding environment with very few restrictions, whereas the use of lower bit rate NIC PCM, ADPCM, and SLC ADM codecs would necessitate more stringent application rules to avoid excessive degradation in tandem encoding situations.