Analysis and Subjective Evaluation of Differential Pulse-Code Modulation Voice Communication Systems

Differential pulse-code modulation (DPCM) has been proposed as a practical encoding scheme for speech, television, and telemetry signals. In this paper, the mean-square error for DPCM systems is derived, and its minimization by proper choice of system variables is examined. A simple analytical method is then presented for optimizing the system bandwidth
W
, the number of quantization bits
L
, and the parameters of the DPCM feedback network when the input is a speech signal and channel noise is negligible. A method is then developed for measuring the subjective quality of voice communication systems as a function of an arbitrary number of system parameters. The method was used to measure the quality of PCM speech and DPCM speech as a function of
W
and
L
. Quantization was logarithmic, the sampling rate
f_{s} = 2.2W
, and channel noise was assumed negligible. Optimum
W
and
L
were found for every bit rate
R = 2.2WL
. For both PCM and DPCM, the equations relating the optimum
W
and
L
are of the form
W = a2^{bL}
, where
a
and
b
are constants independent of
R
. The maximum speech quality obtainable for PCM and DPCM was found as a function of
R
, and the reduction in bit rate resulting when optimum DPCM rather than optimum PCM is used was determined as a function of
R
and as a function of the speech quality. Included in the paper is a brief description of the realtime DPCM systems used in the research.