~ ~ ~~- 2~~~~~~~1~commun ickons, Universitk d; Quebec, 16 ~laie du commerce, ~erdun, Canada-H3E 1H6 Abstract: In this paper, our objective is to devise a fidelity criterion for quantifying the degree of distortion in- troduced by a speech coder. Towards this end, both original speech and its coded version are transformed from the time- domain to a perceptual-domain using a cochlear model. This perceptual-domain representation provides information per- taining to the probability-of-firings in the neural channels. We introduce a cochlear discrimination measure which com- pares these firing probabilit.ies in an information-theoretic sense. This measure, in essence, evaluates the neural-firing cross-entropy of the coded speech with respect to that of the original one. The performance of this objective measure is compared with subjective evaluation results. 1 Introductioxi Distortion measure plays a vital role in the evaluation as well as in the design of a low bit-rate speech coder. The measurement of distortion involves devising a transforma- tion operator for mapping the signals onto an appropriate domain and formulating a suitable comparison in that do- main. In speech communication, the ultimate recipient of information is a human being and hence histher percep- tual abilities govern the precision with which speech data must be processed and transmitted. In this article, we pro- pose a fidelity criterion using a filter bank approach for coded/distorted speech signals. Details of cochlear (inner ear) and other auditory processing involved in the speech perception are imbibed for the transformation of speech sig- nals onto a perceptual-domain. Subsequently, these percep- tual domain parameters of the original and the coded speech signals are compared in an information-theoretic sense. Sec- tion 2 briefly discusses the auditory system. Section 3 de- scribes an electrical model featuring auditory processing and defines the perceptual domain. Section 4 introduces the idea of Cochlear Di~crimination, a perceptual cross-entropy measure-based fidelity criterion, for speech signals. Finally, Section 5 provides the test results with relevant remarks.
[1]
S. Greenberg,et al.
The ear as a speech analyzer
,
1988
.
[2]
Richard E. Blahut,et al.
Hypothesis testing and information theory
,
1974,
IEEE Trans. Inf. Theory.
[3]
K. Nakata.
Synthesis and Perception of Nasal Consonants
,
1959
.
[4]
János Aczél,et al.
Some Recent Results on characterizations of Measures of Information Related to Coding
,
1978,
IEEE Trans. Inf. Theory.
[5]
D. V. Widder.
Some Recent Results on Characterizations of Measures of Information Related to Coding
,
1978
.
[6]
Richard F. Lyon,et al.
A computational model of filtering, detection, and compression in the cochlea
,
1982,
ICASSP.
[7]
J. Allen,et al.
Cochlear modeling
,
1985,
IEEE ASSP Magazine.
[8]
Alfréd Rényi,et al.
Probability Theory
,
1970
.
[9]
S. Kullback,et al.
Information Theory and Statistics
,
1959
.