TABLE OF CONTENTS The mobile-telephone voice signal is essentially a sequence of audible sounds varying at a syllabic rate, with occasional random pauses of silence. This time-varying waveform associated with human speech (200 to 3000 Hz) is not easy to characterize, since the speaker's pauses between phrases and sentences result in a concentration of speech energy in "talk spurts" of about 1-s average duration, separated by random gaps of varying lengths. Thus, the resulting speech signal consists of randomly spaced bursts of energy of random duration. As a consequence, accurate means for analyzing the characteristics of the speech signal are very difficult to define. The following paragraphs will identify and describe certain characteristics of speech waves that can be used to measure the voice-quality performance of a mobile radiotelephone system.
The primary characteristics of the speech wave are (1) the distribution of average speech power, (2) the instantaneous amplitude probability distribution of speech power, and (3) the power spectrum.
The distribution of average speech power among talkers closely follows the lognormal law [1]. Let S be the average speech power of a particular talker, S r some arbitrary reference power, and ? = S/ S r . Then the speech volume V can be expressed:
| (14.1) |  |
The probability distribution for V less than V 1, based on Eq. (2.45), can be expressed as:
| (14.2) |  |
where
| (14.3) |  |
| (14.4) |  |
The quality V is the power that is...
