Speech power estimation with a truncated normal distribution

Abstract

Telephone speech power estimation and related issues in the satellite FDM transmission network are described. The capacity of an FDM transmission system in terms of number of voice channels is closely related to the multichannel loading. Speech is still a major part of the traffic in present day systems. Its characterization is thus an important step, especially in the design of the load-sensitive FDM/SSB/AM satellite transmission system. This paper presents a truncated normal distribution and its application to speech power estimation in the presence of channel and background noise. With the provision of a thresholding mechanism, the new truncated normal distribution is especially suited for the estimation of speech power in the presence of channel noise. It is further suggested that, even in the absence of noise, the truncated normal distribution is in general a more realistic model than the normal distribution.