A Slope-Feedback Method for Speech Compression

This paper deals with a slope-feedback method for compressing the bandwidth of speech signals. The maximum amount of compression is estimated from a quantity called the speech detail factor. This is analogous to the familiar picture detail factor. Measurements of the former show that the following compression ratios are possible 1) a two-fold compression of "telephone" quality speech, 2) almost five-fold compression of band-filtered speech which has a syllable articulation of 60 per cent. A special feature of the coder is a variable-delay unit and the experimental version is described.