Application of Speaker Modification Techniques to Phonetic Vocoding

The goal of the work described in this paper is to develop a very low bit rate vocoding scheme. The vocoder is a typical LPC vocoder, whose parameters are post-processed on a phone-byphone basis, resulting in a variable bit rate segment vocoder. Given the well known speaker recognizability problems presented by vocoders at such low bit rates, we have attempted to integrate a speaker modification method based on altering the formant frequencies and bandwidths of vowel segments. This is done by transmitting the mean value and standard deviation of the radius and angle of the poles corresponding to formant frequencies for each phone. In the decoder stage, the phone index is used to retrieve a set of normalized values from a codebook of `typical' phones. This set is speaker adapted to preserve the static characteristics (average and standard deviation) but relies in the typical phone to represent the dynamic characteristics such as formant trajectories. 1. INTRODUCTION Segment or Phonetic..