Improving intelligibility of a 300 b/s segment vocoder

The problem of increasing the intelligibility of a 300 b/s segment vocoder is addressed by investigating: (1) new line spectral pair (LSP)-based distance measures and (2) new structures and construction methods for segment codebooks. A variety of new distance measures is evaluated, and it is found that, after tuning, all of the distance measures provide almost equal intelligibility, indicating that some other factor, such as codebook template quality, is limiting performance. In an effort to improve the codebook, multiple duration-dependent codebooks constructed by selecting phonetically labeled segments from the TIMIT database are examined.<
>