Improving intelligibility of a 300 b/s segment vocoder

Abstract

We address the problem of increasing the intelligibility of a 300 b/s segment vocoder by investigating: 1) new LSP-based distance measures and 2) new structures and construction methods for segment codebooks. We evaluate a variety of new distance measures and find 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, we examine multiple duration-dependent codebooks constructed by selecting phonetically-labelled segments from the TIMIT database.

Ultra Low Bit-Rate Speech Coding

“…As a result of not using segment to segment warping, [CL94] show that their method outperforms that of [SH88]. Moreover, [CL94] note that a related work Jeanrenaud and Peterson [JP91,PJV90] from BBN, is more close to their formulation, in the sense that [JP91] use duration dependent segment codebooks, which is populated (and designed by means of an iterative algorithm) with segments of variable lengths, but ensuring that it has sub-codebooks each with several segments of a fixed length. However, it should be noted that having a variable length segment codebook and allowing for warping based segment quantization allows a particular segment representing a particular acoustic realization of speech to quantize a varied number of input segments belonging to the same acoustic category; in the absence of such a warping of a particular segment in the codebook, the codebook is constrained to have segments of different lengths to match with the different realizations of the same acoustic category (i.e.…”