Browse Prior Art Database

Phoneme-Class-Based Switch for Selecting Speech-Coding Techniques/Parameters

IP.com Disclosure Number: IPCOM000063375D
Original Publication Date: 1985-Mar-01
Included in the Prior Art Database: 2005-Feb-18

Publishing Venue

IBM

Related People

Authors:
Thomas, JC [+details]

Abstract

A wide variety of speech-coding/decoding techniques have been developed over the last few decades. These have various advantages in terms of bit rates and/or particular types of speech sounds that they tend to reproduce relatively well or poorly. The concept described here provides a general method for optimizing the speech coding technique selected and parameters based on a decision about which the phoneme CLASS is being transmitted. The useful compromise proposed here is to decide on the phoneme CLASS (narrower classes than voiced/unvoiced but considerably broader than an exact phoneme decision). While it may be quite difficult to distinguish, for example, "mmm" from "nnn" from "ng" in a simple way, it is much easier to distinguish nasal from non-nasal sounds.