Voice compression systems for wireless telephony

This paper presents a comparative study between three voice compression systems for wireless telephony: CELP (code excited linear prediction), VSELP (vector sum excited linear prediction) and GSM 06. 10 (Global Standard for Mobile communications), and one system based on artificial neural networks (ANNs). The main result is that the system based on ANNs exceeds the best current performance standard (CELP). However, its speaker dependency hinders its potential standardization.