SELECTION OF ACTIVE SPEAKER(S) IN VoIP CONFERENCE BRIDGES: FROM LINEAR DOMAIN TO CELP PARAMETERS DOMAIN

摘要

This paper presents alternative approaches to select the mixed channels during teleconferencing involving CELP CoDecs. The proposals address the problems related to complexity and delay when classical solutions based on PCM samples are used. The principle consists of avoiding total speech decoding and to extrapolate the speech audio level based on CELP parameters, before channels selection. Only the selected speaker's bit streams are completely decoded for mixing and thus leading to very basic processing. With these proposals, the computational load is reduced as the number of participants increases, enabling more resources to extend the capacity of the conference service. Reduction of complexity also leads to processing delay decreases which is important for real time applications.