MULTI-SPEAKER CONFERENCING OVER NARROWBAND CHANNELS

A terminal is disclosed for use with a system for digital conferencing over narrowband channels which allows use by multiple simultaneous speakers and which allow for interfacing between systems which operate at different bit rates. For each conferee, the terminals interconnect between an A/P converter and a modem, and contain an analyzer and a synthesizer. The analyzer receives digitized speech and encodes it at the system data rate for transmission to the bridge by the modem. The synthesizer receives encoded digital data streams and produces therefrom artificial speech at selected data rates that are inversely proportional to the number of external speakers. One technique used by the synthesizer is the use of overlap-add synthesis which allows for a particularly simple method for synthesizing multiple speakers. With the over-lap add technique a speaker's parameters, pitch voicing, and spectrum, are used to fill out an FFT buffer. The inverse transform is taken and summed in the usual overlap-add manner. With two speakers, the two FFT buffers are added before taking the inverse transform. In this way synthesis of two speaker is no more complex than one speaker.