Linear PCM signal processing for audio processing unit in multipoint video conferencing system

This paper describes the architecture and functions of the audio processing unit (APU) which deals with linear pulse code modulation (PCM) audio signals in the multipoint control unit (MCU), that used for a multipoint video conferencing system. The MCU receives audiovisual data from the subscriber, carries out video merging and audio mixing, and distributes the data to them. It also performs session setup, tear down, and control of the conference. The audio processing unit consists of a local processor and peripheral memories, a VME bus control block, a current speaker detection block, and an audio-mixing block. The current speaker detection block determines the channel of the current speaker by comparing the mean power of each channel. A mixing block mixes linear PCM audio data using speaker information and, performs echo cancellation. We propose another speaker detection algorithm using the scale factor of the MPEG-1 audio elementary stream. In simulation, it showed a similar result in speaker detection with that of linear PCM audio.