A solution to mix low bit-rate speech signal in decentralized multipoint conference

The mixing algorithm for low bit-rate audio streams plays an important part in a decentralized multipoint conference system. In this article, from the fundamental principle of linear prediction codecs, a new mixing scheme for low bit-rate audio streams based on a comparison between the energy of the exciting series of the input bit streams is deduced to overcome the main disadvantages of the traditional mixing schemes. By its embodiment of the speech signals in line with ITU-T Recommendation G.723.1, it can be concluded that this solution is reliable, practical and of low complexity, satisfying the acoustical quality and the effective suppression of the accumulated background noise.