Study on the extensibility of video conferencing with speech mixing

In a video-conferencing system, real time speech mixing is the key technology to make a real and natural conference in which the attendants of the system can exchange information more breezily and naturally. In this paper, three queue models for speech mixing are put forward and analyzed to optimize the program structure and codes for a video-conferencing system. Based on these models, the relation curves among conference size, pre-fetching packet number, and the sub-queue length are found, and the optimum sub-queue length and pre-fetching packet number in case of different conference sizes can be obtained from the curves for models. Tests and experiments show that the extensibility of conferencing systems with real-time speech mixing is enforced because the system can detect the online conference size when running, and automatically choose the optimum parameters from the look-up tables that contain values of the above parameters. When the actual number of attendants changes, the system can detect the changes, and then choose the optimum parameters to save the memory space and reduce the algorithm delay.