Self-adapting multiple microphone system

Abstract An intelligent sound-sensing system is proposed. The system consists of a multiple microphone, a multiple-input linear filter, and a learning component for adapting the filter. This system can pick up a target sound signal out of other disturbing ambient noises with a better signal-to-noise ratio. A point of this system is a new self-learning algorithm for the adaptive filter. For self-learning, an internal learning signal is generated by multiplying a received signal by a cue signal. Any signal correlating with the power level of the target signal may be used as the cue signal. Since such cue signals are easy to obtain, many applications are available. In this paper, two types of intelligent microphone systems are described. The first type of system can pick up a non-stationary sound signal out of stationary disturbance sound signals. The second type of system has a photosensor to obtain a cue signal. The system distinguishes a target sound source from other disturbing sound sources by the visual cue. This new learning algorithm is theoretically analyzed, and the efficiency of the algorithm is demonstrated by experiments.