No abstract

Microphone army technology has been proposed for various audio, teleconference, hearing aid, and voice recognition applications. B y forming a focused beam toward the desired speech soullce, attenuating background noises and rejecting discrete spatial interferers, a microphone a m y can enhance the signal-to-noise-mtio (SNR) in a noisy environment with notable improvement in speech intelligibility. At the threshold of intelligibility, an one dB improvement in SNR can increase 10-15% speech intelligibility. It is also known increasing SNR can result in significant improvement in the recognition mte of v,urious automatic voice recognition systems. We propose an novel electronically steemble microphone army based on the Mazimum Energy (ME) concentmtion criterion, which results in high SNR in moms with reverberations and competing interferences. We present a prototype PC-based microphone a m y system designed for hearing aid applications but also ap'plicable to other tasks.In section 1, a multiple channel microphone army system is described. The autonomous system's basic opemtions of signal-enhancement, beamformation, and detection-and-tmcking are defined as processing mode, calibmtion mode, and search mode respectively which are presented in section 2,9 and 4. Algorithms using ME beamformation army weights obtained from the solution of a genemlized eigenvalue problem utilizing non-ideal microphone characteristics and modified bnmdband near-field MUSIC schemes am discussed. Some army performance results based on both computer simulation and real-time measurement are presented.