A novel hybrid microphone array for hearing aids based on independent multiband beamforming

A microphone beamforming array for hearing aids can potentially provide substantial canceling of spatially distributed interference and competing speech. A new approach of independent multiband beamforming based on Discrete Wavelet Transform (DWT) is introduced, which presents a method of the Fixed Adaptive Beamforming (FAB) to achieve a balance between the effectiveness and robustness. A novel Hybrid Microphone Array (HMA) is used to give excellent performance for eyeglass-mounted hearing aids. A Log Area Ratio (LAR) based Variable Convergence Least-Mean-Square (VC-LMS) algorithm is integrated into the training process in order to preserve maximum speech intelligibility and best speech quality. A comprehensive beamforming simulation and analysis is performed on a self-built Speech Processing Platform (SPP) with a wide range of coverage on different speech database and acoustical environments. Both the Segmental Signal-to-Noise Ratio (SNR) and the objective analysis metrics of LAR distance prove that the FAB-HMA using LAR-based VC-LMS algorithm has significantly enhanced the corrupted speech.