A combined fixed/adaptive beamforming noise-reduction system for hearing aids

Poor speech intelligibility in noise is a major source of dissatisfaction for users of both cochlear implants and conventional hearing aids. Many noise reduction schemes have been proposed so far. The most promising approaches assume that the target signal is emitted in front of the user, while signals from other directions are considered to be noise. These directional or beamforming systems can be realized either with directional microphones or with microphone arrays and fired or adaptive postprocessing. In this work, a novel combined fixed/adaptive beamforming noise reduction system with four head mounted microphones is proposed. Two microphones are mounted on either side of the head in a behind-the-ear hearing aid housing. Each of these pairs of microphones forms a fixed beamformer (Audio-Zoom) and the resulting outputs are then post-processed by an adaptive beamforming scheme. The system has been implemented in real time on a portable digital signal processor system. It was evaluated in a moderately reverberant room, and speech recognition tests with two normal hearing listeners were performed, Preliminary results demonstrate an improved directional pattern and significantly enhanced speech recognition in noise, corresponding to a signal-to-noise advantage of approximately 17 dB over a single omnidirectional microphone.