Speech perception in noise with a two-sensor frequency-domain minimum-variance (FMV) beamforming algorithm

The performance of a two-sensor based, frequency-domain minimum-variance beamforming algorithm (FMV) to extract a signal in the presence of multiple interferers was evaluated. Speech reception thresholds (SRT) and speech intelligibility measures were obtained from listeners with normal hearing or with mild-to-moderate sensorineural hearing loss. Word and sentence length stimuli were processed through the FMV algorithm, directional microphones alone, or with a simple delayand-sum beamformer. Listener's ratings of speech intelligibility, percent of words repeated correctly, and threshold for words and sentences in the presence of four competing signals showed the FMV to provide significant performance benefits across different listening environments.