Speech recognition with a hearing-aid processing scheme combining beamforming with mask-informed speech enhancement

Area of Science:

• Audiology
• Signal Processing
• Hearing Aid Technology

Background:

• Hearing impairment often leads to difficulties in understanding speech in noisy environments.
• Advanced signal processing techniques are crucial for improving hearing aid performance.

Purpose of the Study:

• To evaluate a signal processing approach combining beamforming with mask-informed speech enhancement.
• To compare binaural and bilateral beamforming strategies for hearing aid users in simulated classroom noise.

Main Methods:

• Sentence recognition was measured in listeners with mild-to-moderate hearing impairment.
• Two beamforming types (binaural and bilateral) were compared, with and without mask-informed speech enhancement.
• Deep neural networks were used for mask-informed speech enhancement.

Main Results:

• Both beamformer types significantly improved speech recognition compared to no processing.
• Binaural beamforming provided greater benefit than bilateral beamforming.
• Mask-informed speech enhancement did not improve, and sometimes worsened, performance for both listener groups.

Conclusions:

• Binaural beamforming's signal-to-noise ratio (SNR) improvement outweighs spatial information loss for speech understanding.
• The implemented mask-informed speech enhancement did not enhance speech understanding beyond beamforming alone.