Signal processing for hearing impairment

Area of Science:

• Audiology
• Signal Processing
• Biomedical Engineering

Background:

• Hearing impairment affects millions globally, necessitating advanced sensory aids.
• Effective noise reduction is crucial for improving speech intelligibility in noisy environments.
• Current hearing aid technologies face challenges in complex acoustic conditions.

Purpose of the Study:

• To evaluate the efficacy of four distinct noise reduction algorithms for hearing impairment.
• To compare the performance of adaptive noise cancellation, Wiener filtering, spectrum subtraction, and sinusoidal modeling.
• To determine the optimal method for enhancing speech recognition in sensory aids.

Main Methods:

• Evaluation of a two-microphone adaptive noise canceller.
• Assessment of short-term Wiener filtering.
• Analysis of a transformed spectrum subtraction technique.
• Investigation of sinusoidal modeling for noise reduction.

Main Results:

• The two-microphone adaptive noise canceller yielded the greatest speech recognition improvements in reverberant settings.
• Short-term Wiener filtering provided significant gains for certain hearing-impaired participants.
• Transformed spectrum subtraction enhanced performance as a cochlear implant preprocessor but not as a hearing aid front-end.
• Sinusoidal modeling improved signal-to-noise ratio but not speech intelligibility.

Conclusions:

• The two-microphone adaptive noise canceller is a promising technology for hearing aid applications.
• Algorithm selection for noise reduction should consider the specific hearing aid system and acoustic environment.
• Further research is needed to optimize algorithms like sinusoidal modeling for speech intelligibility.