Smartphone-based single-channel speech enhancement application for hearing aids.

Area of Science:

• Signal Processing
• Acoustics
• Hearing Science

Background:

• Background noise significantly degrades speech intelligibility for hearing-impaired individuals.
• Existing single-channel speech enhancement (SE) methods often struggle with complex noisy conditions.

Purpose of the Study:

• To develop and evaluate a novel single-channel SE framework for hearing improvement.
• To implement the SE algorithm as a real-time smartphone application for practical use.

Main Methods:

• A super-Gaussian extension of the joint maximum a posteriori (SGJMAP) estimation rule was employed.
• A Mel filter-bank was used to smooth the SGJMAP estimation in the frequency domain, creating a Mel-warped estimation.
• A Mel-warped inverse discrete cosine transform (Mel-IDCT) was applied to obtain the impulse response for noise filtering and speech enhancement.

Main Results:

• Objective tests demonstrated superior speech quality and intelligibility compared to conventional SE methods.
• Real-time testing on an iPhone showed substantial improvements in speech recognition for both normal-hearing and hearing-impaired listeners.
• The algorithm effectively filtered background noise, enhancing speech signals in various simulated noisy conditions.

Conclusions:

• The proposed SGJMAP-based SE framework offers a significant advancement in hearing assistance technology.
• The open-source smartphone application provides a viable tool for real-time hearing improvement studies and applications.
• The Mel-warped frequency-domain processing effectively enhances speech perception in challenging acoustic environments.