Realization of improvements to compressive sensing based speech enhancement for hearing aid applications

Enhancing the speech distorted with various noises and making the speech intelligible to humans is the prime requirement in hearing aid applications. In this work, a novel compressive sensing based speech enhancement technique with two improvements of low congruence binary sensing matrix and combined filter gain is proposed and realized using Simulink. The proposed solution has four stages of filtering, sparse representation, adaptive compressive sensing and optimized reconstruction. These four stages are realized in Simulink. The performance of the proposed solution is tested for different noisy speeches and performance measured in terms of various voice quality metrics and delays. The performance of the proposed solution is compared against other compressive sensing-based speech enhancement techniques