Noise Reduction Algorithm for Hearing Aids Based on Smoothed Wiener Filtering

To achieve high-performance noise reduction algorithms under the low computational power and high real-time requirements of hearing aid platforms, this paper proposes an improved noise estimation algorithm for non-stationary noise environments, aiming to enhance noise reduction performance in such environments. First, to improve the speed of noise tracking, we introduce spectral difference features to dynamically adjust the noise reduction coefficients, optimizing the traditional noise estimation algorithm based on minimum power estimation. Second, to ensure a balance between algorithm complexity and noise reduction effectiveness, we employ smoothed Wiener filtering technology to further suppress the estimated noise. Finally, the effectiveness of the proposed algorithm is verified through simulation experiments and actual hearing aid tests. The results show that, in non-stationary noise environments, the proposed algorithm outperforms the comparison algorithms in terms of audio quality and speech intelligibility, demonstrating that the proposed algorithm can effectively improve noise reduction in the more common non-stationary noise scenarios encountered by hearing aids, thereby enhancing the wearer's ability to understand speech in these environments.