Evaluation of Speech Enhancement Algorithms for Hearing Aids

This study evaluates the performance of speech enhancement algorithms aimed at improving speech quality, particularly in the context of hearing aids. Four algorithms, namely Non-Negative Matrix Factorization (NMF), Wiener filtering, Logarithmic Minimum Mean Square Error (LogMMSE), and Minimum Mean Square Error Short-Time Spectral Amplitude (MMSE-STSA), are applied to noisy speech recordings to reduce noise and enhance speech clarity. Evaluation metrics such as Signal-to-Noise Ratio (SNR), Perceptual Evaluation of Speech Quality (PESQ), and Short-Time Objective Intelligibility (STOI) are used to assess algorithm effectiveness. Results indicate that the NMF algorithm achieves the highest SNR (4.33796), demonstrating strong noise reduction capability. However, Wiener and LogMMSE algorithms exhibit the highest PESQ scores (1.61774), suggesting superior perceived speech quality. MMSE-STSA algorithm performs best in intelligibility with an STOI of 0.739275. These findings offer valuable insights into algorithm performance, crucial for optimizing speech enhancement in hearing aids to enhance speech intelligibility and quality. The study underscores the importance of considering multiple metrics when selecting suitable algorithms for hearing aid technology, providing a comprehensive understanding of their effectiveness in real-world applications. Such insights contribute to the development of more effective speech enhancement solutions for hearing aid users.