Speech Enhancement Algorithm Based on Improved Wiener Filtering With Harmonic Model

Abstract

Speech enhancement aims to eliminate noise interference from noisy speech and extract the clean speech. It reduces speech distortion, improves speech quality, alleviates auditory fatigue, and enhances auditory perception. This paper presents a new speech enhancement algorithm based on improved wiener filtering algorithm with harmonic model. The harmonic model is used to represent the voiced part of speech signals. By determining different Fourier lengths according to the fundamental frequency synchronization characteristics, an accurate estimation of the phase spectrum can be achieved. Finally, fundamental frequency synchronized phase and amplitude spectra are inverse Fourier transformed to reconstruct noisy speech signals and enhance clean speech signals. To validate the effectiveness of the proposed algorithm, comparative tests were conducted. The results show that the new algorithm proposed in this paper has definite denoising advantages in both linear white noise and external environmental noise.