Spectral Subtraction with Non-Stationary Noise Estimation Utilizing Harmonic Structure

Abstract:- The spectral subtraction (SS) method is well known as a speech enhancement technique and has been widely used. In this paper we study the noise spectrum estimation required for the SS method. It is set out to estimate directly the noise spectrum from the noise-corrupted speech frame. To accomplish this, a fundamental frequency of speech is detected and harmonic structure is constructed and utilized. We assume that non-stationary noise consists of its stationary part and non-stationary part. The non-stationary part is estimated from the harmon-ics obtained, while the stationary part is estimated in the conventional way. Both parts are combined, resulting in an accurate estimate of the noise spectrum. It is shown by experiments that the proposed method provides a performance improvement relative to the conventional SS methods in non-stationary noise environments. Key–Words:- Speech enhancement, spectral subtraction, noise spectrum, non-stationary noise estimation, har-monic structure