Reduction of musical residual noise for speech enhancement using masking properties and optimal smoothing

Speech noise reduction is a very important research field with applications in many areas, such as voice communication and automatic speech recognition. Most speech enhancement systems produce annoying musical residual noise. This study aims to reduce the effect of musical residual noise. A perceptual-based speech enhancement algorithm that considers masking properties is first employed to enhance a speech signal corrupted by colored noise. Hence, an optimal smoothing factor, adapted by the variation of signal to spectral deviation ratio (SSDR) in successive frames, is used to reduce the effect of musical residual noise. Improved results are obtained by considering both the intra-frame masking properties and inter-frame spectrum smoothing in order to enhance a speech signal corrupted by colored noise.