Reduction of residual noise using directional median filter

Most of speech enhancement systems suffer from annoying musical residual noise which significantly reduces the perception quality of enhanced speech. In this study, we aim at reducing the musical residual noise by a two-stage speech enhancement approach. At the first stage, a noisy speech signal is enhanced by a three-step speech enhancement system which is constituted of two-step noise reduction algorithm and Virag method. Hence the enhanced speech signal is post-processed by a directional-median filter to significantly reduce the amount of residual noise, while the harmonic spectra can be well maintained at an acceptable level. Experimental results show that the proposed approach can significantly improve the performance of a speech enhancement system by reducing the quantity of residual noise.