Speech Enhancement System based on Nonlinear Spectral Attenuation using a Noise Masking Threshold

F.J. Fraga da Silva and L. Krebsky da Silveira Abranches (Brazil)


speech enhancement, nonlinear spectralsubtraction, noise masking.


In this paper a speech enhancement algorithm based on the Ephraim and Malah noise suppressor is presented. The system uses a noise masking threshold in a frame-by-frame basis in order to perform some important modifications in the original algorithm. These modifications were done with the aim of increasing the amount of noise reduction and simultaneously provide a more efficient elimination of the musical noise phenomenon. Evaluation results have shown that an improvement on the overall speech quality was achieved for all types of nearly stationary noise considered in the experiment, in a very wide signaltonoise ratio range of the original noisy signal.

Important Links:

Go Back