Noise Suppression in Speech Using Multi{resolution Sinusoidal Modeling

The multi{resolution sinusoidal transform (MRST) 1] provides a sparse representation for speech signals by utilizing several psychoacoustic phenomena. It is well suited to applications in signal enhancement because the signal is represented in a parametric manner that is easy to manipulate. The MRST has the additional advantage that it is both particularly well suited to typical speech signals and well matched to the human auditory system. The currently reported work removes background noise from a noisy signal by applying an adaptive Wiener lter to the MRST parameters and then conditioning the parameters to eliminate \musical noise." In informal tests MRST based noise reduction was found to reduce background noise better than traditional Wiener ltering and to virtually eliminate the \musical noise" often associated with Wiener ltering.