Suppression of interference in speech signals via multirate filter banks

This paper has presented multirate filter banks based method for detection and suppression of narrow band interference in a speech signal. The Perfect Reconstruction (PR) and easy implementation property of the Cosine Modulated filter bank is considered in designing the Analysis and synthesis filter banks. The original speech signal reconstruction is done by detecting the interference band from the subband signals at the output of analysis bank and eliminating it in the synthesis filter bank resulting in a notch filtering. In this method, the Maximum Entropy principle is used for spectral estimation. The performance of the method is illustrated through implementation results on a contaminated speech signal.