Adaptive algorithm for line and speech enhancement
The area of adaptive signal processing provides one approach for removing the undesirable component from the signal of interest. In this thesis we present algorithms for line enhancement and speech enhancement. Line enhancement deals with removal of a narrowband interference from the broadband signal of interest, while speech enhancement removes the noise from the speech signal. Under low signal to noise ratio (0 dB or less) the performance of line enhancer suffers. A novel AR- based line enhancer in conjuction with a notch filter is proposed. Such a scheme provides fast detectability and removal of interference under low SNR at the input. A Kalman filter based speech enhancement method is investigated in this thesis. The Kalman filter allows for the nonstationarity of speech signal. This concept is extended to a multirate approach and a subband based Kalman filter for speech enhancement is introduced.The basic ideas of multirate digital signal processing are discussed.