An efficient noise reduction by using diagonal and nondiagonal estimation techniques

摘要

This paper describes about to reduce the musical noise by diagonal and non-diagonal estimation procedures in time-frequency domain to achieve better SNR of the musical noise signal.State of the art alogorithms parameterized filtering of spectrogaram coefficients with empirically fixed parameters. A block thresholding estimation procedure is introduced, which adjust all parameters adaptively to signal property by minimizing stein estimation of the risk. The resulting algorithm is robust to variations of signal structures such as short transients and long harmonics. Numerical experiments show that this new adaptive estimator is robust to signal type variations and improves the SNR and the perceived quality with respect to diagonal estimator. By this method noise will be more suppressed. For audio time-frequency denoising, we show that block thresholding regularises the estimate and is thus effective in musical noise reduction.