A New Prediction Based Adaptive Median Filter for Restoration of Degraded Audio Signals

A Prediction based adaptive median filtering algorithm is proposed for restoration of audio signals. This algorithm consists of prediction, detection and adaptive median filtering stages. The proposed algorithm is efficient in detection and suppression of degradations in audio signals compared to the weighted median filter, recursive weighted median filter, adaptive median filters, model based approaches, wavelet approach and SD-ROM algorithm. These algorithms are also effective for the restoration of missing data samples. Normalized Least mean square algorithm is used for prediction technique. Large window size may lead to blurring of the signal in which case the window size is selected based on the number of corrupted samples present in the window. This avoids the unwanted filtering of original samples, which may lead to blurring. Computational results produce better SNR compared to other techniques.