Implementation of Blind Signal Dereverberation Of Speech Signals using Cuckoo-Independent Component Algorithm

Abstract

Blind signal processing of convolutive combinations of unknown time series is asignificant building block in modern schemes connecting broadband signal acquisition bysensor arrays in multipath. Reverberation, a section of any sound produced in a natural atmosphere, can reduce speech intelligibility or more usually the quality of a signal formed within a room.The process of recovering the source signal by removing the unwanted reverberation is called dereverberation.Speech dialog systems interact with the user by recognizing and interpreting the meaning of the received commands.The ICA goes to the category of blind source severance (BSS) and the ICA prominently determined by the key assumption of the physical world character. The BSS estimates the original signal using the mixed-signal information observed from the input channel. The proposed method avoids the drawback of separated sounds with improper localization, directivity and spatial quality of separate sources. The wavelet filter and multi-step linear prediction coding (mLPC)for extraction of coefficients in the late reverberation. The reverberated signals are eliminated using backward differentiation. Cuckoo is an optimization technique is used to improve the effectiveness of ICA. To reduce the redundant bit and hardware cost the new FPGA was proposed to improve the reliability. Hence the reliability and SNR values are increased. Various methods are compared with proposed Cuckoo search algorithm to demonstrate the efficiency of frequency, time delay and power consumption with reduced area utilization.