Enhancement of reverberant speech using LP residual signal

In this paper, we propose a new method of processing speech degraded by rev erberation. The method is based on analysis of short (2 ms) segments of dat a to enhance the regions in the speech signal having high signal-to-reverbe rant component ratio (SRR). The short segment analysis shows that SRR is di fferent in different segments of speech. The processing method involves ide ntifying and manipulating the linear prediction residual signal in three di fferent regions of the speech signal, namely, high SRR region, low SRR regi on, and only reverberation component region. A weight function is derived t o modify the linear prediction residual signal. The weighted residual signa l samples are used to excite a time-varying all-pole filter to obtain perce ptually enhanced speech. The method is robust to noise present in the recor ded speech signal. The performance is illustrated through spectrograms, sub jective and objective evaluations.