Method and system for processing a recorded audio signal are disclosed according to some aspects of the present invention. The near-silent segments of the recorded audio signal are selected and their Mel-frequency cepstrum coefficients are determined by computing an inverse discrete cosine transform of a short-time Mel frequency log spectrum of the recorded audio signal. The entropy coefficients of the Mel-frequency cepstrum coefficients are computed and scaled. Thereafter, a source of the recorded audio signal is identified using one or more of a classification algorithm and a clustering algorithm.