In the previous study [Wu, Too and Lee, Mechanical Systems and Signal Processing, 24 (2010) 1431-1443], audio signal separation has been achieved through the use of time reversal method and deconvolution process. The audio signal after this separation process is recovered from the mixed signal to a signal which is about 60% in correlation with respect to the original signal. In the present study, an adaptive time reversal algorithm has been introduced in this signal separation process. In the adaptive time reversal algorithm, weighting vector is determined by introducing objective function and constrained equation in an optimization process. The simulation is performed in a condition in a reverberation room. Through this adaptive time reversal method, it is found that the audio signal can be extracted from the mixed signal to achieve about 90% in correlation with the original signal.