Audio signal separation via a combination procedure of time-reversal and deconvolution process

摘要

Time-reversal method (TRM) is based on principle of reciprocity of sound by propagating of a reversal signal in time series to compensate distortion due to path effect in propagation and to focus the signal at the original source location. In recent years, the technique has been applied in optics, ultrasound and underwater acoustic communication. In the present study, a procedure for audio signal separation is developed by applying time-reversal and deconvolution process. The procedure separates sources from audio signal in time domain. The advantages of this procedure are to separate a specific source from a combination signal of multiple sources and to reduce reverberation effectively. Finally, a deconvolution process of finding impulse response function (IRF) is developed where single value decomposition (SVD) and Tikhonov regularization process are used to solve ill-conditioned and singular linear systems. In addition, the effect of signal-to-noise ratio (SNR) and number of array sensors are discussed in details.