Simultaneous channel estimation and joint time-frequency domain crosstalk cancellation in multichannel personal audio systems

摘要

In this paper, we present two important contributions. The first is demonstration of the use of subliminal levels of pseudo-random noise to enhance channel estimation, and the second is a joint time and frequency domain algorithm for multichannel inversion. An adaptive system is presented where the acoustic channel is accurately estimated and utilized. In this implementation, maximum length sequences in the form of pseudo-random noise are superimposed on the source signals to aid the estimation quality of the acoustic channels. Upon estimation, crosstalk cancellation filters are designed using a time and frequency domain technique which uses a window to achieve more efficient and effective cancellation of crosstalk. For a 3×2 crosstalk system, the presented results show the improvement in channel estimation quality when low levels of maximum length sequences are superimposed on the source signals. On average −25dB of crosstalk cancellation is achieved.