Abstract
This paper describes a new adaptive filter algorithm based on independent component analysis (ICA) for enhancing a target sound and for suppressing other interference sounds that are known. The technique can provide barge-in capable robot audition systems by utilizing known sound source signals such as self speech. Unlike a conventional ICA-based method, we use the time-frequency domain convolution model to cope with reflections of the sound. Experimental results showed that our method outperformed the conventional ICA-based method and the well-known adaptive filter algorithm called Nomalized Least Mean Squares (LMS) .
