Analysis of inverse kepstrum and innovations-based application to noise cancellation

This paper presents analysis of inverse kepstrum (known elsewhere as complex cepstrum) and its application technique to two-microphone adaptive noise canceling structure. It is based on identification of one acoustic path transfer function from the ratio of two acoustic transfer functions and the estimator results are implemented in an innovations-based form. The method uses that the front-end inverse kepstrum estimates a denominator polynomial from the ratio of two acoustic path transfer functions between two microphones and then the cascaded adaptive FIR filter estimates the remaining numerator polynomial. It is applied to conventional adaptive noise canceling structure to achieve an improved performance in noise cancellation with an efficient processing technique. As a result, it will be shown that the front-end inverse kepstrum gives rise to computational simplicity, therefore could be applied to any noise cancelling method in a real-time processing.