A Method for Detecting Convergence Completion of Adaptive Filter Cancelling Feedback Path Appeared in Active Noise Control Systems
* Presenting author
We present a method for detecting the convergence completion of the feedback control filter working as an adaptive filter. Generally, the feedback path is manually estimated before starting active noise control. This estimation, however, is desired to be automatically completed. We first introduce a step size control method capable of completing the estimation with a prescribed error even where the power of disturbance fluctuates. Then, we propose a method for detecting the convergence completion of the adaptive filter. In practical use, since the impulse response of the feedback path and its power gain are unknown, we cannot directly monitor the difference between the impulse response and the coefficients of the adaptive filter. For solving this problem, we apply two adaptive filters with two different step sizes to estimating the feedback path. Actually, we have already presented a recursive filter expression illustrating the structure of the normalized least mean square algorithm. According to the expression, the square error between the coefficients of the two adaptive filters converges on the prescribed error multiplied by the power of the difference between the two step sizes. We finally verify using computer simulations that the proposed method can successfully detect the convergence.