Analysis of feedback cancellation in hearing aids with Filtered-x LMS and the direct method of closed loop identification

Feedback cancellation in hearing aids based on Filtered-X LMS is analyzed. The data used for identification of the feedback path are the output and input signals of the hearing aid. The identification is thus done in a closed loop. Tracking characteristics and bias of the optimal estimate are discussed. The optimal estimate can be biased when the identification is performed in closed loop and the input signal to the hearing aid is not white. It is shown that the bias could be avoided if the spectrum of the input signal was known and the data used to update the internal feedback is prefiltered. The effects of different choices of the design variables of the Filtered-X LMS are discussed. Three alternatives of the fixed filter are evaluated on feedback paths of hearing aids on human subjects and with alternative spectra of the input signal. The optimal estimate of the Filtered-X LMS with an individually adjusted fixed filter shows the best agreement with the desired estimate