Partitioned block frequency domain prediction error method based acoustic feedback cancellation for long feedback path

In this paper an innovative method of using Acoustic Feedback Cancellation (PEM based PBFD-AFC) in large acoustic spaces is presented. The system under analysis could vary from Single Source Single Receiver (SISR) to a Multiple Sources Multiple Receivers (MSMR). An environment is representative of (e.g.) churches installations or Public Address (PA) systems, thus involving the presence of one or more microphones and corresponding feedback paths. The Partitioned Block approach consists of slicing the feedback path (e.g the impulse response of the system) to improve the algorithm performance. It can be applied either in the time domain or in the frequency domain, where the latter, called Partitioned Block Frequency Domain, shows faster convergence, lower computational cost and higher estimation accuracy. The results of the proposed framework is compared with the state of the art using real acoustic data showing superior performance with up to 20dB Maximum Stable Gain (MSG) and 30 seconds less convergence time.