Parallel feedback active noise control using decomposition of noise signal

Source noise is not measured directly but rather estimated in a so called "feedback" active noise control (FANC) method where no reference noise microphone or sensor is used. Unlike random noise, periodic signals can be attenuated well by a FANC method since they can be accurately estimated knowing the signal period. This paper presents a FANC method in which the estimate of noise signal measured by error microphone is decomposed into a periodic component and a random noise component using a low order adaptive linear prediction (LP) filtering and the signal auto-correlation samples. Each component is then attenuated by two separate NLMS-based FANC systems operating in parallel to cancel the original noise signal. The proposed FANC method is tested for different noisy tonal signals under various signal to noise ratios (SNR). The simulation results show the effectiveness of the proposed method.