An improved acoustic howling control technique with reduced computational load requirement

Of all howling control techniques, the adaptive feedback control (AFC) scheme promises high signal quality and high maximum stable gain (MSG) increase figures at the expense of high computational complexity requirements. The high computational complexity requirement of the AFC scheme generally limits its application in long acoustic feedback path applications because of the significant number of calculations per update cycle required. In this work, the feedback compensation technique having low computational complexity requirements is presented. A MATLAB® model of the proposed technique was developed and simulated for both speech and audio input signals. With a signal analysis frame size and acoustic path model coefficients of 128 and 4410, respectively, the technique requires 1025 multiplication operations; in comparison, the AFC scheme using the NLMS algorithm requires 17,646 calculations. The modified average signal quality figures of the developed technique are 2.3299 and 1.1859 for speech and audio signals respectively. The MSG increase figures are 8.428 dB and 7.76 dB for both speech and audio signals, respectively. These figures show that the developed technique has reduced computational complexity requirements at the expense of poor sound quality figures and less-than-infinite MSG increase figures.