A Numerically Stable Moving Noise Reduction using Interior Point Least Square Algorithm

Abstract

In this paper, we propose a new method for a moving noise reduction with a numerical stability via 2-channel microphone array using the interior point least square(IPLS) algorithm. The noise reduction system is realized by the frequency domain generalized sidelobe canceller(FDGSC) and the IPLS algorithm is applied to form the null of array for the direction of moving noisy source adaptively. The IPLS algorithm is based on the interior point method for the constrained optimization method, and can guarantee the numerical stability at the learning and thus, we obtain the target sound extracted with a high S/N. A superior performance of the proposed method is presented by several computer simulation results when using the speech signal as the noisy source signal.