부산대학교 도서관

This paper proposes a novel combinational Adaptive Noise Canceller (ANC), which is suitable for non stationary noise environments. Pre-filter bank structure improves the convergence rate leading to less error in real-time applications. During our investigations, we have found out adaptive algorithms have better performance for low frequency subbands. Thus, by using Singular Value Decomposition (SVD) in each subband, a simplified version of variable step-size algorithms has been employed in low frequency subbands; while for high frequency subbands, some conventional techniques have been used. Simulation results show that combination of adaptive and conventional methods have better performance with less computational complexity compared to other adaptive algorithms.