Multi-band noise canceller with reduced complexity

This paper presents a new LMS fast multi-band adaptive noise canceller (FMB-ANC) derived from a multi-band structure (MB-ANC) for speech applications with almost the same convergence speed of the original structure but presenting much lower computational costs. In turn, the output distortion due to crosstalk is dramatically decreased, improving the speech quality. The steady-state behavior of the MB-ANC is analyzed for a reference signal containing multiple sinusoids in white noise. It is shown that splitting the input signals in sub-bands increases the level of sinusoids cancellation