The Method of Adaptive Noise Cancellation Based on Frequency Domain Subband Decomposition

For the subband adaptive filtering has the better performance in convergence and computing efficiency, it has been widely used in many signal processing fields, but the aliasing in-band from decimated in subband impair the system performance greatly. In the paper, based on the theory of signal orthogonal decomposition, used self-contained sinusoid basis, a novel subband signal adaptive noise cancellation method DFTSD-LMS is presented, where the analysis filters implemented by DFT have discrete ideal performance. Because there has no aliasing in-band after decimation, so the performance of cancellation is increased. The simulation results indicated that the system had higher convergence speed and SNR gain comparing with full band adaptive noise cancellation system, especially when the reference signal is colored noise.