A Low-computation Active Noise Control Algorithm

In recent years,active noise control(Active Noise Control,ANC)is widely used in daily life,industrial production and other occasions.With the deepening of research,there are more and more problems on ANC.For example,in order to reduce the noise effect of the active water bed system,the existing adaptive noise reduction algorithm must have a large number of constraints.It is an important factor for the wide application of multi-channel system.With the increase of the number of channels,it is usually necessary to reduce the order of the filter to meet the real-time requirements.In practical use,the order selection of the filter also needs many attempts.If the selected order is too large,the amount of computation will be wasted;if the selected order is too small,the noise reduction performance will not be enough.This thesis focuses on the calculation of active noise control algorithm to design a low computation adaptive active noise control algorithm(1)The existing filtered-x least mean square(Filtered-x Least Mean Square,Fx LMS)algorithm is improved,and a double gradient Fx LMS(Double-gradient Fx LMS,DGD-Fx LMS)algorithm is proposed based on whether the actual output of the controller exceeds the constraint.The algorithm improves the noise amplification problem of water bed effect by directly constraining the secondary signal,and reduces the amount of calculation by switching the update direction of different weight vectors.In the active noise reduction earphone,it is verified that the algorithm has good noise reduction effect while improving the water bed effect.Although it switches in two gradient directions,the system will not be unstable in the convergence process.(2)The computational complexity of feedforward multi-channel system and the influence of filter order on the performance of adaptive algorithm are analyzed.Two self-tuning filtering algorithms,DSC-Fx LMS(Desired-signal Self-correcting Fx LMS,DSC-Fx LMS)and ESCFx LMS(Error-signal Self-correcting Fx LMS,ESC-Fx LMS),are studied.By connecting multiple low-order filters in series,the output of the former filter is taken as the input of the latter filter,which simplifies the selection of filter order and reduces the computational complexity.Experimental results on active noise reduction headphones show that the algorithm has lower computational complexity,larger effective noise reduction bandwidth and lower high frequency noise lift compared with Fx LMS algorithm.