Active Noise Control Without Secondary Path Modeling:Algorithm And Implementation

Long time exposure to acoustic noise not only seriously damages the physical and mental health of human beings,but also interferes with normal life.Traditional noise control methods are mostly passive noise control,i.e.,using sound absorption or insulation devices to consume noise energy.However,the materials used in passive noise control are usually bulky,expensive,and the noise reduction effect of low-frequency noise is very disappointing.On the other hand,active noise control has flexible application scenarios,and features low cost,and good noise reduction effect on low-frequency noise.In recent years,it has attracted more and more attention and has been successfully applied in many scenarios.Most traditional active noise control algorithms are based on filtered-x least mean square algorithm or its modified versions,but all these algorithms require modeling the secondary path.Modeling can be done by off-line or on-line manners.The model obtained by off-line modeling is time-invariant.If the secondary path changes,the model will no longer match the actual path,so as to reduce or even lose the noise reduction effect.On-line modeling usually needs to introduce additional white noise at the secondary signal to excite the system,which affects the performance of noise reduction,and the modeling process and the update of filter coefficients will interfere with each other.In this thesis,an active noise control algorithm without secondary path modeling is proposed based on artificial bee colony algorithm,which avoids the problems caused by secondary path modeling.The time-varying nature of noise or measurement noise will lead to the fluctuation of the collected residual noise signal,which will cause the algorithm to fall into the false optimal filter coefficients.This thesis introduces a forgetting factor into the fitness function to make the algorithm have anti-interference ability.Because the artificial bee colony algorithm is good at exploration but poor at exploitation,in this thesis,the least mean square algorithm is integrated to improve it,which balances the relationship between exploration and exploitation,and significantly improves the convergence rate and noise reduction performance.In addition,if the amplitude of residual noise does not change much,the fitness value cannot accurately evaluate the quality of the filter coefficients.In this case,greedy selection mechanism of the artificial bee colony algorithm will result in a large number of high-quality filter coefficients being discarded and the algorithm will stagnate for a long time.To deal with this problem,this thesis introduces resampling and probability-based selection strategies to replace greedy selection,which effectively reduces the loss of high-quality filter coefficients and avoids the algorithm from falling into a long-term stagnation.Finally,the proposed algorithm is implemented in both single channel and dual channel active noise control systems,and experiments are conducted to verify its noise reduction performance.