Reference Signal Selection Algorithm Of Multi-channel Active Noise Cancellation System

Nowadays,with the gradual acceleration of industrialization and urbanization,the problem of noise pollution has become increasingly serious.Noise will seriously interfere with people's normal work and life,and exposure to noise too long will bring great harm to people's physical and mental health.In recent years,active noise cancellation technology has sprung up again.One successful example of its applications is active noise cancellation headphones.Now,active noise cancellation headphones have become a common consumer electronic product,which provides a new solution for people to stay away from noise.The development of science and technology will never stop.In the field of active noise cancellation,people have been pursuing better noise cancellation performance,larger mute area and more efficient noise cancellation algorithms.Generally,the multichannel feedforward active noise cancellation system can perform well with multiple noise sources.In the multi-channel feedforward active noise cancellation system,the quality of the reference signal has a direct effect on performance of the system.However,due to the limitations of equipment volume,cost and the compute resources of the hardware,the number of reference microphones is limited to obtain the characteristics of noise sources,and the bad reference signal will even worsen the performance of the system.Therefore,how to select the best reference signal in complex acoustic filed is a problem worth studying.In this thesis,a multi-channel feedforward active noise cancellation seat experimental platform is built to simulate the noise scene for the study of reference signal selection algorithm.Assuming that the noise sources characteristics change slowly,an acoustic filed recognition algorithm based on the distribution of sound pressure is proposed in this thesis.In the new noise scene,this thesis improves the acoustic filed recognition algorithm,and proposes an acoustic filed recognition algorithm based on autonomous learning.The simulation results show that the average accuracy of the acoustic filed recognition algorithm based on autonomous learning is 99.6%.In each acoustic filed,there must be a set of best reference microphones.According to this principle,this thesis proposes a reference signal selection algorithm based on acoustic filed recognition,and uses the digital twin system of the experimental platform to find the best reference microphones corresponding to each acoustic filed.Finally,the reference signal selection algorithm based on acoustic filed recognition and the ohter two existing reference signal selection algorithms are applied to the multichannel active noise cancellation seat experimental platform.The simulation results show that the average noise reduction of the system under the reference signal selection algorithm proposed in this thesis is 0.8 d B lower than the optimal noise reduction,and the average noise reduction under the other two algorithms is 9.7 d B and 5.2 d B lower than the optimal noise reduction respectively.It can be seen that the reference signal selection algorithm based on acoustic filed recognition proposed in this thesis can choose the best reference signal effectively,and it's performance is better than the performance of the two existing reference signal selection algorithms.