Research On Target Point Noise Estimation Algorithm Based On Microphone Array

With the continuous improvement of China’s industrialization process,products from the pursuit of function into the pursuit of performance and quality stage,some products need to be offline noise detection,even abnormal noise identification,to avoid unqualified products into the market.In order to realize the noise detection and abnormal noise identification of offline products by instrument detection in factories,it is necessary to separate the working noise of products from the environmental noise.Physical isolation and algorithm filtering are two ways to realize noise separation.However,the use of physical noise isolation has major defects.The main research contents are as follows:(1)The basic theory of sound signal and physical topology model of microphone array are introduced,and a reasonable microphone array model is selected according to the actual conditions.(2)The classical spectral subtraction method is studied systematically.The realization of the algorithm should be based on the stable or slowly changing sound field environment.But because the background noise of the factory is complex and varied,it cannot be estimated by spectral subtraction.In this paper,the classical spectral subtraction method is improved.The improved spectral subtraction method can effectively reduce the influence of the change of sound field environment on the noise estimation results.Finally,the experiment verifies that the algorithm can achieve the target point noise estimation under the complex and changeable noise environment.(3)The response estimation algorithm based on transfer function is systematically studied,but the algorithm can not achieve the target point noise estimation under the complex and changeable noise environment.In this paper,the relative transfer function is combined with the measured response signal of the microphone array to achieve the target noise estimation,and a reasonable microphone array model is designed and selected.Finally,the experiment verifies that the algorithm can achieve the target noise estimation under the complex and changeable noise environment.(4)The microphone array verification platform was built,two noise estimation algorithms were written using C#,and the target noise response estimation was realized based on the optimal microphone array model,and the pros and cons of the estimation results of the two algorithms were compared horizontallThe experimental results show that both the improved spectral subtraction method and the relative transfer function can achieve noise estimation of target points,and the two algorithms have their own advantages and disadvantages: The microphone array space occupied by the improved spectral subtraction method is small,which is in line with the actual factory conditions,but the noise estimation accuracy is slightly lower than that of the relative transfer function algorithm,which requires the microphone array to have a large enough space.