Microphone Array Speech Enhancement And Direction Finding Algorithms

At present,with the continuous development of microphone array technology,it is gradually applied in fields such as speech recognition,safe driving of cars,and abnormal fault location in substations.When the microphone array is collecting sound signals,it is susceptible to the influence of noise in the surrounding environment.The collected sound signals are multi-channel mixed noisy sound signals,which not only makes it difficult to effectively extract pure sound signals during analysis,but also affects the direction finding of abnormal sound sources.Under this background,a speech enhancement technology based on Blind signal separation is proposed to extract useful or specific sound sources in noisy environments to obtain pure sound source signals;At the same time,the L-shaped acoustic array direction finding technology is used to accurately determine the position of the sound source.The main work of this article is as follows:Firstly,to address the issue of microphone arrays being susceptible to background noise when collecting multi-channel speech signals in sound source localization,a method based on fast independent component analysis and minimum mean square error estimation is proposed to achieve multi-channel speech signal enhancement.The simulation and experimental results show that the algorithm has good speech enhancement ability and stability under different background noises.Compared with Fast ICA,Fast ICA spectral subtraction,and EMD-ICA algorithms in the indoor environment of substation secondary equipment,it can effectively improve the correlation coefficient and signal-to-noise ratio,and achieve speech enhancement of multi-channel mixed speech signals with background noise in the environment of substation secondary equipment room.Secondly,in order to achieve direction finding of abnormal sound sources in indoor environments,this article conducts research based on the indoor environment of secondary equipment in substations.Due to the complexity and variability of abnormal sound sources in the secondary equipment room of the substation,and the limitations of using contact type vibration sensors to collect vibration signals on equipment with narrow space and difficult to reach.Therefore,a non-contact acoustic array is adopted to solve this problem.We improved the multiple signal classification algorithm through the minimum length description criterion,designed and manufactured an L-shaped acoustic array sound source acquisition device,and conducted sound source acquisition.The results show that the improved multiple signal classification algorithm can effectively estimate the number of sound sources,provide prior conditions for two-dimensional direction finding,and make the direction finding accuracy of the L-shaped acoustic array accurate.Moreover,it ensures that the elevation or azimuth angles of multiple sound sources are the same and will not affect each other.Finally,in order to solve the problem of the main lobe peak being easily affected by the side lobe peak in the two-dimensional MUSIC direction finding algorithm,which leads to significant errors in the direction finding results,an improved MEMP direction finding algorithm without peak search is proposed.This method provides an effective number of sound sources K for the MEMP algorithm through the improved MDL algorithm,and uses the MEMP algorithm to achieve direction finding,which improves direction finding accuracy compared to traditional MUSIC direction finding methods,Reduced the possibility of errors in direction finding results.Using the three L-shaped microphone array positioning method,the point with the smallest sum of the distances between three out of plane straight lines is solved through three L-shaped sound arrays as the spatial position of the sound source,achieving three-dimensional positioning of the sound source.