Kinect-based Source Direction Finding And Microphone Array Calibration

Compared with the single microphone system,the advantage of the microphone array is that it can obtain spatial and temporal information about the sound field.In speech signal processing,the spatial information of the sound field can be used for noise reduction,localization,speech enhancement,and other sound signal processing under the condition of reducing the distortion of the sound source signal.The localization of the sound source plays an important role in microphone array signal processing in many sound field applications.And it is a difficult problem for all kinds of sound source localization algorithms to find the balance point between the accuracy of sound source localization and the computation of the algorithm,which is suitable for common scenes or some specific scenes.The purpose of this thesis is to reduce the computational resources consumed by the algorithm,improve the real-time performance of the algorithm;And reduce the influence of array measurement error on the accuracy of the localization algorithm through the array calibration algorithm while realizing the localization accuracy.The specific work and results of this thesis are as follows:(1)This thesis summarizes the modeling factors that have different influences on the sound source localization algorithm.Starting with the signal receiving model and array topology structure,it introduces the different influences of different structures on the algorithm,which provides a model basis for the research and analysis of the algorithm used in the following chapters.(2)In this thesis,the error factors of time delay algorithm are analyzed.According to the theoretical error caused by the coupling effect of frequency and sample number,an improved delay-and-differential localization algorithm is proposed by change its cost function.Simulation results show that the proposed algorithm improves the performance of the traditional algorithm in precision.(3)In this thesis,a set of hardware and software combined platforms is designed through Microsoft Kinect,which can realize the acquisition and analysis of sound source data.The main system processes include experimental environment building,array signal acquisition(by microphone array of Kinect),speech signal pretreatment(against noise and the effects of truncation effect),array calibration(array element position calibration,to some extent reduced the error caused by mearsurement or other operations),the implementation of sound source localization algorithm.The system can implement a complete process,and the effectiveness of the proposed improved delay-and-difference algorithm is verified in the experimental environment.