Study On Acoustic Source Tracking Algorithm Using Distributed Microphone Arrays

Acoustic source tracking has been widely used in security surveillance, robot navigation,video conference system, and other field. The distributed microphone arrays are composed of a number of microphone arrays which are randomly placed in the space. But for traditional acoustic source tracking method, the microphone arrays must be regular topology. And it is also easy to get. In addition to the above characteristics, the positioning error under distributed microphone arrays is small, compared with the traditional microphone array. Now the research about acoustic source tracking using distributed microphone arrays is still in the beginning. Also there is no mature theory. Therefore, the research on acoustic source tracking using distributed microphone arrays has become a hot topic at home and abroad.This paper reviews the acoustic source tracking algorithm using microphone arrays,gives an introduction about characteristic of distributed microphone arrays and key technology involved in distributed microphone array signal processing, for example time synchronization and data fusion. And on the basis of above, the study on acoustic source tracking method in real acoustic environment is carried out.For the organization of distributed microphone arrays, this paper uses the theory of dynamic clustering to select appropriate number of microphone arrays to participate in acoustic source tracking. Dynamic clustering for acoustic source tracking reduces the computational complexity of the system. In view of the real trajectory of acoustic source in a closed environment, a hybrid movement model is proposed in this paper. By constructing the F matrix of the movement model, this model can change the turning rate parameters to achieve the random combination of the uniform model, the left turning model and the right turning model. In order to track an acoustic source whose motion model is a hybrid model, a distributed interacting multiple model particle filter algorithm is proposed. The algorithm is used to estimate the state and the probability of the model at the head of cluster. The state of the acoustic source at k time is obtained through the weighted sum of above two parameters.This paper uses interactive multiple model(IMM) algorithm to achieve the interaction of the three models. Last but not least, this paper combines generalized cross-correlation(GCC)method and distributed interacting multiple model particle filter to get an acoustic source tracking algorithm. The proposed method can realize single sound source tracking effectively under distributed microphone arrays. Finally, the simulation of the proposed algorithm under different noise and reverberation environment are performed. The results show that the proposed method can estimate the position of acoustic source and obtain a smoothed trajectory robustly in noisy and reverberant environments.