Research On Multiple Target Tracking Algorithms Based On Distributed Microphone Arrays

The distributed microphone arrays can obtain the target space information by measurement.Based on the tracking theory,the time domain characteristics of the target can be combined with the target space information by which the advantages of wide coverage and high precision localization can be achieved.The multi-target tracking is the core content of the multi-sensor information fusion theory,which has been widely applied to military and civil fields.Target tracking mainly involves two aspects.One is to find a stable nonlinear filter.Under the Bayes framework,a new nonlinear filter namely cubature Kalman filter is investigated with combination of the spherical radial transformation and the spherical cubature rule.The cubature Kalman filter(CKF)achieves better stability over the extended Kalman filter and the unscented Kalman filter.Also,the CKF achieves almost the same tracking effect as the particle filterer but at low computational complexity.Another is to find an adaptive target motion model.An improved "current" statistical model has been proposed in this dissertation.By defining a zoon factor,the balance can be achieved between the stability of tracking and the accuracy of strong maneuvering target tracking.The multiple hypothesis tracking algorithm is investigated in the multi-target tracking.A multi-dimensional assignment algorithm is proposed to overcome the computational complexity caused by the increasing number of assumptions.The data association problem in multiple sensor multi-target tracking is transformed into an assignment problem.The Lagrange relaxation approach is utilized to reduce the dimension of multi-dimensional problems and the correct association between data and the trajectories of multiple target can be achieved.Finally,an outdoor experiment of distance speech signals tracking is designed using a dual microphone arrays.The experimental results show that the adaptive multi-target tracking algorithm proposed in this paper can achieve adaptive and stable tracking of multiple targets with different maneuvers.