Target Localization Based On Audio-visual Information

In recent decades,the audio localization has drawn extensive attention of scholars.It is widely used in the hands-free system,robot navigation,human-computer interaction,military and other fields.In addition,the use of video information to locate targets gradually become a reality with the rapid development of digital cameras and image processing technology.However, the performance of audio localization is easily affected by background noise and room reverberation.At the same time,the video localization will be affected by the target block,lighting,similar to the target background color and other factors.Therefore,the fusion of audio and video information to locate targets currently become a hot research.The main innovation of this paperare summarized as follows:1.An improved localization accuracy based on nested uniform circular microphone array configuration is proposed.The nested uniform circular microphone array configuration is proposed in this paper to eliminate the localization ambiguity and improve the localization accuracy,and deduced the choice range of circle radius at the same time.In the first localization,circle radius is not greater than the minimum half wavelength of speech signals.The second localization is dependent on the previous results,using the circle radius larger than previous for precise localization.Moreover,as the utilization of near-field acoustic source 3D localization algorithm with high resolution results in high computing complexity,wideband focusing algorithm is adopted in this paper,at the same time,the second localization reduced the search range, to make the delay time being reduced greatly.Simulation results show that this scheme has better localization effect.2.A statistical method for the target localization is proposed.This paper discusses the target localization problem in visual sensor networks.Additive noises and measurement errors will affect the accuracy of target localization when the visual nodes are equipped with low-resolution cameras.In the goal of improving the accuracy of target localization without prior knowledge of the target,each node extracts multiple feature points from images to represent the target at the sensor node level.A statistical method is presented to match the most correlated feature point pair for merging the position information of different sensor nodes at the base station.Besides,in the case that more than one target exists in the field of interest,a scheme for locating multiple targets is provided.Simulation results show that,our proposed method has desirable performance in improving the accuracy of locating single target or multiple targets.3.A global fusion method based on support matrix is improved in this paper. Firstly,we use partially Kalman filter to estimate the measurement data so that reduce the effects caused by the correlation and uncertainty of observation information, this is because the observing fusion algorithm is easy to be affected by the physical environment and characteristic of sensors. Then, each sensor information is weighted by weights which calculated by the support matrix.Finally,the weighted data is filtered by IMM so that it can be get performance of smoothing the filter and the correlation of time. In addition, our algorithm is simulation in three situation which include constant velocity, constant acceleration and maneuvering those three kind of movement,append high noise,occlusions by other targets.Compared with single-mode localization algorithm, fusion algorithm based on step by step and measurement fusion algorithm, simulation results show that the fusion localization performance of global fusion algorithm based on support matrix is better in any conditions, and he movement is more complex,the advantage of our algorithm is more significant.