Study Of Multi Target Location And Tracking Using The Sound Source Information

The sound source information for locating and tracking technology has been widely used in telephone and video conference, human-computer interaction, the speaker tracking and monitoring, intelligent robot, the sniper system, etc. However, due to the sound source information susceptible to environmental noise, how to accurately and quickly achieve the sound source locating and tracking is of great significance. In the real application scenario, the target number often is not a single, but multiple, and the number constantly will change over time, this makes the sound source locating and tracking algorithm become more complex. Under random set theory, probability hypothesis density(probability hypothesis density, PHD) filter for multiple target tracking algorithm has a strict mathematical theory, breaks through the traditional method of data association, reduces the amount of calculation, opens up a multiple target tracking area of research. This paper studies multi-detection probability hypothesis density(MD-PHD) multiple target tracking algorithm that uses multipath signal propagation to locate and track target, the following is the main research content of this article:First, study the characteristics of sound signal in a real environment and propagation model. Then establish a sound source signal propagation model. Through simulation, analyze the voice signal pulse response of the room. In order to improve the efficiency of subsequent processing, simulation of the sound source signal preprocessing was analyzed.Second, study the variety of generalized cross-correlation method for time delay. Through the simulation experiment, compare the advantages and disadvantages of various kinds of generalized cross correlation, the PHAT-GCC delay algorithm is studied mainly. Because the standard PHAT-GCC performance is affected easily by noise, research an improved PHAT-GCC algorithm. Improved PHAT-GCC has better performance by the simulation the simulation results.Third, on the concept of stochastic finite set, the recursive multiple target tracking model is established, then study and analyze the probability hypothesis density(PHD) filtering multiple target tracking algorithm and the particle filter. Through theoretical analysis and experimental simulation show that the particle filter probability hypothesis density filter for multiple target tracking algorithm(SMC-PHD) can achieve the target locating and tracking effectively.Fourth, in actual indoor application scenarios, receiving signal contains the direct signals and multipath signals, this paper studies Multi-detection probability hypothesis density multiple target tracking algorithm that exploits the multipath signals. The algorithm not only can achieve the target locating and tracking effectively, also reduce the volume of the microphone. Through the simulation experiments verify the effectiveness of the proposed algorithm in single-object scenario and multi-object scenario, and under different signal-to-noise ratio, the tracking performance is evaluated.