Received Signal Strength Based Research Of Sound Source Localization

Sound source location technology is to detect and calculate the sound signal received by acoustic electronic devices to ascertain the position of the natural or non-natural sound source. Its research is related to many subjects, including acoustics, signal processing, electronics, pattern recognition, software design, etc. Sound source location has a very broad application prospect, such as military detection, intelligent robots, mobile communications, video/audio conferencing, customs logistics, banking security system, infrastructure security, file management, health care, human-computer interaction, etc.There are mainly three sound source localization methods based on microphone array:Direction-of-Arrival method, Time-Difference-Of-Arrival (TDOA) method, Received Signal Strength Indication-Based (RSSI) method. This paper is focused on the RSSI method, which has small energy consumption, and is easy to hide from sight impact, and easy to implement even with line-of-sight obstructions.To solve the RSSI localization problem, the maximum likelihood (ML) estimation method is usually used to establish mathematical model, which has high localization accuracy. However, with a fractional nonlinear objective function, the ML model is difficult to solve. By simplifying the denominator of its objective function, ML model becomes a quartic polynomial form. Although simplified, currently known algorithms also need approximation method to solve this model. In this paper, two approximation algorithms are proposed:in Algorithm1, the ML objective function is further simplified into an unconstrained quadratic optimization model. Algorithm1has the advantage of much less computation complexity compared with currently known algorithms. Simulation results show that:under a minor and reasonable condition, the localization accuracy of algorithm1is even superior to currently known algorithms. In Algorithm2, we propose a noise subtraction form of least squares (LS) estimation method. By simplifying the denominator of objective function, this model directly turns into an unconstrained quadratic optimization problem. Simulation results show that:Algorithm2has substantially the same computation complexity and localization accuracy with Algorithm1. Both algorithms can satisfy the requirements of real-time localization.