Research On Audio-Source Localization Technology In Wireless Sensor Networks

Localization technology is one of key technologies in wireless sensor networks(WSNs), and it is a precondition of target recognition, monitoring, tracking and many other applications. Localization of audio-source is extremely important in many military and civilian systems. It has attracted increasing attention from areas of academia and engineering.According to the special requirements of low power consumption and low computational complexity in WSNs, the audio-source localization methods based on GM-WLS (Gauss-Markov Weighted Least Square) and GCC-GA (Generalized Cross Correlation-Genetic Algorithm), respectively, were proposed in this paper. GM-WLS is presented to solve the problem of uncertainty of the anchor nodes'locations (GPS observations). It is based on weight matrix to get the accurate locations of wireless sensor nodes. After all wireless sensor nodes acquiring their locations, GCC-GA method can be utilized to localize the audio-source via a combination of distributed and centralized positioning approachs. The main idea is to determine the region where the sound source is located through distributed method, and then to accurately locate the target by the sensor node in this region. This method can improve the accuracy with low complexity.We design a Zigbee sensor network with nanoPAN5375RF nodes by Nanotron Company and MicaZ nodes by Crossbow. By simulation and experiment, we find that the observations of GPS anchor are not accurate. By introducing a weigh matrix to tackle the observation error, we show that the accuracy of the sensors'locations was improved based on the proposed GM-WLS method. The experiment also proved that it is better than Least Square Estimation (LSE) with lower localization error. It is demonstrated by the experiments that the accuracy of GCC-GA method is dependent on the distance of neighbor nodes, sampling period and the parameters of genetic algorithm. The methods and corresponding experimental results in this thesis can provide a theoretical guidance and implementation example for construction of real localization systems.