Localization in Wireless Sensor Network Based on Multi-Class Support Vector Machines

Localization of sensor nodes is essential for wireless sensor network when it is applied to the special applications. We consider the localization of sensor nodes by integrating with multi-class support vector machines. During the offline training process, the received signal strength of the reference nodes is selected as the input of learning machines and the discrete value of location information is regarded as the class information of multi-class support vector machines. During the online localization process, the decision functions of multi-class support vector machines are used to estimate the location of blindfolded nodes. We demonstrate the practicality and feasibility of our method through simulations in the 100 m times 100 m area.