WSNs中基于隐马尔科夫模型的目标识别问题研究

Target classification based on hidden Markov model in Wireless Sensor Networks

由于无线传感器网络(Wireless Sensor Networks,WSNs)资源受限,如何有效利用资源,提高目标辨别的准确度,是WSNs中目标识别系统的研究难题。以隐马尔科夫模型为分类框架,对一个声音传感器阵列节点簇内的目标识别问题进行建模;基于节点信号的空间关联性,改进了子节点Viterbi最大似然序列的计算状态,设置了子节点报送间隔,从而有效地判别局部状态。实验证明,改进后的算法在维持判别正确率的同时降低信息传输量10%以上。

It is challenging to classify multiple targets in wireless sensor networks based on the time-varying and continuous signals. In this paper, Hidden Markov Model is utilized as a framework for classification. The states in the HMM represent various combinations of vehicles of different types. With a sequence of observations, Viterbi algorithm is used at each sensor node to estimate the most likely sequence of states. Simulation results show that it reduce transmission more than 10%while maintaining identification rate.