基于混合高斯模型测距误差修正和EM-SOM的节点定位算法设计

Design for Sensor Node Localization Based on Gaussian Mixture Model Based on Distance Error Correction and EM-SOM

摘要:

针对RSSI信号强度定位方法中当发射节点和接收节点之间的无线电传播路径被障碍物阻挡而造成噪声误差,使得节点定位欠精确的问题,设计了一种基于混合高斯模型进行测距误差修正和EM-SOM的节点定位算法；首先,通过混合高斯模型对RSSI获得的测量距离误差进行建模,通过EM方法对混合高斯模型中的各参数即各高斯模型的权值、均值和协方差进行训练,采用自组织的SOM对测量距离样本进行聚类,获得各高斯模型的初始权值,最后,将测量距离输入训练后的高斯模型获得较为真实的距离值,并通过极大似然估计进行节点的定位；实验结果表明文中方法能对具有噪声误差的节点进行定位,且相对其它方法,具有平均定位误差和均方根误差小的优点,具有较大的优越性。

Abstract:

Aiming at the RSSI localization method has the detects of inaccuracy of localization, owing to the noise error of barrier interrupt between transmitting sensor node and receiving sensor node, a node localization algorithm based on Gaussian mixture model correcting distance error and EM-SOM is proposed. Firstly, the Gaussian mixture model is used to model the error of the RSSI distance, EM method is used to train the weight, mean value and covariance of all the Gaussian function. The self-organizing SOM is used to cluster for the distance sample, so the initial weights of all the Gaussian model are obtained. Finally, the measured distance is input the Gaussian mixture model to get the final distance value, and then the maximum likelihood estimation is introduce to get the position of all the nodes. The experiment result shows it can realize the sensor node localization, and compared with the other methods, it has the advantage of less average localization error and mean root error, therefore, it has big priority over the other methods. 