我們研究了無線感測器網路(Wireless Sensor Network, WSN) 在混合直視(Line Of Sight, LOS) 與非直視(Non-Line Of Sight, NLOS) 訊號環境中的目標物定位，由於非直視訊號的影響，會使得原本的定位方法準確度下降，加上考量到感測器資源有限，是以接收訊號強度(Received Signal Strength, RSS) 等數據對其做量化(quantization)，融合中心以此量化訊號進行目標物定位。 由於非直視訊號的效應，我們透過建立雙模式高斯混合分佈的測量誤 差，並且假設其混合模型參數是完全未知，因此採用最大期望(Expectation Maximization, EM) 方法來近似目標物位置和混合模型參數的最大似然估計(Maximum Likelihood Estimation, MLE)，而我們所提出的方法修正了最小平方法(Least Squares Estimation, LSE) 在測量誤差屬於高斯混合分佈的狀況。;We studied the target localization method for quantized wireless sensor networks(WSN) in the mixed Line Of Sight (LOS) and Non-Line Of Sight (NLOS) signal environments. Owing to the influence of NLOS signals, the accuracy of conventional localization methods are degraded. And, considering limited power resources of sensors, the Received Signal Strength (RSS) data is usually quantized to several bits. The fusion center can only employ the quantized signals to localize the target position. Due to the effect of non-line of sight signals, we model the measurement noise as a Gaussian mixture distribution and assume that the mixture model parameters are completely unknown. Therefore, the Expectation Maximization (EM) method is used to approximate the Maximum Likelihood Estimation (MLE) target position and the mixed model parameters. The proposed method modifies the Least Squares Estimation (LSE) condition in which the measurement error belongs to a Gaussian mixture distribution.
