Support Vector Machine-Based Human Detection using Channel State Information

• Main Authors: Ching-Chun Tseng, 曾靖鈞
• Other Authors: Po-Hsuan Tseng
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/s4s93j

碩士 === 國立臺北科技大學 === 電子工程系碩士班（碩士在職專班） === 105 === With the rapid development of wireless communication, the human detection implemented based on wireless communication can provide privacy. Channel state information (CSI) is essential for an OFDM-based system to measure the multi-path channel responses on each orthogonal sub-carrier. Compared to the received signal strength indication, more information from CSI can be utilized for human detection. However, there have not yet been effective statistical models of CSI under the situation in the presence of the human body or the absence of the human body. Therefore, in this study, a classification method based on the support vector machine (SVM) is proposed for human detection. In the offline stage, CSI measurements were collected and labeled based on the situation of the presence of the human body or not. The CSI measurements are first calculated based on seven different judgment rules. Then, the SVM builds the models according the calculated results of the associated judgment rules. In the online stage, the presence or the absence of the human body can be classified by comparing the online data with the training model. The research shows that the best scheme is the classification using the basis deviation, which can achieve a 99.38% detection rate while maintaining a 1.88% false alarm rate. The time deviation scheme also performs well to achieve a 96.25% detection rate, while maintaining a 2.41% false alarm rate.