CSI-Based Wireless Localization and Activity Recognition Using Support Vector Machine

With the development of the information technology, human activity recognition and localization has received much attentions, since they can be utilized in many fields. In this paper, a new activity recognition and localization algorithm using channel state information (CSI) measurement is proposed. The problem of activity recognition and localization are formulated as the problem of machine learning which are solve with the support vector machine (SVM) approach. In the off-line phase, after the data normalization and principal component analysis (PCA) preprocessing of the CSI measurements, the (CSI measurement, label of activity) training data set and the (CSI measurement, position information) training data set for each activity can be formed. The SVM technique are utilized for activity based classification learning and position based regression learning. At last, the activity classification function and position regression function are obtained. In the on-line phase, after the data preprocessing of the received CSI measurements, the activity is estimated by the activity classification function at first. Then, the position regression function correspond to the estimated activity result is chosen for position estimation. Experimental results illustrated the activity recognition and localization performance of the proposed algorithm.