Indoor Human Identification Using Advanced Machine-Learning-Based Strategy

Abstract

Major research efforts have been exerted to improve the accuracy of indoor person identification and facilitate the context-aware home services. These researches suffered from the low value of Correct Classification Rate (CCR), due to several technical reasons. In this paper, an advanced system combines pyroelectric infrared and floor-pressure sensors is proposed to identify persons in smart homes. Cooperative Multi-sensor strategy has been adopted to extract explicit information indicating the person's body size to improve the identification accuracy. A novel Machine-Learning-Based strategy is proposed to extract distinctive feature vector that represents the person's body size. Neural Network (NN) and Support Vector Machine (SVM) are used to improve the CCR of person identification. A prototype was designed and implemented. In addition, several test cases were conducted to examine and evaluate the effectiveness of the proposed system in identifying persons with high values of CCR.