A Device-free Human Fall Detection System Based on GMM-HMM Using WiFi Signals

The increase in human life span has created a demand for health care and remote monitoring technologies for the elderly, and falls are one of the major health care threats for those living alone. Traditional fall detection systems based on vision, sensor networks, or wearable devices have some inherent limitations, which makes it difficult to be popularized in engineering applications. In this paper, we propose a real-time, non-contact, low-cost but accurate indoor fall detection system using commercial WiFi equipment. The CSI phase difference expansion matrix is used as the fall detection feature and an effective approach is designed to intercept fall activity signals by using sliding window and labeling methods. Furthermore, the Gaussian Mixture Model-Hidden Markov Model (GMM-HMM) approach is innovatively migrated to a WiFi-based identification system which is originally used for human 3D skeleton-based activity recognition. The approach is of great value for its high accuracy compared with other classification algorithms, such as LSTM, Random forest. Based on the above approaches, our proposed system is implemented on two computers equipped with commercial 802.1 ln NIC, and the system performance is evaluated in three typical indoor scenarios. The experimental results show that the system has superior performance and can realize real-time fall detection for a single person.