Human activity recognition based on Hidden Markov Models using computational RFID

RFID is widely adopted for human activity recognition in interior environments, e.g., elder-caring. Gaining insight through raw RFID data analysis is the key part of the human activity recognition systems. However, the inviolable uncertainty in RFID data, including external environment noise and fragmentary reading (reading collision), increase the difficulty for high-level application widely adoption. In order to address these challenges, we proposing a Hidden Markov Models based data analysis approach in this paper, comparing with previous researches, our method need less limitations and requires only a few prior knowledge about RFID placing, the approach learns from raw RFID data and apply it to analyze the data. Our method analyzes RFID RSSI and 3D-accelerometer data collecting from human movement recognition to overcome aforementioned issues. This system has already been built and successfully deployed in a real experimental room. Result shows that the system run well to obtains an activity recognition with low error rate of 2.5%.