Proposing a System for Unobtrusive Detection of Body Movements during Sleep Based on the Measurement of Wi-Fi RSSI

Although sleeping is essential for healthy life, many people have a problem related to insufficient or poor quality sleep. This problem causes not only loss of concentration but also a decline in immune function and various diseases. Also, children's lack of sleep is known for preventing mental and physical growth. Therefore, many researchers have developed sleep monitoring systems that permit us to measure the quality of sleep. However, because many existing sleep monitoring systems require a user to wear some some device, e.g., wristband, the systems are more or less obtrusive to the users. In addition, several existing systems require an expensive special device and expert knowledge. On the other hand, previous research has found a correlation between body movements during sleep such as rolling over and sleep state. In this paper, we propose a method for unobtrusive detection of body movements during sleep based on the measurement of Wi-Fi Received Signal Strength Indication (RSSI) by using off-the-shelf Wi-Fi devices like smart phones or tablet computers. From the detected body movements, we can estimate a user's sleep state such as Rapid eye movement (REM) sleep or non-REM sleep. In our proposed system, we obtain RSSI data by placing two devices equipped with Wi-Fi modules on the left and right sides of a user's bed when sleeping. One devices measures RSSI of the opposite device and we use the RSSI data to construct a body movement recognition model by employing machine learning approach. We construct a user independent recognition model based on the hidden Markov model, and employ maximum likelihood linear regression (MLLR) to adapt the user independent model to an end user. Therefore, our system is designed so that it does not require an end user's labeled training data.