INDOOR ACTIVITY RECOGNITION AND DEMENTIA RISK DETECTION USING Wi-Fi RECEIVED SIGNAL STRENGTH INDICATOR (RSSI) AND NAIVE BAYES CLASSIFICATION

Increasing life expectancy has resulted in a growing elderly population, making neurodegenerative conditions such as dementia a major global health issue. One of the main behavioral symptoms of dementia is wandering, which is characterized by repetitive and purposeless movement. Activity Recognition (AR) technologies, particularly those based on Wireless Sensor Networks (WSN), have gained attention for monitoring human behavior. Among these, Wi-Fi-based tracking using the Received Signal Strength Indicator (RSSI) offers a promising method for indoor activity monitoring and localization. This study aims to monitor the daily routines of elderly individuals, classify their current activity patterns by comparing them with previously recorded behaviors, and track their locations using Wi-Fi RSSI. A Naïve Bayes algorithm is proposed for activity classification and location tracking, while a time-based behavior graph is used to detect potential wandering behavior, aiding in early dementia risk assessment. The research utilizes primary data, which were collected directly through experiments in a controlled indoor environment. The data source comprises RSSI signals obtained from elderly participants. A purposive sampling method was employed to select participants aged 60 years and above, who were physically capable of performing the required tasks. A total of 4150 RSSI data samples were collected and analyzed. The proposed Naïve Bayes model achieved a classification accuracy of 64.60% using cross-validation, with a minimum average localization error of 0.7 meters, demonstrating the potential of this approach for early detection of dementia-related wandering behavior.