Fall detection and activity monitoring system using dynamic time warping for elderly and disabled people

We present a system that both tracks human movements and detects falling in elderly and disable people. We applied Dynamic Time Warping (DTW) to recognize human activities in daily living. Seven different movements, stand, sit, walk, run, stand-to-sit, sit-to-stand and lyings were considered and recorded as reference database signals. Our system consists of two parts: transmitter and receiver. A transmitter part is a device mounted at the user's waist with a size of a pager case measuring 90x40x20 mm. A sensor used in this device is a 3-axial accelerometer (Hitachi H48C). The signals from the accelerometer are transmitted wirelessly to a personal computer in receiver part using Zigbee Pro 2.4GHz. DTW is used to match the signals from different behaviors online with the databases and classify the data to a known activity. Falls are detected with a rule-based approach, in which the signal values are over thresholds following by the lying activity. Thresholds are computed from the minimum and maximum value in each axis of acceleration in the reference databases. The experiment shows 98.6 percent accuracy in recognizing these behaviors and in detecting fall.