Long-term Detection System for Six Kinds of Abnormal Behavior of the Elderly Living Alone

TL;DR

Numerical evaluations using sensor data that includes a variety of resident behavior patterns over nine years as test data show that the methods for detecting wandering and falls are comparable to previous methods, and the methods for detecting being semi-bedridden, being housebound, and forgetting achieve a sensitivity of over 0.9 with fewer than one false alarm every 50 days.

Abstract

The proportion of elderly people is increasing worldwide, particularly those living alone in Japan. As elderly people get older, their risks of physical disabilities and health issues increase. To automatically discover these issues at a low cost in daily life, sensor-based detection in a smart home is promising. As part of the effort towards early detection of abnormal behaviors, we propose a simulator-based detection systems for six typical anomalies: being semi-bedridden, being housebound, forgetting, wandering, fall while walking and fall while standing. Our detection system can be customized for various room layout, sensor arrangement and resident's characteristics by training detection classifiers using the simulator with the parameters fitted to individual cases. Considering that the six anomalies that our system detects have various occurrence durations, such as being housebound for weeks or lying still for seconds after a fall, the detection classifiers of our system produce anomaly labels depending on each anomaly's occurrence duration, e.g., housebound per day and falls per second. We propose a method that standardizes the processing of sensor data, and uses a simple detection approach. Although the validity depends on the realism of the simulation, numerical evaluations using sensor data that includes a variety of resident behavior patterns over nine years as test data show that (1) the methods for detecting wandering and falls are comparable to previous methods, and (2) the methods for detecting being semi-bedridden, being housebound, and forgetting achieve a sensitivity of over 0.9 with fewer than one false alarm every 50 days.

Figures (3)

• Figure 1: Overview of systems. Anomaly detection classifiers are learned by the simulation data generated from simulators customized based on the information of operational environment. The simulated time series data are transformed into anomaly-dependent features and label interval data to apply different types and durations of anomalies.
• Figure 2: Example of a floor plan and sensor arrangement. The layout is 5 meters by 12 meters. There are chairs (C), a cupboard (CB), a dining table (DT), a kitchen stove (KS), a refrigerator (RF), a trash box (T), a wardrobe (WR), a washing machine (WM) and a water closet (WC). Colors and symbols differentiate sensor types: red circles for passive infrared sensors (#0 to #33), gray squares for pressure sensors (#34 and #35), blue stars for flow sensors (#36 and #37), yellow stars for power sensors (#38 and #39), and green triangles for the door sensor (#40)
• Figure 3: Illustration of our training data construction using a simulator. Initially, the smart home simulator generates a time series data of $S$ sensor firings and anomaly labels. The sensor data is summarized into a one-second interval binary-value vector sequence (data matrix). Each anomaly label sequence is summarized into a binary-value sequence with anomaly-specific intervals. To learn a classifier for each of the six anomalies, the sequence of feature vectors appropriate for each anomaly is extracted from the sensor data matrix.