Validation of an Activity Type Recognition Model Classifying Daily Physical Behavior in Older Adults: The HAR70+ Model

Abstract

Activity monitoring combined with machine learning (ML) methods can contribute to detailed knowledge about daily physical behavior in older adults. The current study (1) evaluated the performance of an existing activity type recognition ML model (HARTH), based on data from healthy young adults, for classifying daily physical behavior in fit-to-frail older adults, (2) compared the performance with a ML model (HAR70+) that included training data from older adults, and (3) evaluated the ML models on older adults with and without walking aids. Eighteen older adults aged 70-95 years who ranged widely in physical function, including usage of walking aids, were equipped with a chest-mounted camera and two accelerometers during a semi-structured free-living protocol. Labeled accelerometer data from video analysis was used as ground truth for the classification of walking, standing, sitting, and lying identified by the ML models. Overall accuracy was high for both the HARTH model (91%) and the HAR70+ model (94%). The performance was lower for those using walking aids in both models, however, the overall accuracy improved from 87% to 93% in the HAR70+ model. The validated HAR70+ model contributes to more accurate classification of daily physical behavior in older adults that is essential for future research.

Keywords: accelerometer; daily physical behavior; free-living; human activity recognition; machine learning; older adults; physical activity; walking aids.

Figure 1

This figure illustrates the positions of the two accelerometers attached to the skin (highlighted with orange lines) and the chest-mounted camera used for the HAR70+ dataset. (a) The back accelerometer was positioned centrally at the third lumbar vertebrae. The z-axis of the coordinate system was pointing forward. (b) The thigh accelerometer was positioned approximately 10 cm above the upper border of the patella. The z-axis was pointing backward. The camera was attached to the chest on the outside of the clothing, pointing downwards.

Figure 2

The 10 features that provide the most information for activity type prediction in the HARTH model (left) and the HAR70+ model (right).

Figure 3

Time distribution of the labeled video data for the activity types during the semi-structured protocol. Light grey color represents the time of level walking with walking aids (WA).

Figure 4

Confusion matrixes for the Human Activity Recognition Trondheim (HARTH) model (a–c) and the Human Activity Recognition 70+ (HAR70+) model (d–f) for all participants (n = 18) and separated for participants not using walking aids (n = 13) (b,e), and participants using walking aids (n = 5) (c,f). The rows represent the labeled activity types, while the columns represent predicted activity types. Values are shown in row percentages.