Effects of Sampling Frequency on Human Activity Recognition with Machine Learning Aiming at Clinical Applications

Abstract

Human activity recognition using wearable accelerometer data can be a useful digital biomarker for severity assessment and the diagnosis of diseases, where the relationship between onset and patient activity is crucial. For long-term monitoring in clinical settings, the volume of data collected over time should be minimized to reduce power consumption, computational load, and communication volume. This study aimed to determine the lowest sampling frequency that maintains recognition accuracy for each activity. Thirty healthy participants wore nine-axis accelerometer sensors at five body locations and performed nine activities. Machine-learning-based activity recognition was conducted using data sampled at 100, 50, 25, 20, 10, and 1 Hz. Data from the non-dominant wrist and chest, which have previously shown high recognition accuracy, were used. Reducing the sampling frequency to 10 Hz did not significantly affect the recognition accuracy for either location. However, lowering the frequency to 1 Hz decreases the accuracy of many activities, particularly brushing teeth. Using data with a 10 Hz sampling frequency can maintain recognition accuracy while decreasing data volume, enabling long-term patient monitoring and device miniaturization for clinical applications.

Keywords: digital biomarkers; digital health; human activity recognition; machine learning; sampling frequency; wearable devices.

Figure 1

Experimental procedure and data analysis.

Figure 2

Waveform data from sensors at two body locations: the non-dominant wrist (red line) and the chest (green line). Labels (0–9) indicate activities: 0 = lying in the supine position, 1 = standing, 2 = sitting, 3 = eating, 4 = brushing teeth, 5 = using the restroom, 6 = walking, 7 = ascending/descending the stairs, 8 = running, 9 = other movements. (a) X-axis acceleration; (b) X-axis angular velocity; (c) X-axis magnetic field intensity.

Figure 3

F-value comparison across sampling frequencies for each activity. (a) Non-dominant wrist sensor; (b) Chest sensor.

Figure 4

Confusion matrices of predicted versus actual activities using data from the non-dominant wrist sensor. Rows show actual activities; columns show classifier predictions: (a) 10 Hz; (b) 1 Hz. Orange highlights indicate misclassifications where brushing teeth was mistaken for standing, sitting, eating, or using the restroom. Purple highlights mark errors where standing or sitting were misidentified as brushing teeth. Reducing the sampling frequency from 10 Hz to 1 Hz markedly increases these misclassification errors.

Figure 5

Confusion matrices of predicted versus actual activities using data from the chest sensor. Rows show actual activities; columns show classifier predictions: (a) 10 Hz; (b) 1 Hz. Red highlights indicate misclassifications where brushing teeth was mistaken for sitting, eating, or other movements. Green highlights mark errors where sitting, eating, or other movements were misidentified as brushing teeth. Reducing the sampling frequency from 10 Hz to 1 Hz markedly increases these misclassification errors.