Application of Nine-Axis Accelerometer-Based Recognition of Daily Activities in Clinical Examination

Abstract

Background: This study aimed to establish an automatic and accurate method for identifying patient activity using wearable devices to facilitate simple measurement of the severity of disease, such as chronic obstructive pulmonary disease (COPD), and accurate diagnosis of arrhythmias using Holter electrocardiogram (ECG). Methods: Nine-axis accelerometers were attached to five different parts of the body of 30 healthy participants, and nine different activities were performed in sequence. Results: Overall, the dominant wrist, non-dominant wrist, and chest yielded high recognition accuracy, whereas the hip and thigh yielded lower recognition accuracy for some activities. Lying in the supine position, standing, walking, and running were identified with high accuracy by the accelerometer on the non-dominant wrist. Lying in the supine position, brushing teeth, walking, ascending/descending the stairs, and running were identified with high accuracy by the accelerometer on the chest. Conclusions: The movements related to the severity of COPD and those related to a diagnosis made via Holter ECG could be identified with reasonable accuracy when the nine-axis accelerometer was attached to one part of the body: the dominant wrist, non-dominant wrist, and chest. The accuracy was higher when the accelerometers were attached to five parts of the body.

Keywords

• wearable devices
• machine learning
• chronic obstructive pulmonary disease
• electrocardiography
• human activity recognition