Thursday 27 March 2025
A team of researchers has made a significant breakthrough in developing more accurate methods for tracking physical activity levels, which could have major implications for our understanding and management of conditions like osteoarthritis.
The study focused on using wearable sensors to track movement patterns and classify physical activities into different intensity levels. The researchers used a dataset collected from nine participants who performed various daily activities such as lying down, sitting, standing, walking, running, and cycling.
To test the accuracy of their methods, the team developed five neural network architectures that analyzed data from three different sensor configurations: wrist-only, wrist-and-chest, and wrist-an- kle. The results showed that incorporating sensors at multiple body locations significantly improved classification accuracy, particularly for high-intensity activities like running.
The wrist-only configuration was found to be effective in classifying low-intensity activities such as lying down or sitting, but struggled with medium- and high-intensity activities. This is because the wrist sensor alone may not capture the full range of motion patterns associated with these activities.
In contrast, adding a sensor at the ankle significantly improved performance, especially for high-intensity activities like running. This makes sense, given that the ankle plays a crucial role in movement kinematics and can provide valuable information about an individual’s physical activity level.
The study also found that the hybrid model combining convolutional neural networks (CNNs) with long short-term memory (LSTM) networks performed best overall, achieving high accuracy rates for all sensor configurations. This suggests that combining spatial feature extraction from CNN layers with temporal modeling of LSTM layers can be an effective way to analyze movement patterns and classify physical activities.
The implications of this research are significant, particularly for individuals with conditions like osteoarthritis who may benefit from personalized physical activity programs tailored to their specific needs and abilities. By using wearable sensors to track movement patterns, healthcare professionals could develop more targeted exercise regimens that help manage symptoms and improve overall quality of life.
Future studies will need to build on these findings by testing the methods in larger and more diverse populations. However, this research has already shown promising results and highlights the potential of machine learning algorithms in improving our understanding of physical activity patterns and their impact on health outcomes.
Cite this article: “Accurate Tracking of Physical Activity Levels Using Wearable Sensors”, The Science Archive, 2025.
Physical Activity Tracking, Wearable Sensors, Machine Learning, Neural Networks, Sensor Configurations, Movement Patterns, Classification Accuracy, Osteoarthritis Management, Exercise Regimens, Health Outcomes
