From Wrist to Ankle: Understanding IMU Sensor Placement in Human Activity Recognition

Human Activity Recognition (HAR) using Inertial Measurement Units (IMUs) is a key technology in various domains, such as fitness monitoring, rehabilitation, and healthcare. The success of HAR systems largely depends on the strategic placement of IMU sensors on the human body, as different sensor locations capture varying motion dynamics. This paper systematically investigates the impact of sensor placement on the performance of HAR systems using a range of machine learning algorithms, including Logistic Regression, Naive Bayes, Random Forest, Decision Tree, Gradient Boosting, k-nearest neighbors (KNN), and Support Vector Machines (SVM), as well as deep learning models such as Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU). Sensor positions at the wrist, waist, and ankle were evaluated for their effectiveness in recognizing diverse physical activities, including walking, standing, lying down, running, jumping, sit-ups, and dancing. The study reveals that sensor location is critical in determining classification accuracy, with certain positions being more suited for specific activities. Moreover, combining multiple sensors across different body locations enhances system robustness and overall performance. The results provide key insights into optimizing sensor placement and model selection, paving the way for more accurate and application-specific HAR systems in wearable technology.