Human Motion Recognition for Adaptive Deep Brain Stimulation Using a Head-mounted Triaxial Accelerometer

Adaptive deep brain stimulation (aDBS) represents a promising advancement in the treatment of Parkinson's disease (PD), offering personalized therapy through real-time adaptability. This study investigates the feasibility of utilizing a head-mounted triaxial accelerometer to detect human motions for a low-power, fully body-embedded aDBS system. We recruited 32 healthy participants to perform daily activities including sitting, standing, walking, and falling, while wearing the accelerometer on their heads. To classify these motions, we employed machine learning algorithms such as Decision Tree, Random Forest, and Support Vector Machine (SVM). Our results indicated that the Radial Basis Function (RBF) kernel SVM achieved the highest classification accuracy, exceeding 80%. However, differentiating between sitting and standing posed a challenge. Notably, the inclusion of a "pre-subsequent motion" feature significantly enhanced performance, resulting in an accuracy of over 96%. This study establishes a robust framework for feature analysis and machine learning algorithm selection in the realm of human motion recognition through a head-mounted accelerometer.Clinical Relevance- This study outlines a method for detecting human motions through head acceleration, which can be deployed in an adaptive deep brain stimulation device.