Multimodal approach for stroke recognition using body-mounted sensors

This paper presents multimodal stroke recognition algorithms using a magnetometer-aided inertial measurement unit (mIMU). At first, we devise a magnetometer-aided (mag-aided) algorithm based on two kinematic constraints. In addition, we propose hybrid algorithms composed of the mag-aided algorithm and the conventional gyro-aided algorithm. The hybrid algorithm outperforms respective algorithm (gyro-aided/mag-aided). Experimental results demonstrate high accuracy in stroke recognition rates.