Partical filtering for orientation determining using inertial sensors IMU

Using a low cost Inertial Measurement Sensor (IMU) to estimate an accurate orientation of human body is a challenging work. Several approaches has been proposed to minimize the sensor error as the Particle filter (PF) which has gain a popularity with nonlinear and non-gaussian noise system models. In the case of orientation estimation and positioning the PF has been employed in the frame of hybridization with other measurement systems such as visual systems, radio-localisation, GPS etc. In this study, we propose to use only the data obtained from inertial sensor, with a constrained PF. The Iterative Mean Density Truncation algorithm (IMeDeT) has been employed with a quaternion algorithm based on acceleration and magnetic field. An improvement in term of accuracy has been clearly detected as well as the filter convergence, against the Extended Kalman Filer (EKF)and an optical measurement system.