An Online Prediction and Trajectory Tracking Method for Human Activity Recognition

Human activity recognition(HAR) is a focus of current research. In this study, through feature selection algorithms, we researched the online posture classification and prediction. With the best selected features and classifiers from each layer, we proposed a method of online prediction and classification, and the prediction accuracy achieved 94%. Support vector machine (SVM) was used to classify seven types of posture with an accuracy of 89.3%. For the trajectory tracking, the acceleration and angular velocity were filtered for denoising, and the gyroscope were calibrated with zero point to reduce the error in trajectory tracking. On the basis of judging posture for walking, the acceleration information in the navigation coordinate system was calculated by the posture solution of quaternion. Then, the motion trajectory was predicted by inertial navigation information. After the experiment, the proposed method could recover the walking trajectory well.