ABSTRACT
Human activity recognition (HAR) is essential for understanding people’s habits and behaviors, providing an important data source for precise marketing and research in psychology and sociology. Different approaches have been proposed and applied to HAR. Data segmentation using a sliding window is a basic step during the HAR procedure, wherein the window length directly affects recognition performance. However, the window length is generally randomly selected without systematic study. In this study, we examined the impact of window length on smartphone sensor-based human motion and pose pattern recognition. With data collected from smartphone sensors, we tested a range of window lengths on five popular machine-learning methods: decision tree, support vector machine, K-nearest neighbor, Gaussian naïve Bayesian, and adaptive boosting. From the results, we provide recommendations for choosing the appropriate window length. Results corroborate that the influence of window length on the recognition of motion modes is significant but largely limited to pose pattern recognition. For motion mode recognition, a window length between 2.5â»3.5 s can provide an optimal tradeoff between recognition performance and speed. Adaptive boosting outperformed the other methods. For pose pattern recognition, 0.5 s was enough to obtain a satisfactory result. In addition, all of the tested methods performed well.
Subject(s)
Machine Learning , Movement , Pattern Recognition, Automated/methods , Posture , Smartphone/instrumentation , Bayes Theorem , Humans , Support Vector MachineABSTRACT
Satellite orbit and clock corrections are always treated as known quantities in GPS positioning models. Therefore, any error in the satellite orbit and clock products will probably cause significant consequences for GPS positioning, especially for real-time applications. Currently three types of satellite products have been made available for real-time positioning, including the broadcast ephemeris, the International GNSS Service (IGS) predicted ultra-rapid product, and the real-time product. In this study, these three predicted/real-time satellite orbit and clock products are first evaluated with respect to the post-mission IGS final product, which demonstrates cm to m level orbit accuracies and sub-ns to ns level clock accuracies. Impacts of real-time satellite orbit and clock products on GPS point and relative positioning are then investigated using the P3 and GAMIT software packages, respectively. Numerical results show that the real-time satellite clock corrections affect the point positioning more significantly than the orbit corrections. On the contrary, only the real-time orbit corrections impact the relative positioning. Compared with the positioning solution using the IGS final product with the nominal orbit accuracy of ~2.5 cm, the real-time broadcast ephemeris with ~2 m orbit accuracy provided <2 cm relative positioning error for baselines no longer than 216 km. As for the baselines ranging from 574 to 2982 km, the cm-dm level positioning error was identified for the relative positioning solution using the broadcast ephemeris. The real-time product could result in <5 mm relative positioning accuracy for baselines within 2982 km, slightly better than the predicted ultra-rapid product.