Search

Activity recognition is fundamental to many applications envisaged in pervasive computing, especially in smart environments where the resident's data collected from sensors will be mapped to human activities. Previous research usually focuses on scripted or pre-segmented sequences related to activities, whereas many real-world deployments require information about the ongoing activities in real time. In this paper, we propose an online activity recognition model on streaming sensor data that incorporates the spatio-temporal correlation-based dynamic segmentation method and the stigmergy-based emergent modeling method to recognize activities when new sensor events are recorded. The dynamic segmentation approach integrating sensor correlation and time correlation judges whether two consecutive sensor events belong to the same window or not, avoiding events from very different functional areas or with a long time interval in the same window, thus obtaining the segmented window for every single event. Then, the emergent paradigm with marker-based stigmergy is adopted to build activity features that are explicitly represented as a directed weighted network to define the context for the last sensor event in this window, which does not need sophisticated domain knowledge. We validate the proposed method utilizing the real-world dataset Aruba from the CASAS project and the results show the effectiveness. [ABSTRACT FROM AUTHOR]

Radio tomographic imaging (RTI) is a technology for target localization by using radio frequency (RF) sensors in a wireless network. The change of the attenuation field caused by the target is represented by a shadowing image, which is then used to estimate the target's position. The shadowing image can be reconstructed from the variation of the received signal strength (RSS) in the wireless network. However, due to the interference from multi-path fading, not all the RSS variations are reliable. If the unreliable RSS variations are used for image reconstruction, some artifacts will appear in the shadowing image, which may cause the target's position being wrongly estimated. Due to the sparse property of the shadowing image, sparse Bayesian learning (SBL) can be employed for signal reconstruction. Aiming at enhancing the robustness to multipath fading, this paper explores the Laplace prior to characterize the shadowing image under the framework of SBL. Bayesian modeling, Bayesian inference and the fast algorithm are presented to achieve the maximum-a-posterior (MAP) solution. Finally, imaging, localization and tracking experiments from three different scenarios are conducted to validate the robustness to multipath fading. Meanwhile, the improved computational efficiency of using Laplace prior is validated in the localization-time experiment as well. [ABSTRACT FROM AUTHOR]

Device-Free Localization (DFL) based on the Radio Frequency (RF) is an emerging wireless sensing technology to perceive the position information of the target. To realize the real-time DFL with lower power, Back-projection Radio Tomographic Imaging (BRTI) has been used as a lightweight method to achieve the goal. However, the multipath noise in the RF sensing network may interfere with the measurement and the BRTI reconstruction performance. To resist the multipath interference in the observed data, it is necessary to recognize the informative RF link measurements that are truly affected by the target appearance. However, the existing methods based on the RF link state analysis are limited by the complex distribution of the RF link state and the high time complexity. In this paper, to enhance the performance of RF link state analysis, the RF link state analysis is transformed into a decomposition problem of the RF link state matrix, and an efficient RF link recognition method based on the low-rank and sparse decomposition is proposed to sense the spatiotemporal variation of the RF link state and accurately figure out the target-affected RF links. From the experimental results, the RF links recognized by the proposed method effectively reflect the target-induced RSS measurement variation with less time. Besides, the proposed method by recognizing the informative measurement is helpful to improve the accuracy of BRTI and enhance the efficiency in actual DFL applications. [ABSTRACT FROM AUTHOR]