Wide-range flexible pressure sensors are in difficulty in research while in demand in application. In this paper, a wide-range capacitive flexible pressure sensor is developed with the foaming agent ammonium bicarbonate (NH₄HCO₃). By controlling the concentration of NH₄HCO₃ doped in the polydimethylsiloxane (PDMS) and repeating the curing process, pressure-sensitive dielectrics with various porosity are fabricated to expand the detection range of the capacitive pressure sensor. The shape and the size of each dielectric is defined by the 3D printed mold. To improve the dielectric property of the dielectric, a 1% weight ratio of multi-walled carbon nanotubes (MWCNTs) are doped into PDMS liquid. Besides that, a 5% weight ratio of MWCNTs is dispersed into deionized water and then coated on the electrodes to improve the contact state between copper electrodes and the dielectric. The laminated dielectric layer and two electrodes are assembled and tested. In order to verify the effectiveness of this design, some reference devices are prepared, such as sensors based on the dielectric with uniform porosity and a sensor with common copper electrodes. According to the testing results of these sensors, it can be seen that the sensor based on the dielectric with various porosity has higher sensitivity and a wider pressure detection range, which can detect the pressure range from 0 kPa to 1200 kPa and is extended to 300 kPa compared with the dielectric with uniform porosity. Finally, the sensor is applied to the fingerprint, finger joint, and knee bending test. The results show that the sensor has the potential to be applied to human motion detection.

WiFi-based human pose estimation has garnered significant interest in deep learning research. However, due to the varying angles of signal transceivers and the differing sensitivities of signal subcarriers to movement, inaccuracies can arise in WiFi-based human pose estimation. For instance, when a person is within a WiFi field, local changes in one or more channels and directions of structure can be detected. This channel interaction generally involves mutual interference, modifying movement localization, and perception sensitivity. To achieve unambiguous localization and identification, we decompose the properties of the Channel State Information spatial structure and its behavior, demonstrating that dual-view observation—spatial direction and channel sensitivity—is sufficient. Furthermore, we propose a CSI-Channel Spatial Decomposition Strategy (CSDS). Specifically, we introduce the Spatial Orientation Attention Module (SOA), which employs angle-dependent weighting to mitigate the error induced by signal transceiver pairs with deviated angles relative to the human body. Subsequently, the Spatial Sensitivity Enhancement Module (SSE) addresses errors from low-sensitivity signal carriers for motion detection by employing channel decoupling. Applying these two modules enables the model to discern potentially valid human pose information more effectively in WiFi transmission signals. The experimental results on the Wi-Pose public dataset demonstrate the effectiveness of CSDS.

The rapid growth of mobile applications and the Internet of Things generates a huge amount of data that can hardly be processed by end devices. Recently, Integrated Cloud and Edge (ICE) networks have been proposed to support smart applications and services by leveraging both remote clouds and edge clusters. However, task assignment over ICE networks has been proved NP-hard, which makes it challenging in practice. In this work, we discover that a polynomial-time optimal solution does exist for the task assignment problem when the ICE network is organized with tree topology. Specifically, the studied problem for the ICE network with tree topology is transformed into an equivalent graph optimization problem. Theoretical analysis is then provided to find the polynomial-time optimal solution. Evaluation results demonstrate the correctness and efficiency of the discovered solution.