Memory network has been extensively used to record prototypical normal patterns to prevent overgeneralization of the network to reconstruct anomalies for video anomaly detection. However, existing memory-based methods only record the lossy representation of normal item prototypes, without recording the rich relationships between them. In this work, we propose an Associative Memory with Spatio-Temporal Enhancement (AMSTE) which introduces the global context information constraint of motion to enhance the appearance features and learn the normal item prototypes and their relationship. Specifically, we utilize two encoders to extract spatio-temporal features with the Spatio-Temporal Enhancement Module (STEM) to enhance appearance features with global motion constraints. Then, the prototypical patterns of normal data and their relationships are recorded in the item memory and relational memory, respectively. Finally, we retrieve features from the memory pools and reconstruct the video frame through the decoder. Extensive experiments on three benchmark datasets demonstrate the effectiveness of our approach. The code will be released at https://github.com/HuYongting/AMSTE .

This paper examines a new technique to improve the figure of merit of laterally vibrating RF-MEMS resonators through an energy-preserving suspended addendum frame structure using finite element analysis. The proposed suspended addendum frame on the sides of the resonant plate helps as a mechanical vibration isolator from the supporting substrate. This enables the resonator to have a low acoustic energy loss, resulting in a higher quality factor. The simulated attenuation characteristics of the suspended addendum frame are up to an order of magnitude larger than those achieved with the conventional structure. Even though the deployed technique does not have a significant impact on increasing the effective electromechanical coupling coefficient, due to a gigantic improvement in the unloaded quality factor, from 4106 to 51,136, the resonator with the suspended frame achieved an 11-folds improvement in the figure of merit compared to that of the conventional resonator. Moreover, the insertion loss was improved from 5 dB down to a value as low as 0.7 dB. Furthermore, a method of suppressing spurious mode is demonstrated to remove the one incurred by the reflected waves due to the proposed energy-preserving structure.

In the study of image compressed sensing (CS), various priors are explored for regularization to achieve better reconstruction and provide different additional information. However, it is a challenge to combine different priors for better performance. In this article, we propose a novel framework (dubbed SCS-GNet) that integrates the merits of multiple priors for image CS. The framework is composed of a measurement matrix generator network and a reconstruction generator network, in which the implicit image prior is explored in the process of measurement matrix generation, and deep image and sparsity priors are explored in the reconstruction process. The two generator networks are separate in SCS-GNet, and the partial consistency of loss functions is explored for connection to boost the performance of image CS. Different from existing end-to-end joint training deep network-based methods, our measurement matrix generation process is performed separately, while the reconstruction process is training-free. Moreover, a novel training strategy is applied in the process of measurement matrix generation to enable the framework to be available for different measurement rates. Extensive experiments demonstrate that the proposed SCS-GNet achieves adaptive and scalable measurement and reconstruction at different measurement rates and offers competitive performance compared with other state-of-the-art methods.

A SAW device with a multi-layered piezoelectric substrate has excellent performance due to its high Q value. A multi-layer piezoelectric substrate combined with phononic crystal structures capable of acoustic wave reflection with a very small array can achieve miniaturization and high performance. In this paper, a honeycomb-shaped phononic crystal structure based on 42°Y-X LT/SiO₂/poly-Si/Si-layered substrate is proposed. The analysis of the bandgap distribution under various filling fractions was carried out using dispersion and transmission characteristics. In order to study the application of PnCs in SAW devices, one-port resonators with different reflectors were compared and analyzed. Based on the frequency response curves and Bode-Q value curves, it was found that when the HC-PnC structure is used as a reflector, it can not only improve the transmission loss of the resonator but also reduce the size of the device.

This paper studies the radial alternating material phononic crystal (RAM-PnC). By simulating the band gap structure of the phononic crystal, a complete acoustic band gap was verified at the resonant frequency of 175.14 MHz, which can prevent the propagation of elastic waves in a specific direction. The proposed alternately arranged radial phononic crystal structure is applied to the thin-film piezoelectric-on-silicon (TPOS) MEMS resonator. The finite element simulation method increases the anchor quality factor (Q_anchor) from 60,596 to 659,536,011 at the operating frequency of 175.14 MHz, which is about 10,000 times higher. The motion resistance of the RAM-PnC resonator is reduced from 156.25 Ω to 48.31 Ω compared with the traditional resonator. At the same time, the insertion loss of the RAM-PnC resonator is reduced by 1.1 dB compared with the traditional resonator.

At present, with the original point cloud as input, most of the object detectors use Pointnet++ to extract features of the point cloud based on the Farthest Point Sampling (FPS). However, affected by FPS, feature extraction is incomplete and unstable. Moreover, high-level semantic features lack the internal vertex properties of Regions of Interest (RoI). In order to solve the above problems, we propose the MaskSL RCNN (Mask-Spherical-neighborhood-global-feature-Layer Region-CNN), a feature-enhanced 3D object detection network. It improves sampling of the farthest point through point-level feature enhancement. In addition, we propose Spherical neighborhood global feature Layer (SL) to supplement the global features and improve the learning ability of network. At last, based on semantic-level feature enhancement, we design grid pooling layer based on vertex attention, which fully explores the boundary characteristics of RoI and increases ability to learn advanced features in RoI. Our network improves detection precision of small objects such as pedestrians. Compared with PointRCNN, it has improved the mAP of simple, medium, and difficult object detection in the KITTI dataset by 2.66%, 1.69%, and 0.67%, respectively.