In this paper, we propose a lightweight facial expression recognition algorithm based on joint model compression. Firstly, self-distillation method recognition, which not only considers the overall output of but also focuses prediction results each category, thereby improving performance and generalization ability model. Secondly, in order to further compress model, apply channel pruning transformed <i>L</i>₁ regularization. This imposes non-convex constraints parameters through...

High-level features can represent the semantics of original data and it is a plausible way to avoid problem hand-crafted for face recognition. This paper proposes an effective deep learning framework by stacking multiple output that learned through each stage Convolutional Neural Network (CNN). Different from traditional network, we use Principal Component Analysis (PCA) get filter kernels convolutional layer, which name as Stacked PCA (SPCANet). Our SPCANet model follows basic architecture...

Monocular depth estimation, as one of the fundamental tasks computer vision, plays a crucial role in three-dimensional (3D) scene understanding and perception. Usually, deep learning methods recover monocular maps using continuous regression manners by minimizing errors between ground-truth predicted depth. However, fine features may not be fully captured through layer-by-layer coding, which is prone to low spatial resolution insufficient details. Furthermore, it usually converges slowly...

Automatic speech recognition (ASR) is a fundamental technology in the field of artificial intelligence. End-to-end (E2E) ASR favored for its state-of-the-art performance. However, E2E still faces spatial information loss and text local loss, which results increase deletion substitution errors during inference. To overcome this challenge, we propose novel Enhancing Hybrid Architecture with Fast Attention Capsule Network (termed En-HACN), can model position relationships between different...

This letter considers a compute-and-forward two-way relaying channel with simultaneous wireless information and power transfer. Specifically, two single-antenna users exchange via multi-antenna relay station based on nested lattice codes. Meanwhile, energies flow from the to for circuit consumption uplink transmission. Based this model, an optimization problem is formulated minimize transmit at relay, while guaranteeing minimal transmission rate each user. To solve problem, we propose...

It is well known that higher level features can represent the abstract semantics of original data. We propose a multiple scales combined deep learning network to learn set high-level feature representations through each stage convolutional neural for face recognition, which named as multiscaled principle component analysis (PCA) Network (MS-PCANet). There are two main differences between our model and traditional network. On one hand, we get prefixed filter kernels by principal images’...

With the rapid development of 3D somatosensory technology, human behavior recognition has become an important research field. Human feature analysis evolved from traditional 2D features to features. In order improve performance activity recognition, a method is proposed, which based on hybrid texture-edge local pattern coding extraction and integration RGB depth videos information. The paper mainly focuses background subtraction video sequences behaviors, extracting integrating historical...

Nowadays, OpenID is a popular specification and provided by several large websites. It decentralized, open standard to authenticate user's identity. We don't deny that has some disadvantages. In our proposal, we try improve the performance of OpenID. The proposed authentication mechanism embedded with One-time password (OTP) technology which solves security problems also illustrate architecture OTP typical process between mobile terminal provider. Finally, analyses system results demonstrate...

Robust dynamic gesture recognition algorithm is of great value for kinds intelligent interactive systems. Most current researches on this field are based trajectory time-series, which unstable and limited. In paper, we proposed a novel method to realize by analyzing the static images with Convolutional Neural Networks (CNN). First all, new motion-capture device named Leap Motion used track fingertip positions. An effective spotting applied identify start/end points gestures. Then, map 3D...

A mobile payment solution, called SIMpass is presented in this paper, a contactless technology solution based on RFID-SIM implemented with Dual-interface SIM card. Moreover, the scheme described paper. Compare traditional NFC more cost effective and easy to use. The applications card are also discussed

In this paper, we present a new spectrum-hole prediction model for cognitive radio (CR) systems based on the IEEE 802.11 wireless local areas networks. We have also analyzed distribution of interval between two consecutive packets (ITCP) simple two-user network. For multiuser networks, investigated ITCP and study its characteristics using Bayesian inference. The effectiveness presented is validated by MATLAB simulation programs.

A middleware for the Internet of Things (IoT) diverse application protocol is proposed in this paper, which based on a system - intelligent work scene SmartScene using RFID and ZigBee protocols. With middleware, shields difference between data processing by defining conversion mapping model. common service interface upper applications also provided, allows to focus level development reduce coupling with underlying hardware protocol. Moreover, it reduces effort greatly, enhances versatility...

To reduce the blind zone in network coverage, we propose a coverage optimization algorithm of wireless sensor based on mobile nodes. This calculates irregularity and obtains minimum approximate numerical solution by utilizing quantitative relationship between energy consumption related nodes position After determining optimal relative nodes, problem static is addressed. Simulation result shows that proposed has high dynamic adaptability can address maximally. Besides increasing also reduces...

A full-duplex multiple-input multiple-output (MIMO) power splitting based relay system with channel uncertainty is considered in this paper. Full-duplex mode allows the to harvest energy and forward wireless information simultaneously from same radio frequency (RF) signal. With consideration of imperfect state (CSI), self-interference cannot be suppressed completely at node. Hence, a robust beamforming problem formulated minimum total mean squared error (MSE) network while guaranteeing...

With the increasing growth of applications deployed to cloud computing platform, providers have made a higher request for resource utilization efficiency, but there is bottlenecks in service provisioning base on VM. Against these problems, we proposed VM-Container hybrid hierarchical scheduling mechanism which classified tasks into different levels according comprehensive virtualization type and state. By using characters lightweight easy migrate, formulate initial strategy design...

PCANet, as one noticeable shallow network, employs the histogram representation for feature pooling. However, there are three main problems about this kind of pooling method. First, histogram-based method binarizes maps and leads to inevitable discriminative information loss. Second, it is difficult effectively combine other visual cues into a compact representation, because simple concatenation various inefficiency. Third, dimensionality output grows exponentially with number used. In order...

Nowadays the Internet of Things is a new concept that attracts more and attention researchers. A large amount smart objects with various communication interfaces will access Things. Unique addressing schemes unified mechanism are two basic issues. In this paper, we introduce our solutions. We designed data platform which can work as middleware between physical virtual digital world. It provides RESTful Web Service, means any object has unique URI its address be accessed through standard HTTP...