Search

The data enhancement related technology was used to augment the small sample data set, and the multi-sample nature of the sample was enhanced. Then the face recognition model has been established based on the siamese neural network, and the face recognition model was trained by using the expanded data set. Combined with the simulation calculation, the paper tests the trained face recognition model and obtains a face recognition rate of over 90%.The result proves that the appropriate face recognition model can be established under the small sample data, and the recognition model with good recognition rate can also be trained. At the same time, the paper also analyzes the effects of different activation and loss functions on face recognition rate.

ZnO thin films were deposited by RF magnetron sputtering on glass substrates, and the Au-ZnO-Au ultraviolet photodetectors were fabricated. The effects of annealing temperature on the properties of ZnO thin films and ultraviolet photodetectors were investigated, systematically. The results show that ZnO film has a c-axis preferred orientation, the resistivity and the carrier concentration of thin film can be effecttively improved from changing annealing temperature. The average transmittance of ZnO is above 85 % in the visible range and the strong absorption of ultraviolet light can be observed. After annealing at 500°C, the dark current and photocurrent of photo-detectors can reach $1.5 \mu A$ and 3.6 mA at 10 V bias respectively.

P-Cu 2 O/n-ZnO thin film heterojunction was fabricated onto ITO glass substrate by radio frequency magnetron sputtering. Meanwhile, the characteristics of p-Cu 2 O/n-ZnO thin film heterojunction were studied systematically. The results show that ZnO thin films have the c-axis preferred orientation, and Cu2O thin films have the preferred orientation along the (110). The optical band gap energies of ZnO and Cu2O thin films are 3.3 eV and 2.8 eV, respectively. The carrier concentrations of Cu2O and ZnO thin films are thin films are $6.41\times 10^{15} {\mathrm {cm}}^{-3}$ and $1.4\times 10^{18} {\mathrm {cm}}^{-3}$, the values of resistivity are 6.41 $\Omega\bullet {\mathrm {cm}}$ and 1.4 $\Omega\bullet {\mathrm {cm}}$, respectively. Furthermore, the results demonstrated that the absorption intensity of the Cu 2 O/ZnO thin film heterojunction is significantly enhanced in visible region compared with the ZnO thin films. The forward current of the heterojunction increases obviously under illumination.

We adopted and integrated GIMMS NDVI (1982–2006) and MODIS NDVI (2001–2013) to analyze desert vegetation dynamics and its response to climatic variations in the Hexi Corridor, Northwest China. We summarized the results as follows. (1) In this study, the desert vegetation mainly showed an improvement trend during the past thirty years, In the growth season, the maximum of NDVI increased with a rate of 0.65×10−3/a (P < 0.05), and it was significantly increased in the northwestern of the Hexi Corridor and lower reaches of different inland river basin. (2) Generally, the climate of this region showed a warm-wet pattern over the last three decades, and precipitation had significant effect on the growth of desert vegetation than temperature. (3) The relationship between NDVI and precipitation became weaker in the recent ten years, especially in the lower reaches of Shiyang river basin, may be due to human activities.

The exchange of health information between patients and multiple stakeholders empowers digital home based self-care, which can deal with the challenge of increasing chronic disease prevalence. Cloud computing has the potential to provide a more imaginative long-term solution. However, concerns over pattern of data sharing, confidentiality of sensitive health data, and service integration need to be addressed. This paper aims to identify a promising and achievable cloud based solution which enables patients to share the health information when multiple stakeholders are involved for different purposes. A prototype on the case of type 2 diabetes is implemented as a proof of concept for the proposed hybrid cloud based solution. The hybrid cloud environment is composed of Open Stack and Amazon Web Services with open source libs. And the exchanging health data contains in HL7 CCD format. According to the simulation, it shows that the health data sharing between patients and multiple stakeholders has been achieved. And the problems of interoperability and privacy have been partly solved within the proposed solution.

This paper introduces a fully differential rail-to-rail input/output CMOS operational amplifier consisting of a bulk-driven input stage and an interleaved class-AB output stage. The input stage features a partial positive feedback loop to improve the gain and gain-bandwidth-product. Post-layout simulated results have shown that the proposed opamp, implemented in standard 65nm CMOS technology and with a 1.2V supply, obtains an open-loop DC gain higher than 72dB and gain-bandwidth-product above 15MHz when a 6pF load is connected to the differential output.

Video real-time communication of Intelligent terminal is the core technology for video telephone, telemedicine and tele-education. By discussing the video communication hardware platform of both IP carmer and STB and software of IP crammer based on RTSP protocol on server client, an adaptor stream transmission method for client is proposed. The real-time video communication between digital device and IP is achieved. Not only does it support both standard RTSP protocol and Microsoft RTSP protocol in same time, but also solves the communication problems between video devices and PC. It will be widely used in oriented 3C interactive applications.

Recommender systems apply knowledge discovery techniques to the problem of making products recommendations during a live customer interaction and they are achieving widespread success in e-commerce nowadays. But the traditional recommendation algorithm makes the quality of system decreased dramatically. In particular, we present an improved recommendation algorithm based on clustering and association rule to calculate the customer's nearest neighbor, and then provide the most appropriate products to meet his needs. The experimental results show the efficiency of our method.

With the evolution of mobile communication networks, three-dimensional graphics applications and services are fast developing fields in mobile graphics and wireless networks. While the data size of 3D meshes is increasing, the available bandwidth of wireless networks and the weak resources of mobile devices have become relatively limited. The transmission of graphics multimedia over wireless networks is a bottleneck problem. In this paper, a novel mobile 3D graphics compression approach for progressive transmission over wireless networks is presented. By improving the progressive coding framework using reverse Butterfly Scheme, a new progressive mesh coding algorithm based on the reverse Modified Loop scheme is proposed. The dense triangle mesh is decomposed into the progressive mesh which consists of a base mesh and a series of displacement values. Then the progressive mesh is compressed by the embedded zerotree mesh coding. The proposed approach allows the data of 3D mesh to be transmitted over wireless network using TCP and UDP protocols respectively. It speeds up 3D graphics browsing on mobile devices as well as optimizes the data size of 3D mesh for progressive transmission.