Human activity recognition (HAR) for dense prediction is proven to be of good performance, but it relies on labeling every point in time series with the high cost. In addition, the performance of HAR model will show significant degradation when tested on the sensor data with different distribution from the training data, where the training data and the test data are usually collected from different sensor locations or sensor users. Therefore, the adaptive transfer learning framework for dense prediction of HAR is introduced to implement cross-domain transfer, where the proposed multi-level unsupervised domain adaptation (MLUDA) approach combines the global domain adaptation and the specific task adaptation to adapt the source and target domain in multiple levels. The multi-connected global domain adaptation architecture is proposed for the first time, which can adapt the output layer of the encoder and the decoder in dense prediction model. After this, the specific task adaptation is proposed to ensure alignment of each class centroid in source domain and target domain by introducing the cosine distance loss and the moving average method. Experiments on three public human activity recognition datasets demonstrate that the proposed MLUDA improves the prediction accuracy of target data by 20% compared to the source domain pre-trained model and it is more effective than the other three deep transfer learning methods with an improvement of 10% to 18% in accuracy.

To improve the evolutionary algorithm performance, especially in convergence speed and global optimization ability, a self-adaptive mechanism is designed both for the conventional genetic algorithm (CGA) and the quantum inspired genetic algorithm (QIGA). For the self-adaptive mechanism, each individual was assigned with suitable evolutionary parameter according to its current evolutionary state. Therefore, each individual can evolve toward to the currently best solution. Moreover, to reduce the running time of the proposed self-adaptive mechanism based QIGA (SAM-QIGA), a multi-universe parallel structure was employed in the paper. Simulation results show that the proposed SAM-QIGA have better performances both in convergence and global optimization ability.

Cloud computing makes it possible for users to share computing power. The framework of multiple data centers gains a greater popularity in modern cloud computing. Due to the uncertainty of the requests from users, the loads of CPU(Center Processing Unit) of different data centers differ. High CPU utilization rate of a data center affects the service provided for users, while low CPU utilization rate of a data center causes high energy consumption. Therefore, it is important to balance the CPU resource across data centers in modern cloud computing framework. A virtual machine(VM) migration algorithm was proposed to balance the CPU resource across data centers. The simulation results suggest that the proposed algorithm has a good performance in the balance of CPU resource across data centers and reducing energy consumption.

Paxos is a well-known distributed algorithm that provides strong consistency. However, the original Paxos has several shortcomings, including those of slow elections, redundant communications and excessive traffic of the coordinator node. In order to tackle the above dificiencies,the design of advanced edition of Paxos(Adv Paxos) was proposed, which is a new distributed consensus algorithm that is derived from Basic Paxos. This paper analyzes the behavior of each character of the original algotithm during each of its phases.  By optimizing the behavior of the proposer and acceptor, a series of behavioral optimization measures was proposed, which included distance related waiting mechanisms, optimization of the number of proposals, self-learning and a reduction in broadcast communications. Through theoretical analysis and experimentation, it is demonstrated that the new algorithm has a lower probability of livelock without a reduction in reliability, faster agreement reaching speeds, lower communication costs among server clusters and higher percentage of successful proposals.

In order to improve the visibility and contrast of low-light images and better preserve the edge and details of images, a new low-light color image enhancement algorithm is proposed in this paper. The steps of the proposed algorithm are described as follows. First, the image is converted from the red, green and blue (RGB) color space to the hue , saturation and value (HSV) color space, and the histogram equalization (HE) is performed on the value component. Next, non-subsampled shearlet transform (NSST) is used on the value component to decompose the image into a low frequency sub-band and several high frequency sub-bands. Then, the low frequency sub-band and high frequency sub-bands are enhanced respectively by Gamma correction and improved guided image filtering (IGIF), and the enhanced value component is formed by inverse NSST transform. Finally, the image is converted back to the RGB color space to obtain the enhanced image. Experimental results show that the proposed method not only significantly improves the visibility and contrast, but also better preserves the edge and details of images.

An ensemble learning algorithm based on game theory is proposed to evaluate algorithms of image analysis and image feature extraction. A competition system is established to implement the algorithm for evaluating the applicability and efficiency of different edge detection algorithms. Through the game in the algorithm competition system, the most suitable algorithm as a winner in the competition can be selected. A group of optimal parameters for the corresponding edge detection can also be found. Firstly, based on the evolutionary game theory, a strategy of the competition of edge extraction algorithms is developed. Secondly, after selecting the most suitable algorithm from the candidates, the overall parameters are optimized. Experiments show that for a specific class of images, several candidate algorithms can be used as a class of preference algorithms based on the final evolutionary result. When analyzing the images, the priority algorithm can be recommended as the best edge detection algorithm from these reference algorithms. It is more effective than traditional methods in determining an algorithm and choosing parameters.

aphics processing is an increasing important application domain with the demand of real-time rendering, video streaming, virtual reality, and so on. Illumination is a critical module in graphics rendering and is typically compute-bound, memory-bound, and power-bound in different application cases. It is crucial to decide how to schedule different illumination algorithms with different features according to the practical requirements in reconfigurable graphics hardware. This paper analyze the performance characteristics of four main-stream lighting algorithms, Lambert illumination algorithm, Phong illumination algorithm, Blinn-Phong illumination algorithm, and Cook-Torrance illumination algorithm, using hardware performance counters on x86 processor platform KabyLake (KBL). The data movement, computation, power consumption, and memory accessing are evaluated over a range of application scenarios. Further, by analyzing the system-level behavior of these illumination algorithms, obtains the cons and pros of these specific algorithms were obtained. The associated relationship between performance/energy and the evaluated metrics was analyzed through Pearson correlation coefficient(PCC)analysis. According to these performance characterization data, this paper presents some reconfiguration suggestions in reconfigurable graphics processor.

High efficiency video coding (HEVC) uses half of the bitrate compared to H.264/advanced video coding(AVC) for encoding the same sequence with similar quality. Because of the advanced hierarchical structures of coding units (CUs), predicting units (PUs), and transform units (TUs), HEVC can better adapt when encoding full high definition (HD) and ultra high definition (UHD) videos. At the expense of encoding efficiency, the complexity of HEVC sharply increases compared to H.264/AVC, mainly due to its quad-tree structure that splits pictures. In this study, the probability distribution, which is generated by a rate distortion optimizing (RDO) cost, is analyzed. Then, an early terminating method is proposed to decrease the complexity of the HEVC based on probability distributions. The experiment shows that the coding time is reduced by 44.9% for HEVC intra coding, at the cost of a 0.61% increase in the Bjøntegaard delta rate (BD-rate), on average.

The indoor positioning system based on fingerprint receives more and more attention due to its high positioning accuracy and time efficiency. In the existing positioning approaches, much consideration is given to the positioning accuracy improvement by using the angle of signal, but the optimization of access points (APs) deployment is ignored. In this circumstance, an adaptive APs deployment approach is proposed. First of all, the criterion of reference points (RPs) effective coverage is proposed, and the number of deployed APs in target environment is obtained by using the region partition algorithm and full coverage algorithm. Secondly, the wireless signal propagation model is established for target environment, and meanwhile based on the initial APs deployment, the simulation fingerprint database is constructed for the sake of establishing the discrimination function with respect to fingerprint database. Thirdly, the greedy algorithm is applied to optimize APs deployment. Finally, the extensive experiments show that the proposed approach is capable of achieving adaptive APs deployment as well as improving positioning accuracy.

A novel single-cavity equilateral triangular substrate integrated waveguide (TSIW) bandpass filter (BPF) with a complementary triangular split ring resonator (CTSRR) is designed in this paper. A metallic via-hole is used to split the degenerate modes and adjust the transmission zeros (TZs) properly. Meanwhile, the CTSRR is utilized as a resonator to work together with the degenerate modes of the TSIW cavity. The resonant frequency of the CTSRR can be adjusted by its own size. Meanwhile, a TZ is observed in the lower band due to the CTSRR. Finally, a 16% 3dB fractional bandwidth (FBW) triple-mode TSIW BPF with three TZs in both lower and upper bands is simulated, fabricated, and measured. There is a good agreement between the simulated and measured ones.