To improve TCP throughput even if the maximum receiving window size is small, a TCP performance enhancing proxy (PEP) using a UDP-like packet sending policy with error control has been proposed. The PEP operates on a router along a TCP connection. When the PEP receives a data packet from the source host, it transmits the packet to the destination host, copies the packet into the local buffer (PEP buffer) in case the packets need to be transmitted and sends a premature ACK acknowledging receipt of the packet to the source host. In the PEP, the number of prematurely acknowledged packets in the PEP buffer is limited to a fixed threshold (watermark) value to avoid network congestion. Although the watermark value should be adjusted to changes in the network conditions, watermark adjusting algorithms have not been investigated. In this paper, we propose a watermark adjusting algorithm the goal of which is to maximize the throughput of each connection as much as possible without excessively suppressing the throughputs of the other connections. In our proposed algorithm, a newly established connection uses the initial watermark value of zero to avoid drastic network congestion and increases the value as long as its throughput increases. In addition, when a new connection is established, every already-established connection halves its watermark value to allow the newly established connection to use some portion of the bandwidth and increases again as long as its throughput increases. We compare the proposed algorithm (CW method) with other methods: the FW method that uses a fixed large watermark value and the NP method that does not use the PEP. Numerical results with respect to throughput and fairness showed that the CW method is generally superior to the other two methods.

Multi-view video can provide users with three-dimensional (3-D) and virtual reality perception through multiple viewing angles. In recent years, depth image-based rendering (DIBR) has been generally used to synthesize virtual view images in free viewpoint television (FTV) and 3-D video. To conceal the zero-region more accurately and improve the quality of a virtual view synthesized frame, an integrated hole-filling algorithm for view synthesis is proposed in this paper. The proposed algorithm contains five parts: an algorithm for distinguishing different regions, foreground and background boundary detection, texture image isophotes detection, a textural and structural isophote prediction algorithm, and an in-painting algorithm with gradient priority order. Based on the texture isophote prediction with a geometrical principle and the in-painting algorithm with a gradient priority order, the boundary information of the foreground is considerably clearer and the texture information in the zero-region can be concealed much more accurately than in previous works. The vision quality mainly depends on the distortion of the structural information. Experimental results indicate that the proposed algorithm improves not only the objective quality of the virtual image, but also its subjective quality considerably; human vision is also clearly improved based on the subjective results. In particular, the algorithm ensures the boundary contours of the foreground objects and the textural and structural information.

BEAN is a newly proposed lightweight stream cipher adopting Fibonacci FCSRs. It is designed for very constrained environments and aims at providing a balance between security, efficiency and cost. A weakness in BEAN was first found by Å gren and Hell in 2011, resulting in a key recovery attack slightly better than brute force. In this paper, we present new correlations between state and keystream with large statistical advantage, leading to a much more efficient key recovery attack. The time and data complexities of this attack are 257.53 and 259.94, respectively. Moreover, two new output functions are provided as alternatives, which are more efficent than the function used in BEAN and are immune to all attacks proposed on the cipher. Also, suggestions for improving the FCSRs are given.

RSPICE, a fast timing simulator for large digital MOS circuits, is presented in this paper. A new table-based region-wise linear MOS transistor model and the analytical solution of the generic sub-circuit primitive are applied to calculate the transient response of digital MOS circuits. The body effect of pass transistors is included in the MOS model and the floating capacitor network can be handled by this sub-circuit primitive as well. In RSPICE, MOS transistors with a DC path are grouped into a DC-connected block (DCCB), and DCCBs with a feedback path are combined as a strongly connected component (SCC). RSPICE orders SCCs by Tarjan's algorithm and simulates ordered SCCs one by one. DCCBs are basic cells in RSPICE and any DCCB can be mapped into one or more sub-circuit primitives. In order to calculate the transient response of these primitives analytically, RSPICE approximates the input signals of the primitive by piecewise linear functions. To compromise the simulation accuracy and run time, partial waveform and partial time convergent (PWPTC) combined with dynamic windowing technique is applied to simulate SCCs. Other key issues of RSPICE, such as circuit partition, pass-transistor and floating-capacitor processing, simulation-flow control and waveform modification are also discussed in detail. Compared with HSPICE , the simulation result of RSPICE is very accurate with an error less than 3%, but the speed is 1-2 orders over HSPICE.

As a research hotspot and difficulty in the field of computer vision, pedestrian detection has been widely used in intelligent driving and traffic monitoring. The popular detection method at present uses region proposal network (RPN) to generate candidate regions, and then classifies the regions. But the RPN produces many erroneous candidate areas, causing region proposals for false positives to increase. This letter uses improved residual attention network to capture the visual attention map of images, then normalized to get the attention score map. The attention score map is used to guide the RPN network to generate more precise candidate regions containing potential target objects. The region proposals, confidence scores, and features generated by the RPN are used to train a cascaded boosted forest classifier to obtain the final results. The experimental results show that our proposed approach achieves highly competitive results on the Caltech and ETH datasets.

A frontal and realistic face image was synthesized from a single profile face image. It has a wide range of applications in face recognition. Although the frontal face method based on deep learning has made substantial progress in recent years, there is still no guarantee that the generated face has identity consistency and illumination consistency in a significant posture. This paper proposes a novel pixel-based feature regression generative adversarial network (PFR-GAN), which can learn to recover local high-frequency details and preserve identity and illumination frontal face images in an uncontrolled environment. We first propose a Reslu block to obtain richer feature representation and improve the convergence speed of training. We then introduce a feature conversion module to reduce the artifacts caused by face rotation discrepancy, enhance image generation quality, and preserve more high-frequency details of the profile image. We also construct a 30,000 face pose dataset to learn about various uncontrolled field environments. Our dataset includes ages of different races and wild backgrounds, allowing us to handle other datasets and obtain better results. Finally, we introduce a discriminator used for recovering the facial structure of the frontal face images. Quantitative and qualitative experimental results show our PFR-GAN can generate high-quality and high-fidelity frontal face images, and our results are better than the state-of-art results.

Laser Doppler Vibrometers (LDVs) enable the acquisition of remote speech signals by measuring small-scale vibrations around a target. They are now widely used in the fields of information acquisition and national security. However, in remote speech detection, the coherent measurement signal is subject to environmental noise, making detecting and reconstructing speech signals challenging. To improve the detection distance and speech quality, this paper proposes a highly accurate real-time speech measurement method that can reconstruct speech from noisy coherent signals. First, the I/Q demodulation and arctangent phase discrimination are used to extract the phase transformation caused by the acoustic vibration from coherent signals. Then, an innovative smoothness criterion and a novel phase difference-based dynamic bilateral compensation phase unwrapping algorithm are used to remove any ambiguity caused by the arctangent phase discrimination in the previous step. This important innovation results in the highly accurate detection of phase jumps. After this, a further innovation is used to enhance the reconstructed speech by applying an improved waveform-based linear prediction coding method, together with adaptive spectral subtraction. This removes any impulsive or background noise. The accuracy and performance of the proposed method were validated by conducting extensive simulations and comparisons with existing techniques. The results show that the proposed algorithm can significantly improve the measurement of speech and the quality of reconstructed speech signals. The viability of the method was further assessed by undertaking a physical experiment, where LDV equipment was used to measure speech at a distance of 310m in an outdoor environment. The intelligibility rate for the reconstructed speech exceeded 95%, confirming the effectiveness and superiority of the method for long-distance laser speech measurement.

Time synchronization is an essential service for wireless sensor networks (WSNs). However, fixed-period time synchronization can not serve multiple users efficiently in terms of energy consumption. This paper proposes a lightweight precision-adaptive protocol for cluster-based multi-user networks. It consists of a basic average time synchronization algorithm and an adaptive control loop. The basic average time synchronization algorithm achieves 1 µs instantaneous synchronization error performance. It also prolongs re-synchronization period by taking the average of two specified nodes' local time to be cluster global time. The adaptive control loop realizes diverse levels of synchronization precision based on the proportional relationship between sync error and re-synchronization period. Experimental results show that the proposed precision-adaptive protocol can respond to the sync error bound change within 2 steps. It is faster than the exponential convergence of the adaptive protocols based on multiplicative iterations.

In this letter, we propose a novel and effective haze removal method by using the structure-aware atmospheric veil. More specifically, the initial atmospheric veil is first estimated based on dark channel prior and morphological operator. Furthermore, an energy optimization function considering the structure feature of the input image is constructed to refine the initial atmospheric veil. At last, the haze-free image can be restored by inverting the atmospheric scattering model. Additionally, brightness adjustment is also performed for preventing the dehazing result too dark. Experimental results on hazy images reveal that the proposed method can effectively remove the haze and yield dehazing results with vivid color and high scene visibility.

This paper puts forward the concept of cellular network location with less information which can overcome the weaknesses of the cellular location technology in practical applications. After a systematic introduction of less-information location model, this paper presents a location algorithm based on AGA (Adaptive Genetic Algorithm) and an optimized RBF (Radical Basis Function) neural network. The virtues of this algorithm are that it has high location accuracy, reduces the location measurement parameters and effectively enhances the robustness. The simulation results show that under the condition of less information, the optimized location algorithm can effectively solve the fuzzy points in the location model and satisfy the FCC's (Federal Communications Commission) requirements on location accuracy.

In parallelizing compilers on distributed memory systems, distributions of irregular sized array blocks are provided for load balancing and irregular problems. The irregular data redistribution is different from the regular block-cyclic redistribution. This paper is devoted to scheduling message for irregular data redistribution that attempt to obtain suboptimal solutions while satisfying the minimal communication costs condition and the minimal step condition. Based on the list scheduling, an efficient algorithm is developed and its experimental results are compared with previous algorithms. The improved list algorithm provides more chance for conflict messages in its relocation phase, since it allocates conflict messages through methods used in a divide-and-conquer algorithm and a relocation algorithm proposed previously. The method of selecting the smallest relocation cost guarantees that the improved list algorithm is more efficient than the other two in average.

This paper studies the problem of recovering an arbitrarily distributed sparse matrix from its one-bit (1-bit) compressive measurements. We propose a matrix sketching based binary method iterative hard thresholding (MSBIHT) algorithm by combining the two dimensional version of BIHT (2DBIHT) and the matrix sketching method, to solve the sparse matrix recovery problem in matrix form. In contrast to traditional one-dimensional BIHT (BIHT), the proposed algorithm can reduce computational complexity. Besides, the MSBIHT can also improve the recovery performance comparing to the 2DBIHT method. A brief theoretical analysis and numerical experiments show the proposed algorithm outperforms traditional ones.

This paper proposes a region-of-interest (ROI) based H.264 encoder and the VLSI architecture of the ROI detection algorithm. In ROI based video coding system, pre-processing unit to detect ROI should only introduce low computational complexity overhead due to the low power requirement. The Macroblocks (MBs) in ROIs are detected sequentially in the same order of H.264 encoding to satisfy the MB level pipelining of ROI detector and H.264 encoder. ROI detection is performed in a novel estimation-and-verification process with an ROI contour template. Proposed architecture can be configured to detect either single ROI or multiple ROIs in each frame and the throughput of single detection mode is 5.5 times of multiple detection mode. 98.01% and 97.89% of MBs in ROIs can be detected in single and multiple detection modes respectively. Hardware cost of proposed architecture is only 4.68 k gates. Detection speed is 753 fps for CIF format video at the operation frequency of 200 MHz in multiple detection mode with power consumption of 0.47 mW. Compared with previous fast ROI detection algorithms for video coding application, the proposed architecture obtains more accurate and smaller ROI. Therefore, more efficient ROI based computation complexity and compression efficiency optimization can be implemented in H.264 encoder.

A low-power low-noise intermediate-frequency (IF) circuit is proposed for Gaussian frequency shift keying (GFSK) low-IF receivers. The proposed IF circuit is realized by an all-analog architecture composed of a couple of limiting amplifiers (LAs) and received signal strength indicators (RSSIs), a couple of band-pass filters (BPFs), a frequency detector (FD), a low-pass filter (LPF) and a slicer. The LA and RSSI are realized by an optimized combination of folded amplifiers and current subtractor based rectifiers to avoid the process induced depressing on accuracy. In addition, taking into account the nonlinearity and static current of rectifiers, we propose an analytical model as an accurate approximation of RSSIs' transfer character. An active-RC based GFSK demodulation scheme is proposed, and then both low power consumption and a large dynamic range are obtained. The chip is implemented with HJTC 0.18 µm CMOS technology and measured under an intermediate frequency of 200 kHz, a data rate of 100 kb/s and a modulation index of 1. The RSSI has a dynamic range of 51 dB with a logarithmic linearity error of less than 1 dB, and the slope is 23.9 mV/dB. For 0.1% bit error ratio (BER), the proposed IF circuit has the minimum input signal-to-noise ratio (SNR) of 5 dB and an input dynamic range of 55.4 dB, whereas it can tolerate a frequency offset of -3%+9.5% at 6 dB input SNR. The total power consumption is 5.655.89 mW.

In this paper, we explicitly construct a large class of symmetric Boolean functions on 2k variables with algebraic immunity not less than d, where integer k is given arbitrarily and d is a given suffix of k in binary representation. If let d = k, our constructed functions achieve the maximum algebraic immunity. Remarkably, 2⌊ log2k ⌋ + 2 symmetric Boolean functions on 2k variables with maximum algebraic immunity are constructed, which are much more than the previous constructions. Based on our construction, a lower bound of symmetric Boolean functions with algebraic immunity not less than d is derived, which is 2⌊ log2d ⌋ + 2(k-d+1). As far as we know, this is the first lower bound of this kind.

SUMPLE, one of important signal combining approaches, its combining loss increases when a sensor in an array fails. A novel failure detection circuit for SUMPLE is proposed by using variability index. This circuit can effectively judge whether a sensor fails or not. Simulation results validate its effectiveness with respect to the existing algorithms.

This paper presents a high-performance dual-issue 32-core SIMD platform for image and video processing. The SIMD cores support 8/16 bits SIMD MAC instructions, and vertical vector access. Eight cores with a 4-ports L2 cache are connected by CIB bus as a cluster. Four clusters are connected by mesh network. This hierarchical network can provide more than 192 GB/s low latency inter-core BW in average. The 4-ports L2 cache architecture is also designed to provide 192 GB/s L2 cache BW. To reduce coherence operation in large-scale SMP, an application specified protocol is proposed. Compared with MOESI, 67.8% of L1 cache energy can be saved in 32 cores case. The whole system including 32 vector cores, 256 KB L2 cache, 64-bit DDRII PHY and two PLL units, occupy 25 mm2 in 65 nm CMOS. It can achieve a peak performance of 375 GMACs and 98 GMACs/W at 1.2 V.

Construction of multiple output functions is one of the most important problems in the design and analysis of stream ciphers. Generally, such a function has to be satisfied with several criteria, such as high nonlinearity, resiliency and high algebraic degree. But there are mutual restraints among the cryptographic parameters. Finding a way to achieve the optimization is always regarded as a hard task. In this paper, by using the disjoint linear codes and disjoint spectral functions, two classes of resilient multiple output functions are obtained. It has been proved that the obtained functions have high nonlinearity and high algebraic degree.

Thin film glucose biosensors were fabricated with organic/inorganic hybrid films based on glucose oxidase (GOx) and Prussian Blue nano-clusters. The biosensors composed of hybrid films were characterized by the low operating potential and the advantage to interference-free detection. In this research, we employed two kinds of thin films for GOx immobilization: Langmuir-Blodgett (LB) and self-assembled monolayer (SAM). The LB film immobilizes GOx in its inside through the electrostatic force, while the SAM immobilizes GOx with the covalent bond. The sensors with LB film produced a relatively high current signal, while the non-linear behavior and a low stability were recognized. On the other hand, the sensors with SAM presented a good linear relationship and a very stable performance.

Selective attention mechanism, plays an important role in human visual perception, can be investigated by developing an approach to perceiving the multi-meaningful-dotted-pattern in a color blindness plate (CBP). In this Letter, a perception model driven by a simple active vision mechanism is presented for the image segmentation and understanding of a CBP. Experiments show that to understand one meaningful pattern in an image containing multi-meaningful patterns, the active visual search (i.e., pattern attention) is a very useful function.