Break arcs are generated in a 48VDC resistive circuit. Circuit current I0 when electrical contacts are closed is changed from 50A to 300A. The break arcs are observed by a high-speed camera with appropriate settings of exposure from horizontal direction. Length of the break arcs L is measured from images of the break arcs. Time evolutions of the length L and gap voltage Vg are investigated. The following results are obtained. By appropriate settings of the high-speed camera, the time evolution of the length L is obtained from just after ignition to before arc extinction. Tendency of increase of the length L is similar to that of increase of the voltage Vg for each current I0.

Data rate requirements, from consumer application to automotive and aerospace grew rapidly in the last years. This led to the development of a series of communication protocols (i.e. Ethernet, PCI-Express, RapidIO and SpaceFibre), which use more than one communication lane, both to speed up data rate and to increase link reliability. Some of these protocols, such as SpaceFibre, are able to detect real-time changes in the number of active lanes and to adapt the data flow appropriately, providing a flexible solution, robust to lane failures. This results in a real time varying data path in the lower layers of the data handling system. The aim of this paper is to propose the architecture of a hardware block capable of reading a fixed number of words from a host FIFO and shaping them on a real time variable number of words equal to the number of active lanes.

With the continuous refinement of Deep Neural Networks (DNNs), a series of deep and complex networks such as Residual Networks (ResNets) show impressive prediction accuracy in image classification tasks. Unfortunately, the structural complexity and computational cost of residual networks make hardware implementation difficult. In this paper, we present the quantized and reconstructed deep neural network (QR-DNN) technique, which first inserts batch normalization (BN) layers in the network during training, and later removes them to facilitate efficient hardware implementation. Moreover, an accurate and efficient residual network accelerator (RNA) is presented based on QR-DNN with batch-normalization-free structures and weights represented in a logarithmic number system. RNA employs a systolic array architecture to perform shift-and-accumulate operations instead of multiplication operations. QR-DNN is shown to achieve a 1∼2% improvement in accuracy over existing techniques, and RNA over previous best fixed-point accelerators. An FPGA implementation on a Xilinx Zynq XC7Z045 device achieves 804.03 GOPS, 104.15 FPS and 91.41% top-5 accuracy for the ResNet-50 benchmark, and state-of-the-art results are also reported for AlexNet and VGG.

A novel image enhancement method for vein recognition is introduced. Inspired by observation that the intensity of the vein vessel changes rapidly during the smoothing process compared to that of background (i.e., skin tissue) due to its thin and long shape, we propose to exploit the smoothing speed as a restoration weight for the vein image enhancement. Experimental results based on the CASIA multispectral palm vein database demonstrate that the proposed method is effective to improve the performance of vein recognition.

The aim of this research is to support real-time drawingin talking by using multimodal user interface technologies. In this situation, if talking and drawing are considered as commands by mistake during presentation, it will disturb users' natural talking and drawing. To prevent this problem, we introduce two modes of a command mode and a free mode, and explore smooth mode switching techniques that does not interfere with users' natural talking and drawing. We evaluate four techniques. Among them, a technique that specifies the command mode after actions using a pen gesture was the most effective. In this technique, users could quickly draw diagrams, and specifying mode switching didn't interfere with users' natural talk.

The GUINNESS (GUI based binarized neural network synthesizer) is an open-source tool flow for a binarized deep neural network toward FPGA implementation based on the GUI including both the training on the GPU and inference on the FPGA. Since all the operation is done on the GUI, the software designer is not necessary to write any scripts to design the neural network structure, training behavior, only specify the values for hyperparameters. After finishing the training, it automatically generates C++ codes to synthesis the bit-stream using the Xilinx SDSoC system design tool flow. Thus, our tool flow is suitable for the software programmers who are not familiar with the FPGA design. In our tool flow, we modify the training algorithms both the training and the inference for a binarized CNN hardware. Since the hardware has a limited number of bit precision, it lacks minimal bias in training. Also, for the inference on the hardware, the conventional batch normalization technique requires additional hardware. Our modifications solve these problems. We implemented the VGG-11 benchmark CNN on the Digilent Inc. Zedboard. Compared with the conventional binarized implementations on an FPGA, the classification accuracy was almost the same, the performance per power efficiency is 5.1 times better, as for the performance per area efficiency, it is 8.0 times better, and as for the performance per memory, it is 8.2 times better. We compare the proposed FPGA design with the CPU and the GPU designs. Compared with the ARM Cortex-A57, it was 1776.3 times faster, it dissipated 3.0 times lower power, and its performance per power efficiency was 5706.3 times better. Also, compared with the Maxwell GPU, it was 11.5 times faster, it dissipated 7.3 times lower power, and its performance per power efficiency was 83.0 times better. The disadvantage of our FPGA based design requires additional time to synthesize the FPGA executable codes. From the experiment, it consumed more three hours, and the total FPGA design took 75 hours. Since the training of the CNN is dominant, it is considerable.

Saliency quality assessment aims at estimating the objective quality of a saliency map without access to the ground-truth. Existing works typically evaluate saliency quality by utilizing information from saliency maps to assess its compactness and closedness while ignoring the information from image content which can be used to assess the consistence and completeness of foreground. In this letter, we propose a novel multi-information fusion network to capture the information from both the saliency map and image content. The key idea is to introduce a siamese module to collect information from foreground and background, aiming to assess the consistence and completeness of foreground and the difference between foreground and background. Experiments demonstrate that by incorporating image content information, the performance of the proposed method is significantly boosted. Furthermore, we validate our method on two applications: saliency detection and segmentation. Our method is utilized to choose optimal saliency map from a set of candidate saliency maps, and the selected saliency map is feeded into an segmentation algorithm to generate a segmentation map. Experimental results verify the effectiveness of our method.

The IEEE 802.11p Enhanced Distributed Channel Access (EDCA) is a standardization for vehicle-to-vehicle and road-to-vehicle communications. The saturated throughputs of the IEEE 802.11p EDCA obtained from previous analytical expressions differ from those of simulations. The purpose of this paper is to explain the reason why the differences appear in the previous analytical model of the EDCA. It is clarified that there is a special state wherein the Backoff Timer (BT) is decremented in the first time slot of after a frame transmission, which cannot be expressed in the previous Markov model. In addition, this paper proposes modified Markov models, which allow the IEEE 802.11p EDCA to be correctly analyzed. The proposed models describe BT-decrement procedure in the first time slot accurately by adding new states to the previous model. As a result, the proposed models provide accurate transmission probabilities of network nodes. The validity of the proposed models is confirmed by the quantitative agreements between analytical predictions and simulation results.

This study presents a joint dictionary learning approach for speech emotion recognition named locality preserved joint nonnegative matrix factorization (LP-JNMF). The learned representations are shared between the learned dictionaries and annotation matrix. Moreover, a locality penalty term is incorporated into the objective function. Thus, the system's discriminability is further improved.

The recent widespread use of high-performance terminals has resulted in a rapid increase in mobile data traffic. Therefore, public wireless local area networks (WLANs) are being used often to supplement the cellular networks. Capacity improvement through the dense deployment of access points (APs) is being considered. However, the effective throughput degrades significantly when many users connect to a single AP. In this paper, users are classified into guaranteed bit rate (GBR) users and best effort (BE) users, and we propose a network model to provide those services. In the proposed model, physical APs and the bandwidths are assigned to each service class dynamically using a virtual AP configuration and a virtualized backhaul network, for reducing the call-blocking probability of GBR users and improving the satisfaction degree of BE users. Finally, we evaluate the performance of the proposed model through simulation experiments and discuss its feasibility.

To recognize transportation modes without any additional sensor devices, we demonstrate that the transportation modes can be recognized from communication quality factors. In the demonstration, instead of using global positioning system (GPS) and accelerometer sensors, we collect mobile TCP throughputs, received-signal strength indicators (RSSIs), and cellular base-station IDs (Cell IDs) through in-line network measurement when the user enjoys mobile services, such as video streaming. In accuracy evaluations, we conduct two different field experiments to collect the data in six typical transportation modes (static, walking, riding a bicycle, riding a bus, riding a train and riding a subway), and then construct the classifiers by applying a support-vector machine (SVM), k-nearest neighbor (k-NN), random forest (RF), and convolutional neural network (CNN). Our results show that these transportation modes can be recognized with high accuracy by using communication quality factors as well as the use of accelerometer sensors.

This paper presents a novel model in the field of image co-saliency detection. Previous works simply design low level handcrafted features or extract deep features based on image patches for co-saliency calculation, which neglect the entire object perception properties. Besides, they also neglect the problem of visual similar region's mismatching when designing co-saliency calculation model. To solve these problems, we propose a novel strategy by considering both local prediction and global refinement (LPGR). In the local prediction stage, we train a deep convolutional saliency detection network in an end-to-end manner which only use the fully convolutional layers for saliency map prediction to capture the entire object perception properties and reduce feature redundancy. In the global refinement stage, we construct a unified co-saliency refinement model by integrating global appearance similarity into a co-saliency diffusion function, realizing the propagation and optimization of local saliency values in the context of entire image group. To overcome the adverse effects of visual similar regions' mismatching, we innovatively incorporates the inter-images saliency spread constraint (ISC) term into our co-saliency calculation function. Experimental results on public datasets demonstrate consistent performance gains of the proposed model over the state-of-the-art methods.

Biometric authentication, namely using biometric features for authentication is gaining popularity in recent years as further modalities, such as fingerprint, iris, face, voice, gait, and others are exploited. We explore the effectiveness of three simple Electroencephalography (EEG) related biometric authentication tasks, namely resting, thinking about a picture, and moving a single finger. We present details of the data processing steps we exploit for authentication, including extracting features from the frequency power spectrum and MFCC, and training a multilayer perceptron classifier for authentication. For evaluation purposes, we record an EEG dataset of 27 test subjects. We use three setups, baseline, task-agnostic, and task-specific, to investigate whether person-specific features can be detected across different tasks for authentication. We further evaluate, whether different tasks can be distinguished. Our results suggest that tasks are distinguishable, as well as that our authentication approach can work both exploiting features from a specific, fixed, task as well as using features across different tasks.

Given a graph G=(V,E), where V and E are vertex and edge sets of G, and a subset VNT of vertices called a non-terminal set, the minimum spanning tree with a non-terminal set VNT, denoted by MSTNT, is a connected and acyclic spanning subgraph of G that contains all vertices of V with the minimum weight where each vertex in a non-terminal set is not a leaf. On general graphs, the problem of finding an MSTNT of G is NP-hard. We show that if G is a series-parallel graph then finding an MSTNT of G is linearly solvable with respect to the number of vertices.

Most recent work used raw electroencephalograph (EEG) data to train deep learning (DL) models, with the assumption that DL models can learn discriminative features by itself. It is not yet clear what kind of RSVP specific features can be selected and combined with EEG raw data to improve the RSVP classification performance of DL models. In this paper, we tried to extract RSVP specific features and combined them with EEG raw data to capture more spatial and temporal correlations of target or non-target event and improve the EEG-based RSVP target detection performance. We tested on X2 Expertise RSVP dataset to show the experiment results. We conducted detailed performance evaluations among different features and feature combinations with traditional classification models and different CNN models for within-subject and cross-subject test. Compared with state-of-the-art traditional Bagging Tree (BT) and Bayesian Linear Discriminant Analysis (BLDA) classifiers, our proposed combined features with CNN models achieved 1.1% better performance in within-subject test and 2% better performance in cross-subject test. This shed light on the ability for the combined features to be an efficient tool in RSVP target detection with deep learning models and thus improved the performance of RSVP target detection.

Most wireless networks are specified as using the data link protocol, i.e. layer 2 (L2). Recently, IoT and big data processing have promoted the use of wireless sensor networks to connect and send data to data center applications over the Internet. To do so, the implementation of an IP stack on the wireless node, or the gateway of the IP and wireless L2 network, has been proposed. Both approaches were developed to allow applications on the IP network to access L2 wireless network nodes. However, since wireless sensor networks do not require any network protocol, an IP is not essential for collecting data. Therefore, we propose a novel bridging VPN for connecting wireless networks, in which the application and wireless end nodes are not required to acknowledge an IP address or network protocol. In this way, the IP network merely serves to transport the data link frames of wireless networks. We believe that this is another style of IoT and recommend that this VPN be used as a test bed for small IoT businesses and institutions before they start to implement an IP stack on their systems.

DHT routing algorithms can provide efficient mechanisms for resource placement and lookup for distributed file sharing systems. However, we must still deal with irregular and frequent join/leave of nodes and the problem of load unbalancing between nodes in DHT-based file sharing systems. This paper presents an efficient file backup scheme based on dynamic DHT key space clustering in order to guarantee data availability and support load balancing. The main idea of our method is to dynamically divide the DHT network into a number of clusters, each of which locally stores and maintains data chunks of data files to guarantee the data availability of user data files even when node churn occurs. Further, high-capacity nodes in clusters are selected as backup nodes to achieve adequate load balancing. Simulation results demonstrate the superior effectiveness of the proposed scheme over other file replication schemes.

By taking advantages of deep learning and reinforcement learning, ADNet (Action Decision Network) outperforms other approaches. However, its speed and performance are still limited by factors such as unreliable confidence score estimation and redundant historical actions. To address the above limitations, a faster and more accurate approach named Faster-ADNet is proposed in this paper. By optimizing the tracking process via a status re-identification network, the proposed approach is more efficient and 6 times faster than ADNet. At the same time, the accuracy and stability are enhanced by historical actions removal. Experiments demonstrate the advantages of Faster-ADNet.

Given two feasible solutions A and B, a reconfiguration problem asks whether there exists a reconfiguration sequence (A0=A, A1,...,Aℓ=B) such that (i) A0,...,Aℓ are feasible solutions and (ii) we can obtain Ai from Ai-1 under the prescribed rule (the reconfiguration rule) for each i ∈ {1,...,ℓ}. In this paper, we address the reconfiguration problem for induced trees, where an induced tree is a connected and acyclic induced subgraph of an input graph. We consider the following two rules as the prescribed rules: Token Jumping: removing u from an induced tree and adding v to the tree, and Token Sliding: removing u from an induced tree and adding v adjacent to u to the tree, where u and v are vertices of an input graph. As the main results, we show that (I) the reconfiguration problemis PSPACE-complete even if the input graph is of bounded maximum degree, (II) the reconfiguration problem is W[1]-hard when parameterized by both the size of induced trees and the length of the reconfiguration sequence, and (III) there exists an FPT algorithm when the problem is parameterized by both the size of induced trees and the maximum degree of an input graph under Token Jumping and Token Sliding.

This paper introduces a new noise generation algorithm for vocoder-based speech waveform generation. White noise is generally used for generating an aperiodic component. Since short-term white noise includes a zero-frequency component (ZFC) and inaudible components below 20 Hz, they are reduced in advance when synthesizing. We propose a new noise generation algorithm based on that for velvet noise to overcome the problem. The objective evaluation demonstrated that the proposed algorithm can reduce the unwanted components.