Recently, cloud systems have started to be utilized for services which analyze user's data in the field of computer vision. In these services, keypoints are extracted from images or videos, and the data is identified by machine learning with a large database in the cloud. To reduce the number of keypoints which are sent to the cloud, Keypoints of Interest (KOI) extraction has been proposed. However, since its computational complexity is large, hardware implementation is required for real-time processing. Moreover, the hardware resource must be low because it is embedded in devices of users. This paper proposes a hardware-friendly KOI algorithm with low amount of computations and its real-time hardware implementation based on dual threshold keypoint detection by gradient histogram and parallelization of connectivity of adjacent keypoint-utilizing register counters. The algorithm utilizes dual-histogram based detection and keypoint-matching based calculation of motion information and dense-clustering based keypoint smoothing. The hardware architecture is composed of a detection module utilizing descriptor, and grid-region-parallelization based density clustering. Finally, the evaluation results of hardware implementation show that the implemented hardware achieves Full-HD (1920x1080)-60 fps spatio-temporal keypoint extraction. Further, it is 47 times faster than low complexity keypoint extraction on software and 12 times faster than spatio-temporal keypoint extraction on software, and the hardware resources are almost the same as SIFT hardware implementation, maintaining accuracy.

A method of color scheme is proposed considering contrast of luminance between adjacent regions and design property. This method aims at setting the contrast of luminance high, in order to make the image understandable to visually handicapped people. This method also realizes preferable color design for visually normal people by assigning color components from color combination samples. Interactive evolutionary computing is adopted to design the luminance and the color, so that the luminance and color components are assigned to each region appropriately on the basis of human subjective criteria. Here, the luminance is designed first, and then color components are assigned, keeping the luminance unchanged. Since samples of fine color combinations are applied, the obtained color design is also fine and harmonic. Computer simulations verify the high performance of this system.

Although many approaches about ideal channels have been proposed in previous researches, few authors considered the situation of nonideal communication links. In this paper, we study the problem of distributed decision fusion over nonideal channels by using the scan statistics. In order to obtain the fusion rule under nonideal channels, we set up the nonideal channels model with the modulation error, noise and signal attenuation. Under this model, we update the fusion rule by using the scan statstics. We firstly consider the fusion rule when sensors are distributed in grid, then derive the expressions of the detection probability and false alarm probability when sensors follow an uniform distribution. Extensive simulations are conducted in order to investigate the performance of our fusion rule and the influence of signal-noise ratio (SNR) on the detection and false alarm probability. These simulations show that the theoretical values of the global detection probability and the global false alarm probability are close to the experimental results, and the fusion rule also has high performance at the high SNR region. But there are some further researches need to do for solving the large computational complexity.

A number of parallel applications run on a high-performance computing (HPC) system simultaneously. Job mapping and scheduling become crucial to improve system utilization, because fragmentation prevents an incoming job from being assigned even if there are enough compute nodes unused. Wireless supercomputers and datacenters with free-space optical (FSO) terminals have been proposed to replace the conventional wired interconnection so that a diverse application workload can be better supported by changing their network topologies. In this study we firstly present an efficient job mapping by swapping the endpoints of FSO links in a wireless HPC system. Our evaluation shows that an FSO-equipped wireless HPC system can achieve shorter average queuing length and queuing time for all the dispatched user jobs. Secondly, we consider the use of a more complicated and enhanced scheduling algorithm, which can further improve the system utilization over different host networks, as well as the average response time for all the dispatched user jobs. Finally, we present the performance advantages of the proposed wireless HPC system under more practical assumptions such as different cabinet capacities and diverse subtopology packings.

In this paper, we introduce a fully automatic approach to construct action datasets from noisy Web video search results. The idea is based on combining cluster structure analysis and density-based outlier detection. For a specific action concept, first, we download its Web top search videos and segment them into video shots. We then organize these shots into subsets using density-based hierarchy clustering. For each set, we rank its shots by their outlier degrees which are determined as their isolatedness with respect to their surroundings. Finally, we collect high ranked shots as training data for the action concept. We demonstrate that with action models trained by our data, we can obtain promising precision rates in the task of action classification while offering the advantage of fully automatic, scalable learning. Experiment results on UCF11, a challenging action dataset, show the effectiveness of our method.

Diabetes mellitus is a group of metabolic diseases that cause high blood sugar due to functional problems with the pancreas or metabolism. Diabetic patients have few subjective symptoms and may experience decreased sensation without being aware of it. The commonly performed tests for sensory disorders are qualitative in nature. The authors pay attention to the decline of the sensitivity of tactile sensations, and develop a non-invasive method to detect the level of tactile sensation using a novel micro-vibration actuator that employs shape-memory alloy wires. Previously, we performed a pilot study that applied the device to 15 diabetic patients and confirmed a significant reduction in the tactile sensation in diabetic patients when compared to healthy subjects. In this study, we focus on the asymptomatic development of decreased sensation associated with diabetes mellitus. The objectives are to examine diabetic patients who are unaware of abnormal or decreased sensation using the quantitative tactile sensation measurement device and to determine whether tactile sensation is decreased in patients compared to healthy controls. The finger method is used to measure the Tactile Sensation Threshold (TST) score of the index and middle fingers using the new device and the following three procedures: TST-1, TST-4, and TST-8. TST scores ranged from 1 to 30 were compared between the two groups. The TST scores were significantly higher for the diabetic patients (P<0.05). The TST scores for the left fingers of diabetic patients and healthy controls were 5.9±6.2 and 2.7±2.9 for TST-1, 15.3±7.0 and 8.7±6.4 for TST-4, and 19.3±7.8 and 12.7±9.1 for TST-8. Our data suggest that the use of the new quantitative tactile sensation measurement device enables the detection of decreased tactile sensation in diabetic patients who are unaware of abnormal or decreased sensation compared to controls.

In this letter, we propose a novel discriminative representation for patterned fabric defect inspection when only limited negative samples are available. Fisher criterion is introduced into the loss function of deep learning, which can guide the learning direction of deep networks and make the extracted features more discriminating. A deep neural network constructed from the encoder part of trained autoencoders is utilized to classify each pixel in the images into defective or defectless categories, using as context a patch centered on the pixel. Sequentially the confidence map is processed by median filtering and binary thresholding, and then the defect areas are located. Experimental results demonstrate that our method achieves state-of-the-art performance on the benchmark fabric images.

In general, in-focus images are used in visual object detection because image blur is considered as a factor reducing detection accuracy. However, in-focus images make it difficult to separate target objects from background images, because of that, visual object detection becomes a hard task. Background subtraction and inter-frame difference are famous schemes for separating target objects from background but they have a critical disadvantage that they cannot be used if illumination changes or the point of view moves. Considering these problems, the authors aim to improve detection accuracy by using images with out-of-focus blur obtained from a camera with a shallow depth of field. In these images, it is expected that target objects become in-focus and other regions are blurred. To enable visual object detection based on such image blur, this paper proposes a novel scheme using DFT-based feature extraction. The experimental results using synthetic images including, circle, star, and square objects as targets showed that a classifier constructed by the proposed scheme showed 2.40% miss rate at 0.1 FPPI and perfect detection has been achieved for detection of star and square objects. In addition, the proposed scheme achieved perfect detection of humans in natural images when the upper half of the human body was trained. The accuracy of the proposed scheme is better than the Filtered Channel Features, one of the state-of-the-art schemes for visual object detection. Analyzing the result, it is convincing that the proposed scheme is very feasible for visual object detection based on image blur.

This paper proposes a Watts and Strogatz-model based routing method for wireless sensor network along with link-exchange operation. The proposed routing achieves low data-collection delay because of hub-node existence. By applying the link exchanges, node with low remaining battery level can escape from a hub node. Therefore, the proposed routing method achieves the fair battery-power consumptions among sensor nodes. It is possible for the proposed method to prolong the network lifetime with keeping the small-world properties. Simulation results show the effectiveness of the proposed method.

This letter introduces a non-local means (NLM) denoising algorithm that uses a weight function based on a switching norm. The noise level and local activity are incorporated into the NLM denoising algorithm which enhances performance. This is done by selecting a norm among l1, l2, and l4 norms to determine a weighting function. The experimental results show the capability of the proposed algorithm. In addition, the proposed algorithm is verified as effective for enhancing the performance of other NLM algorithms.

This paper describes a novel harmonic-based robust voice activity detection (H-RVAD) method with harmonic spectral local peak (HSLP) feature. HSLP is extracted by spectral amplitude analysis between the adjacent formants, and such characteristic can be used to identify and verify audio stream containing meaningful human speech accurately in low SNR environment. And, an enhanced low SNR noisy speech recognition system framework with wakeup module, speech recognition module and confirmation module is proposed. Users can determine or reject the system feedback while a recognition result was given in the framework, to prevent any chance that the voiced noise misleads the recognition result. The H-RVAD method is evaluated by the AURORA2 corpus in eight types of noise and three SNR levels and increased overall average performance from 4% to 20%. In home noise, the performance of H-RVAD method can be performed from 4% to 14% sentence recognition rate in average.

This paper presents a novel control method based on predictions of a neural network in coordination with a conventional PID control to improve transient characteristics of digitally controlled switching dc-dc converters. Power supplies in communication systems require to achieve a superior operation for electronic equipment installed to them. Especially, it is important to improve transient characteristics in any required conditions since they affect to the operation of power supplies. Therefore, dc-dc converters in power supplies need a superior control method which can suppress transient undershoot and overshoot of output voltage. In the presented method, the neural network is trained to predict the output voltage and is adopted to modify the reference value in the PID control to reduce the difference between the output voltage and its desired one in the transient state. The transient characteristics are gradually improved as the training procedure of the neural network is proceeded repetitively. Furthermore, the timing and duration of neural network control are also investigated and devised since the time delay, which is one of the main issue in digital control methods, affects to the improvement of transient characteristics. The repetitive training and duration adjustment of the neural network are performed simultaneously to obtain more improvement of the transient characteristics. From simulated and experimental results, it is confirmed that the presented method realizes superior transient characteristics compared to the conventional PID control.

Hough-based voting approaches have been widely used to solve many detection problems such as object and action detection. These approaches for action detection cast votes for action classes and positions based on the local spatio-temporal features of given videos. The voting process of each local feature is performed independently of the other local features. This independence enables the method to be robust to occlusions because votes based on visible local features are not influenced by occluded local features. However, such independence makes discrimination of similar motions between different classes difficult and causes the method to cast many false votes. We propose a novel Hough-based action detection method to overcome the problem of false votes. The false votes do not occur randomly such that they depend on relevant action classes. We introduce vote distributions, which represent the number of votes for each action class. We assume that the distribution of false votes include important information necessary to improving action detection. These distributions are used to build a model that represents the characteristics of Hough voting that include false votes. The method estimates the likelihood using the model and reduces the influence of false votes. In experiments, we confirmed that the proposed method reduces false positive detection and improves action detection accuracy when using the IXMAS dataset and the UT-Interaction dataset.

Automatic detection of immunoreactive areas in fluorescence microscopic images is becoming a key technique in the field of biology including neuroscience, although it is still challenging because of several reasons such as low signal-to-noise ratio and contrast variation within an image. In this study, we developed a new algorithm that exhaustively detects co-localized areas in multi-channel fluorescence images, where shapes of target objects may differ among channels. Different adaptive binarization thresholds for different local regions in different channels are introduced and the condition of each segment is assessed to recognize the target objects. The proposed method was applied to detect immunoreactive spots that labeled membrane receptors on dendritic spines of mouse cerebellar Purkinje cells. Our method achieved the best detection performance over five pre-existing methods.

This paper describes an equivalent circuit analysis of a meta-surface using a double-layered patch-type frequency-selective surface (FSS); the analysis considers the coupling between FSSs. Two types of double-layered structures are examined. One is a stacked structure and the other is an alternated structure. The results calculated using the equivalent circuit are in agreement with the results of the FDTD analysis. In addition, it is clarified that the stacked and alternated structures exhibit the common mode and the differential mode coupling, respectively. Moreover, experiments support analysis results for both stacked and alternated structures.

Microwave heating is expected to increase the yield of product, to decrease the reaction time, and to discover the new reaction system. The Rolling Circle Amplification (RCA) is an enzymatic synthesis method of deoxyribonucleic acid (DNA) strands with repeated sequence of a circulate template-DNA. In previous study, controlled microwave heating accelerated the maximum 4-fold compared with the conventional condition. Further, we indicated that the selectively heat of some buffer components by microwave irradiation induced the acceleration of RCA. The purpose of this research is to clarify the relationship between the microwave heating and buffer components. The understanding of role of ion-containing buffer components under microwave will be able to control the microwave-assisted enzymatic reaction. We studied the relation between the microwave power loss and RCA components via dielectric measurements, cavity resonator feature measurement, and electromagnetic simulation. Electromagnetic simulation of the TM010 cavity showed that the sample tube was heated only by an electric field. The buffer containing ions of the RCA components was selectively heated via microwave irradiation in the TM010 cavity resonator.

Due to rapid increases in the number of users and diversity of devices, temporal fluctuation of traffic on information communication network is becoming large and rapid recently. Especially, sudden traffic changes such as flash crowds often cause serious congestion on the network and result in nearly fatal slow down of date-communication speed. In order to keep communication quality high on the network, routing protocols that are scalable and able to quickly respond to rapid, and often unexpected, traffic fluctuation are highly desired. One of the promising approaches is the distributed routing protocol, which works without referring global information of the whole network but requires only limited informatin of it to realize route selection. These approaches include biologically inspired routing protocols based on the Adaptive Response by Attractor Selection model (ARAS), in which routing tables are updated along with only a scalar value reflecting communication quality measured on each router without evaluating communication quality over the whole network. However, the lack of global knowledge of the current status of the network often makes it difficult to respond promptly to traffic changes on the network that occurs at outside of the local scope of the protocol and causes inefficient use of network resources. In order to solve the essential problem of the local scope, we extend ARAS and propose a routing protocol with active and stochastic route exploration. The proposed protocol can obtain current communication quality of the network beyond its local scope and promptly responds to traffic changes occur on the network by utilizing the route exploration. In order to compensate destabilization of routing itself due to the active and stochastic exploration, we also introduce a short-term memory to the dynamics of the proposed attractor selection model. We conform by numerical simulations that the proposed protocol successfully balances rapid exploration with reliable routing owning to the memory term.

A black mask (BM) is a layer used to improve the display quality by suppressing light leakage. In general, the BM is formed by a photolithography process. In this study, a novel technique for the fabrication of a quasi-black mask (q-BM) is proposed; the q-BM was composed of vertical and hybrid orientation areas, patterned by a separation coating technique using an electro-spray deposition method. Using our technique, the q-BM can be formed easily without the additional masks used for the BM.

This paper proposes an address power reduction method for plasma display panels (PDPs) using subfield data smoothing based on a visual masking effect. High-resolution, high-frame-rate PDPs have large address power loss caused by parasitic capacitance. Although the address power is reduced by smoothing the subfield data, noise is generated. The proposed method reduces the address power while maintaining the image quality by choosing the smoothing area of the address data based on the visual masking effect. The results of subjective assessment for the images based on smoothed address data indicate that image quality is maintained.

In this paper, we propose and analyze the opportunistic amplify-and-forward (AF) relaying scheme using antenna selection in conjunction with different adaptive transmission techniques over Rayleigh fading channels. In this scheme, the best antenna of a source and the best relay are selected for communication between the source and destination. Closed-form expressions for the outage probability and average symbol error rate (SER) are derived to confirm that increasing the number of antennas is the best option as compared with increasing the number of relays. We also obtain closed-form expressions for the average channel capacity under three different adaptive transmission techniques: 1) optimal power and rate adaptation; 2) constant power with optimal rate adaptation; and 3) channel inversion with a fixed rate. The channel capacity performance of the considered adaptive transmission techniques is evaluated and compared with a different number of relays and various antennas configurations for each adaptive technique. Our derived analytical results are verified through extensive Monte Carlo simulations.