We propose new CAPTCHA image generation systems by using generative adversarial network (GAN) techniques to strengthen against CAPTCHA solvers. To verify whether a user is human, CAPTCHA images are widely used on the web industry today. We introduce two different systems for generating CAPTCHA images, namely, the distance GAN (D-GAN) and composite GAN (C-GAN). The D-GAN adds distance values to the original CAPTCHA images to generate new ones, and the C-GAN generates a CAPTCHA image by composing multiple source images. To evaluate the performance of the proposed schemes, we used the CAPTCHA breaker software as CAPTCHA solver. Then, we compared the resistance of the original source images and the generated CAPTCHA images against the CAPTCHA solver. The results show that the proposed schemes improve the resistance to the CAPTCHA solver by over 67.1% and 89.8% depending on the system.

Three-dimensional (3D) field-programmable gate arrays (FPGAs) are expected to offer higher logic density as well as improved delay and power performance by utilizing 3D integrated circuit technology. However, because through-silicon-vias (TSVs) for conventional 3D FPGA interlayer connections have a large area overhead, there is an inherent tradeoff between connectivity and small size. To find a balance between cost and performance, and to explore 3D FPGAs with realistic 3D integration processes, we propose two types of 3D FPGA and construct design tool sets for architecture exploration. In previous research, we created a TSV-free 3D FPGA with a face-down integration method; however, this was limited to two layers. In this paper, we discuss the face-up stacking of several face-down stacked FPGAs. To minimize the number of TSVs, we placed TSVs peripheral to the FPGAs for 3D-FPGA with 4 layers. According to our results, a 2-layer 3D FPGA has reasonable performance when limiting the design to two layers, but a 4-layer 3D FPGA is a better choice when area is emphasized.

We propose a wideband antenna that has both vertical and horizontal polarization to create access points with enhanced connectivity. The antenna is composed of a rectangular plate and a ground plate, and the rectangular plate is fed sideways from the ground plate. Its -10dB fractional bandwidth is approximately 162%. It is shown that the offset feed of the rectangular plate is important to attain wideband impedance matching and vertical polarized wave. The results of a parametric study to characterize the first- and second-lowest resonant frequencies are presented. Moreover, the behavior of the impedance matching and polarization is interpreted by dividing the current distribution around the feed port on the rectangular plate into the same direction current mode and the opposite direction current mode. The measured results for the return loss and the radiation pattern of a prototype antenna agree well with the simulation results, therefore the wideband property was experimentally confirmed.

Many concurrency control protocols for B-trees use latch-coupling because its execution is efficient on a single machine. Some studies have indicated that latch-coupling may involve a performance bottleneck when using multicore processors in a shared-everything environment, but no studies have considered the possible performance bottleneck caused by sending messages between processing elements (PEs) in shared-nothing environments. We propose two new concurrency control protocols, “LCFB” and “LCFB-link”, which require no latch-coupling in optimistic processes. The LCFB-link also innovates B-link approach within each PE to reduce the cost of modifications in the PE, as a solution to the difficulty of consistency management for the side pointers in a parallel B-tree. The B-link algorithm is well known as a protocol without latch-coupling, but B-link has the difficulty of guaranteeing the consistency of the side pointers in a parallel B-tree. Experimental results in various environments indicated that the system throughput of the proposed protocols was always superior to those of the conventional protocols, particularly in large-scale configurations, and using LCFB-link was effective for higher update ratios. In addition, to mitigate access skew, data should migrate between PEs. We have demonstrated that our protocols always improve the system throughput and are effective as concurrency controls for data migration.

This paper presents a pivot-set generation algorithm for accelerating exact similarity search in a large-scale data set. To deal with the large-scale data set, it is important to efficiently construct a search index offline as well as to perform fast exact similarity search online. Our proposed algorithm efficiently generates competent pivots with two novel techniques: hierarchical data partitioning and fast pivot optimization techniques. To make effective use of a small number of pivots, the former recursively partitions a data set into two subsets with the same size depending on the rank order from each of two assigned pivots, resulting in a complete binary tree. The latter calculates a defined objective function for pivot optimization with a low computational cost by skillfully operating data objects mapped into a pivot space. Since the generated pivots provide the tight lower bounds on distances between a query object and the data objects, an exact similarity search algorithm effectively avoids unnecessary distance calculations. We demonstrate that the search algorithm using the pivots generated by the proposed algorithm reduces distance calculations with an extremely high rate regarding a range query problem for real large-scale image data sets.

In this paper, we present a new three-way split formula for binary polynomial multiplication (PM) with five recursive multiplications. The scheme is based on a recently proposed multievaluation and interpolation approach using field extension. The proposed PM formula achieves the smallest space complexity. Moreover, it has about 40% reduced time complexity compared to best known results. In addition, using developed techniques for PM formulas, we propose a three-way split formula for Toeplitz matrix vector product with five recursive products which has a considerably improved complexity compared to previous known one.

In this study, toward the development of smartphone-based monitoring system for life logging, we collect over 1,400 hours of data by recording including both the outdoor and indoor daily activities of 19 subjects, under practical conditions with a smartphone and a small camera. We then construct a huge human activity database which consists of an environmental sound signal, triaxial acceleration signals and manually annotated activity tags. Using our constructed database, we evaluate the activity recognition performance of deep neural networks (DNNs), which have achieved great performance in various fields, and apply DNN-based adaptation techniques to improve the performance with only a small amount of subject-specific training data. We experimentally demonstrate that; 1) the use of multi-modal signal, including environmental sound and triaxial acceleration signals with a DNN is effective for the improvement of activity recognition performance, 2) the DNN can discriminate specified activities from a mixture of ambiguous activities, and 3) DNN-based adaptation methods are effective even if only a small amount of subject-specific training data is available.

A robust pedestrian detection approach in thermal infrared imageries for an all-day surveillance is proposed. Firstly, the candidate regions which are likely to contain pedestrians are extracted based on a saliency detection method. Then a deep convolutional network with a multi-task loss is constructed to recognize the pedestrians. The experimental results show the superiority of the proposed approach in pedestrian detection.

This paper presents a new GPU-based rasterization algorithm for Boolean operations that handles arbitary closed polygons. We construct an efficient data structure for interoperation of CPU and GPU and propose a fast GPU-based contour extraction method to ensure the performance of our algorithm. We then design a novel traversing strategy to achieve an error-free calculation of intersection point for correct Boolean operations. We finally give a detail evaluation and the results show that our algorithm has a higher performance than exsiting algorithms on processing polygons with large amount of vertices.

Regular expression matching is essential in network and big-data applications; however, it still has a serious performance bottleneck. The state-of-the-art schemes use a multi-pattern exact string-matching algorithm as a filtering module placed before a heavy regular expression engine. We design a new approximate string-matching filter using multiple q-grams; this filter not only achieves better space compactness, but it also has higher throughput than the existing filters.

This paper deals with the scattering of a cylindrical wave by a perfectly conductive periodic surface. This problem is equivalent to finding the Green's function G(x,z|xs,zs), where (x,z) and (xs,zs) are the observation and radiation source positions above the periodic surface, respectively. It is widely known that the Green's function satisfies the reciprocity: G(x,z|xs,zs)=G(xs,zs|x,z), where G(xs,zs|x,z) is named the reversal Green's function in this paper. So far, there is no numerical method to synthesize the Green's function with the reciprocal property in the grating theory. By combining the shadow theory, the reciprocity theorem for scattering factors and the average filter introduced previously, this paper gives a new numerical method to synthesize the Green's function with reciprocal property. The reciprocity means that any properties of the Green's function can be obtained from the reversal Green's function. Taking this fact, this paper obtains several new formulae on the radiation and scattering from the reversal Green's function, such as a spectral representation of the Green's function, an asymptotic expression of the Green's function in the far region, the angular distribution of radiation power, the total power of radiation and the relative error of power balance. These formulae are simple and easy to use. Numerical examples are given for a very rough periodic surface. Several properties of the radiation and scattering are calculated for a transverse magnetic (TM) case and illustrated in figures.

This paper newly proposes a fast computation technique on the method of image Green's function for p-characteristic calculations, when a plane wave with the transverse wavenumber p is incident on a periodic rough surface having perfect conductivity. In the computation of p-characteristics, based on a spectral domain periodicity of the periodic image Green's function, the image integral equation for a given incidence p maintains the same form for other particular incidences except for the excitation term. By means of a quadrature method, such image integral equations lead to matrix equations. Once the first given matrix equation is performed by a solution procedure as calculations of its matrix elements and its inverse matrix, the other matrix equations for other particular incidences no longer need such a solution procedure. Thus, the total CPU time for the computation of p-characteristics is largely reduced in complex shaped surface cases, huge roughness cases or large period cases.

A wideband CMOS common-gate low-noise amplifier (LNA) with high linearity is proposed. The linearity is improved by dual cross-coupled feedback technique. A passive cross-coupled feedback removes the second-order harmonic feedback effect to the input-referred third-order intercept point (IIP3), which is known as one of the limitations for linearity enhancement using feedback. An active cross-coupled feedback, constituted by a voltage combiner and a feedback capacitor is employed to enhance loop gain, and acquire further linearity improvement. An enhanced LC-match input network and forward isolation of active cross-coupled feedback enable the proposed LNA with wideband input matching and flat gain performance. Fabricated in a 0.13 µm RF CMOS process, the LNA achieves a flat voltage gain of 13 dB, an NF of 2.6∼3.8 dB, and an IIP3 of 3.6∼4.9 dBm over a 3 dB bandwidth of 0.1∼1.3 GHz. It consumes only 3.2 mA from a 1.2 V supply and occupies an area of 480×418 um2. In contrast to those of reported wideband LNAs, the proposed LNA has the merit of low power consumption and high linearity.

A method for encoding detection and bit rate classification of AMR-coded speech is proposed. For each texture frame, 184 features consisting of the short-term and long-term temporal statistics of speech parameters are extracted, which can effectively measure the amount of distortion due to AMR. The deep neural network then classifies the bit rate of speech after analyzing the extracted features. It is confirmed that the proposed features provide better performance than the conventional spectral features designed for bit rate classification of coded audio.

This paper analyzes the time domain Reed Solomon Decoder with FPGA implementation. Data throughput and area is carefully evaluated compared with typical frequency domain Reed Solomon Decoder. In this analysis, three hardware architecture to enhance the data throughput, namely, the pipelined architecture, the parallel architecture, and the truncated arrays, is evaluated, too. The evaluation reveals that the number of the consumed resources of RS(255, 239) is about 20% smaller than those of the frequency domain decoder although data throughput is less than 10% of the frequency domain decoder. The number of the consumed resources of the pipelined architecture is 28% smaller than that of the parallel architecture when data throughput is same. It is because the pipeline architecture requires less extra logics than the parallel architecture. To get higher data throughput, the pipelined architecture is better than the parallel architecture from the viewpoint of consumed resources.

One of the problems associated with voice conversion from a nonparallel corpus is how to find the best match or alignment between the source and the target vector sequences without linguistic information. In a previous study, alignment was achieved by minimizing the distance between the source vector and the transformed vector. This method, however, yielded a sequence of feature vectors that were not well matched with the underlying speaker model. In this letter, the vectors were selected from the candidates by maximizing the overall likelihood of the selected vectors with respect to the target model in the HMM context. Both objective and subjective evaluations were carried out using the CMU ARCTIC database to verify the effectiveness of the proposed method.

In this paper, we propose a deep learning based model for classifying speech-acts using a convolutional neural network (CNN). The model uses some bigram features including parts-of-speech (POS) tags and dependency-relation bigrams, which represent syntactic structural information in utterances. Previous classification approaches using CNN have commonly exploited word embeddings using morpheme unigrams. However, the proposed model first extracts two different bigram features that well reflect the syntactic structure of utterances and then represents them as a vector representation using a word embedding technique. As a result, the proposed model using bigram embeddings achieves an accuracy of 89.05%. Furthermore, the accuracy of this model is relatively 2.8% higher than that of competitive models in previous studies.

An improved beamformer, which uses joint estimation of the reconstructed interference-plus-noise (IPN) covariance matrix and array steering vector (ASV), is proposed. It can mitigate the problem of performance degradation in situations where the desired signal exists in the sample covariance matrix and the steering vector pointing has large errors. In the proposed method, the covariance matrix is reconstructed by weighted sum of the exterior products of the interferences' ASV and their individual power to reject the desired signal component, the coefficients of which can be accurately estimated by the compressed sensing (CS) and total least squares (TLS) techniques. Moreover, according to the theorem of sequential vector space projection, the actual ASV is estimated from an intersection of two subspaces by applying the alternating projection algorithm. Simulation results are provided to demonstrate the performance of the proposed beamformer, which is clearly better than the existing robust adaptive beamformers.

This paper considers a design method of a discrete-time adaptive output feedback control system with a feedforward input based on almost strict positive realness (ASPR-ness). The proposed scheme utilizes the property of ASPR of the controlled plant, and the reference signal is used as feedforward input. The parallel feedforward compensator (PFC) which renders an ASPR augmented controlled plant is also investigated. Besides, it is shown that the output of original plant can track reference signal perfectly without any steady state error. The effectiveness of the proposed scheme is confirmed through a pilot-scale temperature control system.

Wireless sensor networks usually deploy sensor nodes with limited energy resources in unattended environments so that people have difficulty in replacing or recharging the depleted devices. In order to balance the energy dissipation and prolong the network lifetime, this paper proposes a routing spanning tree-based clustering algorithm (RSTCA) which uses routing spanning tree to analyze clustering. In this study, the proposed scheme consists of three phases: setup phase, cluster head (CH) selection phase and steady phase. In the setup phase, several clusters are formed by adopting the K-means algorithm to balance network load on the basis of geographic location, which solves the randomness problem in traditional distributed clustering algorithm. Meanwhile, a conditional inter-cluster data traffic routing strategy is created to simplify the networks into subsystems. For the CH selection phase, a novel CH selection method, where CH is selected by a probability based on the residual energy of each node and its estimated next-time energy consumption as a function of distance, is formulated for optimizing the energy dissipation among the nodes in the same cluster. In the steady phase, an effective modification that counters the boundary node problem by adjusting the data traffic routing is designed. Additionally, by the simulation, the construction procedure of routing spanning tree (RST) and the effect of the three phases are presented. Finally, a comparison is made between the RSTCA and the current distributed clustering protocols such as LEACH and LEACH-DT. The results show that RSTCA outperforms other protocols in terms of network lifetime, energy dissipation and coverage ratio.