Crack detection is a vital task to maintain a bridge's health and safety condition. Traditional computer-vision based methods easily suffer from disturbance of noise and clutters for a real bridge inspection. To address this limitation, we propose a two-stage crack detection approach based on Convolutional Neural Networks (CNN) in this letter. A predictor of small receptive field is exploited in the first detection stage, while another predictor of large receptive field is used to refine the detection results in the second stage. Benefiting from data fusion of confidence maps produced by both predictors, our method can predict the probability belongs to cracked areas of each pixel accurately. Experimental results show that the proposed method is superior to an up-to-date method on real concrete surface images.

This letter proposes a novel speech enhancement system based on the ‘L’ shaped triple-microphone. The modified coherence-based algorithm and the first-order differential beamforming are combined to filter the spatial distributed noise. The experimental results reveal that the proposed algorithm achieves significant performance in spatial filtering under different noise scenarios.

For voice-enabled car navigation systems that use a multi-purpose cloud speech recognition service (cloud ASR), utterance classification that is robust against speech recognition errors is needed to realize a user-friendly voice interface. The purpose of this study is to improve the accuracy of utterance classification for voice-enabled car navigation systems when inputs to a classifier are error-prone speech recognition results obtained from a cloud ASR. The role of utterance classification is to predict which car navigation function a user wants to execute from a spontaneous utterance. A cloud ASR causes speech recognition errors due to the noises that occur when traveling in a car, and the errors degrade the accuracy of utterance classification. There are many methods for reducing the number of speech recognition errors by modifying the inside of a speech recognizer. However, application developers cannot apply these methods to cloud ASRs because they cannot customize the ASRs. In this paper, we propose a system for improving the accuracy of utterance classification by modifying both speech-signal inputs to a cloud ASR and recognized-sentence outputs from an ASR. First, our system performs speech enhancement on a user's utterance and then sends both enhanced and non-enhanced speech signals to a cloud ASR. Speech recognition results from both speech signals are merged to reduce the number of recognition errors. Second, to reduce that of utterance classification errors, we propose a data augmentation method, which we call “optimal doping,” where not only accurate transcriptions but also error-prone recognized sentences are added to training data. An evaluation with real user utterances spoken to car navigation products showed that our system reduces the number of utterance classification errors by 54% from a baseline condition. Finally, we propose a semi-automatic upgrading approach for classifiers to benefit from the improved performance of cloud ASRs.

Multiplication is a key fundamental function for many error-tolerant applications. Approximate multiplication is considered to be an efficient technique for trading off energy against performance and accuracy. This paper proposes an accuracy-controllable multiplier whose final product is generated by a carry-maskable adder. The proposed scheme can dynamically select the length of the carry propagation to satisfy the accuracy requirements flexibly. The partial product tree of the multiplier is approximated by the proposed tree compressor. An 8×8 multiplier design is implemented by employing the carry-maskable adder and the compressor. Compared with a conventional Wallace tree multiplier, the proposed multiplier reduced power consumption by between 47.3% and 56.2% and critical path delay by between 29.9% and 60.5%, depending on the required accuracy. Its silicon area was also 44.6% smaller. In addition, results from two image processing applications demonstrate that the quality of the processed images can be controlled by the proposed multiplier design.

In this paper, the analog code modulation characteristics of distributed-based transversal filters (DTFs) suitable for use in spectrally encoded CDMA systems are presented. The DTF is verified as an appropriate method to use in high-speed CDMA systems as opposed to previously proposed methods, which are intended for Direct Sequence (DS) CDMA systems. The large degree of freedom of DTF design permits controlling the filter pulse response to generate well specified temporal phase-coded signals. A decoder structure that performs bipolar detection of user subbands giving rise to a Spectral-Amplitude Encoded CDMA system is considered. Practical implementations require truncating the spreading signals by a time window of duration equal to the span time of the tapped delay line. Filter functions are chosen to demodulate the matched channel and achieve improved user interference rejection avoiding the need for transversal filters featuring a large number of taps. As a proof-of-concept of the electronic SAE scheme, practical circuit designs are developed at low speeds (3-dB point at 1 GHz) demonstrating the viability of the proposal.

In the successive refinement problem, a fixed-length sequence emitted from an information source is encoded into two codewords by two encoders in order to give two reconstructions of the sequence. One of two reconstructions is obtained by one of two codewords, and the other reconstruction is obtained by all two codewords. For this coding problem, we give non-asymptotic inner and outer bounds on pairs of numbers of codewords of two encoders such that each probability that a distortion exceeds a given distortion level is less than a given probability level. We also give a general formula for the rate-distortion region for general sources, where the rate-distortion region is the set of rate pairs of two encoders such that each maximum value of possible distortions is less than a given distortion level.

Compressed Sensing (CS) is known to provide better channel estimation performance than the Least Square (LS) method for channel estimation. However, multipath delays may not be resolved if they span between the grids. This grid problem of CS is an obstacle to super resolution channel estimation. An Atomic Norm (AN) minimization is one of the methods for estimating continuous parameters. The AN minimization can successfully recover a spectrally sparse signal from a few time-domain samples even though the dictionary is continuous. There are studies showing that the AN minimization method has better resolution than conventional CS methods. In this paper, we propose a channel estimation method based on the AN minimization for Spread Spectrum (SS) systems. The accuracy of the proposed channel estimation is compared with the conventional LS method and Dantzig Selector (DS) of the CS. In addition to the application of channel estimation in wireless communication, we also show that the AN minimization can be applied to Global Positioning System (GPS) using Gold sequence.

In 1973, Arimoto proved the strong converse theorem for the discrete memoryless channels stating that when transmission rate R is above channel capacity C, the error probability of decoding goes to one as the block length n of code word tends to infinity. He proved the theorem by deriving the exponent function of error probability of correct decoding that is positive if and only if R > C. Subsequently, in 1979, Dueck and Körner determined the optimal exponent of correct decoding. Recently the author determined the optimal exponent on the correct probability of decoding have the form similar to that of Dueck and Körner determined. In this paper we give a rigorous proof of the equivalence of the above exponet function of Dueck and Körner to a exponent function which can be regarded as an extention of Arimoto's bound to the case with the cost constraint on the channel input.

In this paper, we propose a design of reversible adder/subtractor blocks and arithmetic logic units (ALUs). The main concept of our approach is different from that of the existing related studies; we emphasize the function design. Our approach of investigating the reversible functions includes (a) the embedding of irreversible functions into incompletely-specified reversible functions, (b) the operation assignment, and (c) the permutation of function outputs. We give some extensions of these techniques for further improvements in the design of reversible functions. The resulting reversible circuits are smaller than that of the existing design in terms of the number of multiple-control Toffoli gates. To evaluate the quantum cost of the obtained circuits, we convert the circuits to reduced quantum circuits for experiments. The results also show the superiority of our realization of adder/subtractor blocks and ALUs in quantum cost.

Spectrum sensing is the first task performed by cognitive radio (CR) networks. In this paper we propose a spectrum sensing algorithm for orthogonal frequency division multiplex (OFDM) signal based on deep learning and covariance matrix graph. The advantage of deep learning in image processing is applied to the spectrum sensing of OFDM signals. We start by building the spectrum sensing model of OFDM signal, and then analyze structural characteristics of covariance matrix (CM). Once CM has been normalized and transformed into a gray level representation, the gray scale map of covariance matrix (GSM-CM) is established. Then, the convolutional neural network (CNN) is designed based on the LeNet-5 network, which is used to learn the training data to obtain more abstract features hierarchically. Finally, the test data is input into the trained spectrum sensing network model, based on which spectrum sensing of OFDM signals is completed. Simulation results show that this method can complete the spectrum sensing task by taking advantage of the GSM-CM model, which has better spectrum sensing performance for OFDM signals under low SNR than existing methods.

Data integrity is a key metric of security for Internet of Things (IoT) which refers to accuracy and reliability of data during transmission, storage and retrieval. Cryptographic hash functions are common means used for data integrity verification. Newly announced SHA-3 is the next generation hash function standard to replace existing SHA-1 and SHA-2 standards for better security. However, its underlying Keccak algorithm is computation intensive and thus limits its deployment on IoT systems which are normally equipped with 32-bit resource constrained embedded processors. This paper proposes two efficient SHA-3 ASIPs based on an open 32-bit RISC-V embedded processor named Z-scale. The first operation-oriented ASIP (OASIP) focuses on accelerating time-consuming operations with instruction set extensions to improve resource efficiency. And next datapath-oriented ASIP (DASIP) targets exploiting advance data and instruction level parallelism with extended auxiliary registers and customized datapath to achieve high performance. Implementation results show that both proposed ASIPs can effectively accelerate SHA-3 algorithm with 14.6% and 26.9% code size reductions, 30% and 87% resource efficiency improvements, 71% and 262% better maximum throughputs as well as 40% and 288% better power efficiencies than reference design. This work makes SHA-3 algorithm integration practical for both low-cost and high-performance IoT systems.

This paper proposes gain relaxation in signal enhancement designed for speech recognition. Gain relaxation selectively applies softer enhancement of a target signal to eliminate potential degradation in speech recognition caused by small undesirable distortion in the target signal components. The softer enhancement is a solution to overlooked performance degradation in signal enhancement combined with speech recognition which is encountered in commercial products with an unaware small local noise source. Evaluation of directional interference suppression with signals recorded by a commercial PC (personal computer) demonstrates that signal enhancement over the input is achieved without sacrificing the performance for clean speech.

In this paper, we propose a novel primary user detection scheme for spectrum sensing in cognitive radio. Inspired by the conventional signal classification approach, the spectrum sensing is translated into a classification problem. On the basis of feature-based classification, the spectral correlation of a second-order cyclostationary analysis is applied as the feature extraction method, whereas a stacked denoising autoencoders network is applied as the classifier. Two training methods for signal detection, interception-based detection and simulation-based detection, are considered, for different prior information and implementation conditions. In an interception-based detection method, inspired by the two-step sensing, we obtain training data from the interception of actual signals after a sophisticated sensing procedure, to achieve detection without priori information. In addition, benefiting from practical training data, this interception-based detection is superior under actual transmission environment conditions. The alternative, a simulation-based detection method utilizes some undisguised parameters of the primary user in the spectrum of interest. Owing to the diversified predetermined training data, simulation-based detection exhibits transcendental robustness against harsh noise environments, although it demands a more complicated classifier network structure. Additionally, for the above-described training methods, we discuss the classifier complexity over implementation conditions and the trade-off between robustness and detection performance. The simulation results show the advantages of the proposed method over conventional spectrum-sensing schemes.

Recently, a control technique of light distribution pattern has become important to improve the functionality and the light utilization efficiency of electronic displays, illumination devices and so on. As a light control technique, polymer-dispersed liquid crystals (PDLCs) have been commonly used so far. However, a precise control of the light diffusion distribution of conventional PDLC has been difficult due to the random polymer network structure, which results in the low light utilization efficiency. On the other hand, reverse-mode PDLCs with homogeneously aligned molecules can anisotropically diffuse light. The reverse-mode PDLC, however, has polarization dependency in the haze value due to homogeneously aligned molecules, which also results in the low light utilization efficiency. Therefore, it is necessary to establish the optimization method of light diffusion distribution without the molecules alignment treatment, and we have proposed a novel PDLC with structure-controlled polymer network which was fabricated by the irradiation with uni-directionally diffused UV light. In this paper, we investigated the effect of the process temperature during UV irradiation on the internal structure and light diffusion distribution of the proposed PDLC. As a result, in case that the mixture during UV irradiation was in isotropic phase, we clarified that the structure-controlled PDLCs with alternating striped LCs/polymer pattern could be obtained because the mixture was sufficiently irradiated with uni-directionally diffused UV light. For the high haze, this structure-controlled PDLC should be fabricated as low temperature as possible with maintaining the mixture in isotropic phase so that the mixture was not a nano-scaled molecular mixing state. Also, this PDLC had no polarization dependency in the haze value and could electrically switch the light distribution pattern between anisotropic light diffusion and light transmission. From the above results, we concluded that the proposed PDLC could precisely control the light diffusion distribution, and realize the high light utilization efficiency.

We have developed the high speed bandpass liquid crystal filter with narrow full width at half maximum (FWHM) of 5nm for real-time multi spectral imaging systems. We have successfully achieved short wavelength-switching time of 30ms by the optimization of phase retardation of thin liquid crystal cells.

A new hybrid brain-computer interface (BCI), which is based on sequential controls by eye tracking and steady-state visual evoked potentials (SSVEPs), has been proposed for high-speed spelling in virtual reality (VR) with a 40-target virtual keyboard. During target selection, gaze point was first detected by an eye-tracking accessory. A 4-target block was then selected for further target selection by a 4-class SSVEP BCI. The system can type at a speed of 1.25 character/sec in a cue-guided target selection task. Online experiments on three subjects achieved an averaged information transfer rate (ITR) of 360.7 bits/min.

A wind-noise suppressor with SNR based wind-noise detection and speech-wind discrimination is proposed. Wind-noise detection is performed in each frame and frequency based on the power ratio of the noisy speech and an estimated stationary noise. The detection result is modified by speech presence likelihood representing spectral smoothness to eliminate speech components. To suppress wind noise with little speech distortion, spectral gains are made smaller in the frame and the frequency where wind-noise is detected. Subjective evaluation results show that the 5-grade MOS for the proposed wind-noise suppressor reaches 3.4 and is 0.56 higher than that by a conventional noise suppressor with a statistically significant difference.

Spectrum databases are required to assist the process of radio propagation estimation for spectrum sharing. Especially, a measurement-based spectrum database achieves highly efficient spectrum sharing by storing the observed radio environment information such as the signal power transmitted from a primary user. However, when the average received signal power is calculated in a given square mesh, the bias of the observation locations within the mesh strongly degrades the accuracy of the statistics because of the influence of terrain and buildings. This paper proposes a method for determining the statistics by using mesh clustering. The proposed method clusters the feature vectors of the measured data by using the k-means and Gaussian mixture model methods. Simulation results show that the proposed method can decrease the error between the measured value and the statistically processed value even if only a small amount of data is available in the spectrum database.

Electrical contacts are separated at constant speed and break arcs are generated in nitrogen or air in a 200V-450VDC/10A resistive circuit. The break arcs are extinguished by magnetic blow-out. Arc duration for the silver and copper contact pairs is investigated for each supply voltage. Following results are shown. The arc duration for Cu contacts in nitrogen is the shortest. For Cu contacts, the arc dwell time in air was considerably longer than that of nitrogen. For Ag contacts, the arc duration in nitrogen was almost the same as that in air.

We are developing an Ultra-Low Field (ULF) Magnetic Resonance Imaging (MRI) system with a tuned high-Tc (HTS)-rf-SQUID for food inspection. We previously reported that a small hole in a piece of cucumber can be detected. The acquired image was based on filtered back-projection reconstruction using a polarizing permanent magnet. However the resolution of the image was insufficient for food inspection and took a long time to process. The purpose of this study is to improve image quality and shorten processing time. We constructed a specially designed cryostat, which consists of a liquid nitrogen tank for cooling an electromagnetic polarizing coil (135mT) at 77K and a room temperature bore. A Cu pickup coil was installed at the room temperature bore and detected an NMR signal from a sample. The signal was then transferred to an HTS SQUID via an input coil. Following a proper MRI sequence, spatial frequency data at 64×32 points in k-space were obtained. Then, a 2D-FFT (Fast Fourier Transformation) method was applied to reconstruct the 2D-MR images. As a result, we successfully obtained a clear water image of the characters “TUT”, which contains a narrowest width of 0.5mm. The imaging time was also shortened by a factor of 10 when compared to the previous system.