Recognition of second language (L2) speech is a challenging task even for state-of-the-art automatic speech recognition (ASR) systems, partly because pronunciation by L2 speakers is usually significantly influenced by the mother tongue of the speakers. Considering that the expressions of non-native speakers are usually simpler than those of native ones, and that second language speech usually includes mispronunciation and less fluent pronunciation, we propose a novel method that maximizes unified acoustic and linguistic objective function to derive a phoneme set for second language speech recognition. The authors verify the efficacy of the proposed method using second language speech collected with a translation game type dialogue-based computer assisted language learning (CALL) system. In this paper, the authors examine the performance based on acoustic likelihood, linguistic discrimination ability and integrated objective function for second language speech. Experiments demonstrate the validity of the phoneme set derived by the proposed method.

Narrow swipe sensor based systems have drawn more and more attention in recent years. However, the size of captured image is significantly smaller than that obtained from the traditional area fingerprint sensor. Under this condition the available minutiae number is also limited. Therefore, only employing minutiae with the standard associated feature can hardly achieve high verification accuracy. To solve this problem, we present a novel Hybrid Minutiae Descriptor (HMD) which consists of two modules. The first one: Minutiae Ridge-Valley Orientation Descriptor captures the orientation information around minutia and also the trace points located at associated ridge and valley. The second one: Gabor Binary Code extracts and codes the image patch around minutiae. The proposed HMD enhances the representation capability of minutiae feature, and can be matched very efficiently. Experiments conducted over public databases and the database captured by the narrow swipe sensor show that this innovative method gives rise to significant improvements in reducing FRR (False Reject Rate) and EER (Equal Error Rate).

The solution of the standard 2-norm-based multiple kernel regression problem and the theoretical limit of the considered model space are discussed in this paper. We prove that 1) The solution of the 2-norm-based multiple kernel regressor constructed by a given training data set does not generally attain the theoretical limit of the considered model space in terms of the generalization errors, even if the training data set is noise-free, 2) The solution of the 2-norm-based multiple kernel regressor is identical to the solution of the single kernel regressor under a noise free setting, in which the adopted single kernel is the sum of the same kernels used in the multiple kernel regressor; and it is also true for a noisy setting with the 2-norm-based regularizer. The first result motivates us to develop a novel framework for the multiple kernel regression problems which yields a better solution close to the theoretical limit, and the second result implies that it is enough to use the single kernel regressors with the sum of given multiple kernels instead of the multiple kernel regressors as long as the 2-norm based criterion is used.

We propose a digital image watermarking method satisfying information hiding criteria (IHC) for robustness against JPEG compression, cropping, scaling, and rotation. When a stego-image is cropped, the marking positions of watermarks are unclear. To detect the position in a cropped stego-image, a marker or synchronization code is embedded with the watermarks in a lattice pattern. Attacks by JPEG compression, scaling, and rotation cause errors in extracted watermarks. Against such errors, the same watermarks are repeatedly embedded in several areas. The number of errors in the extracted watermarks can be reduced by using a weighted majority voting (WMV) algorithm. To correct residual errors in output of the WMV algorithm, we use a high-performance error-correcting code: a low-density parity-check (LDPC) code constructed by progressive edge-growth (PEG). In computer simulations using the IHC ver. 4 the proposed method could a bit error rate of 0, the average PSNR was 41.136 dB, and the computational time for synchronization recovery was less than 10 seconds. The proposed method can thus provide high image quality and fast synchronization recovery.

Within information hiding technology, digital watermarking is one of the most important technologies for copyright protection of digital content. Many digital watermarking schemes have been proposed in academia. However, these schemes are not used, because they are not practical; one reason for this is that the evaluation criteria are loosely defined. To make the evaluation more concrete and improve the practicality of digital watermarking, watermarking schemes must use common evaluation criteria. To realize such criteria, we organized the Information Hiding and its Criteria for Evaluation (IHC) Committee to create useful, globally accepted evaluation criteria for information hiding technology. The IHC Committee improves their evaluation criteria every year, and holds a competition for digital watermarking based on state-of-the-art evaluation criteria. In this paper, we describe the activities of the IHC Committee and its evaluation criteria for digital watermarking of still images, videos, and audio.

Since digital ICs are often designed and fabricated by third parties at any phases today, we must eliminate risks that malicious attackers may implement Hardware Trojans (HTs) on them. In particular, they can easily insert HTs during design phase. This paper proposes an HT rank which is a new quantitative analysis criterion against HTs at gate-level netlists. We have carefully analyzed all the gate-level netlists in Trust-HUB benchmark suite and found out several Trojan net features in them. Then we design the three types of Trojan points: feature point, count point, and location point. By assigning these points to every net and summing up them, we have the maximum Trojan point in a gate-level netlist. This point gives our HT rank. The HT rank can be calculated just by net features and we do not perform any logic simulation nor random test. When all the gate-level netlists in Trust-HUB, ISCAS85, ISCAS89 and ITC99 benchmark suites as well as several OpenCores designs, HT-free and HT-inserted AES netlists are ranked by our HT rank, we can completely distinguish HT-inserted ones (which HT rank is ten or more) from HT-free ones (which HT rank is nine or less). The HT rank is the world-first quantitative criterion which distinguishes HT-inserted netlists from HT-free ones in all the gate-level netlists in Trust-HUB, ISCAS85, ISCAS89, and ITC99.

Kernel discriminant analysis (KDA) is the mainstream approach of nonlinear discriminant analysis (NDA). Since it uses the kernel trick, KDA does not consider its nonlinear discriminant mapping explicitly. In this paper, another NDA approach where the nonlinear discriminant mapping is analytically given is developed. This study is based on the theory of optimal nonlinear discriminant analysis (ONDA) of which the nonlinear mapping is exactly expressed by using the Bayesian posterior probability. This theory indicates that various NDA can be derived by estimating the Bayesian posterior probability in ONDA with various estimation methods. Also, ONDA brings an insight about novel kernel functions, called discriminant kernel (DK), which is defined by also using the posterior probabilities. In this paper, several NDA and DK derived from ONDA with several posterior probability estimators are developed and evaluated. Given fine estimation methods of the Bayesian posterior probability, they give good discriminant spaces for visualization or classification.

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.

The copy-move or region duplication forgery technique is a very common type of image manipulation, where a region of the image is copied and then pasted in the same image in order to hide some details. In this paper, a keypoint-based method for copy-move forgery detection is proposed. Firstly, the feature points are detected from the image by using the Förstner Operator. Secondly, the algorithm extracts the features by using MROGH feature descriptor, and then matching the features. Finally, the affine transformation parameters can be estimated using the RANSAC algorithm. Experimental results are presented to confirm that the proposed method is effective to locate the altered region with geometric transformation (rotation and scaling).

In this paper, we propose improved methods of liquid-phase detection of biological targets utilizing magnetic markers and a high-critical-temperature superconducting quantum interference device (SQUID). For liquid-phase detection, the bound and unbound (free) markers are magnetically distinguished by using Brownian relaxation of free markers. Although a signal from the free markers is zero in an ideal case, it exists in a real sample on account of the aggregation and precipitation of free markers. This signal is called a blank signal, and it degrades the sensitivity of target detection. To solve this problem, we propose improved detection methods. First, we introduce a reaction field, Bre, during the binding reaction between the markers and targets. We additionally introduce a dispersion process after magnetization of the bound markers. Using these methods, we can obtain a strong signal from the bound markers without increasing the aggregation of the free markers. Next, we introduce a field-reversal method in the measurement procedure to differentiate the signal from the markers in suspension from that of the precipitated markers. Using this procedure, we can eliminate the signal from the precipitated markers. Then, we detect biotin molecules by using these methods. In an experiment, the biotins were immobilized on the surfaces of large polymer beads with diameters of 3.3 µm. They were detected with streptavidin-conjugated magnetic markers. The minimum detectable molecular number concentration was 1.8×10-19 mol/ml, which indicates the high sensitivity of the proposed method.

Recent studies have obtained superior performance in image recognition tasks by using, as an image representation, the fully connected layer activations of Convolutional Neural Networks (CNN) trained with various kinds of images. However, the CNN representation is not very suitable for fine-grained image recognition tasks involving food image recognition. For improving performance of the CNN representation in food image recognition, we propose a novel image representation that is comprised of the covariances of convolutional layer feature maps. In the experiment on the ETHZ Food-101 dataset, our method achieved 58.65% averaged accuracy, which outperforms the previous methods such as the Bag-of-Visual-Words Histogram, the Improved Fisher Vector, and CNN-SVM.

Connection Service Providers (CSP) are wishing to increase their Return on Investment (ROI) by utilizing the data assets generated by tracking subscriber behaviors. This results in the ability to apply personalized policies, monitor and control the service traffic to subscribers and gain more revenue through the usage of subscriber data with ad networks. In this paper, a system is proposed to monitor and analyze the Internet access of the subscribers of a regional SP in order to classify the subscribers into interest categories from the Interactive Advertising Bureau (IAB) categories. The study employs the categorization engine to build category vectors for all individuals using Internet services through the subscription. The proposal makes it easy to detect changes in the interests of individuals/subscribers over time.

Due to the periodic recovery of virtual machines regardless of whether malicious intrusions exist, proactive recovery-based Intrusion Tolerant Systems (ITSs) are being considered for mission-critical applications. However, the virtual replicas can easily be exposed to attacks during their working period, and additionally, proactive recovery-based ITSs are ineffective in eliminating the vulnerability of exposure time, which is closely related to service availability. To address these problems, we propose a novel hybrid recovery-based ITS in this paper. The proposed method utilizes availability-driven recovery and dynamic cluster resizing. The availability-driven recovery method operates the recovery process by both proactive and reactive ways for the system to gain shorter exposure times and higher success rates. The dynamic cluster resizing method reduces the overhead of the system that occurs from dynamic workload fluctuations. The performance of the proposed ITS with various synthetic and real workloads using CloudSim showed that it guarantees higher availability and reliability of the system, even under malicious intrusions such as DDoS attacks.

Time performance optimization and resource conflict resolution are two important challenges in multiple project management contexts. Compared with traditional project management, multi-project management usually suffers limited and insufficient resources, and a tight and urgent deadline to finish all concurrent projects. In this case, time performance optimization of the global project management is badly needed. To our best knowledge, existing work seldom pays attention to the formal modeling and analyzing of multi-project management in an effort to eliminate resource conflicts and optimizing the project execution time. This work proposes such a method based on PRT-Net, which is a Petri net-based formulism tailored for a kind of project constrained by resource and time. The detailed modeling approaches based on PRT-Net are first presented. Then, resource conflict detection method with corresponding algorithm is proposed. Next, the priority criteria including a key-activity priority strategy and a waiting-short priority strategy are presented to resolve resource conflicts. Finally, we show how to construct a conflict-free PRT-Net by designing resource conflict resolution controllers. By experiments, we prove that our proposed priority strategy can ensure the execution time of global multiple projects much shorter than those without using any strategies.

Nonnegative matrix factorization (NMF) is an unsupervised technique to represent nonnegative data as linear combinations of nonnegative bases, which has shown impressive performance for source separation. However, its source separation performance degrades when one signal can also be described well with the bases for the interfering source signals. In this paper, we propose a discriminative NMF (DNMF) algorithm which exploits the reconstruction error for the interfering signals as well as the target signal based on target bases. The objective function for training the bases is constructed so as to yield high reconstruction error for the interfering source signals while guaranteeing low reconstruction error for the target source signals. Experiments show that the proposed method outperformed the standard NMF and another DNMF method in terms of both the perceptual evaluation of speech quality score and signal-to-distortion ratio in various noisy environments.

Middle-level parts have attracted great attention in the computer vision community, acting as discriminative elements for objects. In this paper we propose an unsupervised approach to mine discriminative parts for object detection. This work features three aspects. First, we introduce an unsupervised, exemplar-based training process for part detection. We generate initial parts by selective search and then train part detectors by exemplar SVM. Second, a part selection model based on consistency and distinctiveness is constructed to select effective parts from the candidate pool. Third, we combine discriminative part mining with the deformable part model (DPM) for object detection. The proposed method is evaluated on the PASCAL VOC2007 and VOC2010 datasets. The experimental results demons-trate the effectiveness of our method for object detection.

Collective activity recognition plays an important role in high-level video analysis. Most current feature representations look at contextual information extracted from the behaviour of nearby people. Every person needs to be detected and his pose should be estimated. After extracting the feature, hierarchical graphical models are always employed to model the spatio-temporal patterns of individuals and their interactions, and so can not avoid complex preprocessing and inference operations. To overcome these drawbacks, we present a new feature representation method, called attribute-based spatio-temporal (AST) descriptor. First, two types of information, spatio-temporal (ST) features and attribute features, are exploited. Attribute-based features are manually specified. An attribute classifier is trained to model the relationship between the ST features and attribute-based features, according to which the attribute features are refreshed. Then, the ST features, attribute features and the relationship between the attributes are combined to form the AST descriptor. An objective classifier can be specified on the AST descriptor and the weight parameters of the classifier are used for recognition. Experiments on standard collective activity benchmark sets show the effectiveness of the proposed descriptor.

With recent developments in machine learning technology, the predictions by systems incorporating machine learning can now have a significant impact on the lives and activities of individuals. In some cases, predictions made by machine learning can result unexpectedly in unfair treatments to individuals. For example, if the results are highly dependent on personal attributes, such as gender or ethnicity, hiring decisions might be discriminatory. This paper investigates the neutralization of a probabilistic model with respect to another probabilistic model, referred to as a viewpoint. We present a novel definition of neutrality for probabilistic models, η-neutrality, and introduce a systematic method that uses the maximum likelihood estimation to enforce the neutrality of a prediction model. Our method can be applied to various machine learning algorithms, as demonstrated by η-neutral logistic regression and η-neutral linear regression.

The paper addresses a scheme of lightly supervised training of an acoustic model, which exploits a large amount of data with closed caption texts but not faithful transcripts. In the proposed scheme, a sequence of the closed caption text and that of the ASR hypothesis by the baseline system are aligned. Then, a set of dedicated classifiers is designed and trained to select the correct one among them or reject both. It is demonstrated that the classifiers can effectively filter the usable data for acoustic model training. The scheme realizes automatic training of the acoustic model with an increased amount of data. A significant improvement in the ASR accuracy is achieved from the baseline system and also in comparison with the conventional method of lightly supervised training based on simple matching.

We demonstrated that load balancing using actual subscriber extension numbers was practical and effective against traffic congestion after a disaster based on actual data. We investigated the ratios of the same subscriber extension numbers in each prefecture and found that most of them were located almost evenly all over the country without being concentrated in a particular area. The ratio of every number except for the fourth-last digit in the last group of four numbers in a telephone number was used almost equally and located almost evenly all over the country. Tolerance against overload in the last, second-, and third-last single digits stays close to that in the ideal situation if we assume that each session initiation protocol server has a capacity in accordance with the ratio of each number on every single digit in the last group of four numbers in Japan. Although tolerance against overload in double-, triple-, and quadruple-digit numbers does not stay close to that in the ideal situation, it still remains sufficiently high in the case of double- and triple-digit numbers. Although tolerance against overload in the quadruple-digit numbers becomes low, disaster congestion is still not likely to occur in almost half of the area of Japan (23 out of 47 prefectures).