Bloom Filter is a bit array (a one-dimensional storage structure) that provides a compact representation for a set of data, which can be used to answer the membership query in an efficient manner with a small number of false positives. It has a lot of applications in many areas. In this paper, we extend the design of Bloom Filter by using a multi-dimensional matrix to replace the one-dimensional structure with three different implementations, namely OFFF, WOFF, FFF. We refer the extended Bloom Filter as Feng Filter. We show the false positive rates of our method. We compare the false positive rate of OFFF with that of the traditional one-dimensional Bloom Filter and show that under certain condition, OFFF has a lower false positive rate. Traditional Bloom Filter can be regarded as a special case of our Feng Filter.

A single-dimensional interface which enables users to obtain diverse localizations of audio sources is proposed. In many conventional interfaces for arranging audio sources, there are multiple arrangement parameters, some of which allow users to control positions of audio sources. However, it is difficult for users who are unfamiliar with these systems to optimize the arrangement parameters since the number of possible settings is huge. We propose a simple, single-dimensional interface for adjusting arrangement parameters, allowing users to sample several diverse audio source arrangements and easily find their preferred auditory localizations. To select subsets of arrangement parameters from all of the possible choices, auditory-localization space vectors (ASVs) are defined to represent the auditory localization of each arrangement parameter. By selecting subsets of ASVs which are approximately orthogonal, we can choose arrangement parameters which will produce diverse auditory localizations. Experimental evaluations were conducted using music composed of three audio sources. Subjective evaluations confirmed that novice users can obtain diverse localizations using the proposed interface.

Multi-label classification is an appealing and challenging supervised learning problem, where multiple labels, rather than a single label, are associated with an unseen test instance. To remove possible noises in labels and features of high-dimensionality, multi-label dimension reduction has attracted more and more attentions in recent years. The existing methods usually suffer from several problems, such as ignoring label outliers and label correlations. In addition, most of them emphasize on conducting dimension reduction in an unsupervised or supervised way, therefore, unable to utilize the label information or a large amount of unlabeled data to improve the performance. In order to cope with these problems, we propose a novel method termed Robust sEmi-supervised multi-lAbel DimEnsion Reduction, shortly READER. From the viewpoint of empirical risk minimization, READER selects most discriminative features for all the labels in a semi-supervised way. Specifically, the ℓ2,1-norm induced loss function and regularization term make READER robust to the outliers in the data points. READER finds a feature subspace so as to keep originally neighbor instances close and embeds labels into a low-dimensional latent space nonlinearly. To optimize the objective function, an efficient algorithm is developed with convergence property. Extensive empirical studies on real-world datasets demonstrate the superior performance of the proposed method.

We propose a non-blind deconvolution algorithm of point cloud attributes inspired by multi-Wiener SURE-LET deconvolution for images. The image reconstructed by the SURE-LET approach is expressed as a linear combination of multiple filtered images where the filters are defined on the frequency domain. The coefficients of the linear combination are calculated so that the estimate of mean squared error between the original and restored images is minimized. Although the approach is very effective, it is only applicable to images. Recently we have to handle signals on irregular grids, e.g., texture data on 3D models, which are often blurred due to diffusion or motions of objects. However, we cannot utilize image processing-based approaches straightforwardly since these high-dimensional signals cannot be transformed into their frequency domain. To overcome the problem, we use graph signal processing (GSP) for deblurring the complex-structured data. That is, the SURE-LET approach is redefined on GSP, where the Wiener-like filtering is followed by the subband decomposition with an analysis graph filter bank, and then thresholding for each subband is performed. In the experiments, the proposed method is applied to blurred textures on 3D models and synthetic sparse data. The experimental results show clearly deblurred signals with SNR improvements.

Recent research on the assessment of low probability of interception (LPI) radar waveforms is mainly based on limiting spectral properties of Wigner matrices. As the dimension of actual operating data is constrained by the sampling frequency, it is very urgent and necessary to research the finite theory of Wigner matrices. This paper derives a closed-form expression of the spectral cumulative distribution function (CDF) for Wigner matrices of finite sizes. The expression does not involve any derivatives and integrals, and therefore can be easily computed. Then we apply it to quantifying the LPI performance of radar waveforms, and the Kullback-Leibler divergence (KLD) is also used in the process of quantification. Simulation results show that the proposed LPI metric which considers the finite sample size and signal-to-noise ratio is more effective and practical.

A bi-dimensional empirical mode decomposition (BEMD) is one of the powerful methods for decomposing non-linear and non-stationary signals without a prior function. It can be applied in many applications such as feature extraction, image compression, and image filtering. Although modified BEMDs are proposed in several approaches, computational cost and quality of their bi-dimensional intrinsic mode function (BIMF) still require an improvement. In this paper, an iteration-free computation method for bi-dimensional empirical mode decomposition, called iBEMD, is proposed. The locally partial correlation for principal component analysis (LPC-PCA) is a novel technique to extract BIMFs from an original signal without using extrema detection. This dramatically reduces the computation time. The LPC-PCA technique also enhances the quality of BIMFs by reducing artifacts. The experimental results, when compared with state-of-the-art methods, show that the proposed iBEMD method can achieve the faster computation of BIMF extraction and the higher quality of BIMF image. Furthermore, the iBEMD method can clearly remove an illumination component of nature scene images under illumination change, thereby improving the performance of text localization and recognition.

A novel dimensionality reduction method, Fisher Discriminant Locality Preserving Canonical Correlation Analysis (FDLP-CCA), for visualizing Web images is presented in this paper. FDLP-CCA can integrate two modalities and discriminate target items in terms of their semantics by considering unique characteristics of the two modalities. In this paper, we focus on Web images with text uploaded on Social Networking Services for these two modalities. Specifically, text features have high discriminate power in terms of semantics. On the other hand, visual features of images give their perceptual relationships. In order to consider both of the above unique characteristics of these two modalities, FDLP-CCA estimates the correlation between the text and visual features with consideration of the cluster structure based on the text features and the local structures based on the visual features. Thus, FDLP-CCA can integrate the different modalities and provide separated manifolds to organize enhanced compactness within each natural cluster.

This paper presents a capacitor-loaded 4x4 planar loop array for three-dimensional near-field beamforming of magnetic resonance wireless power transfer (WPT). This planar loop array provides three important functions: beamforming, selective power transfer, and the ability to work alignment free with the receiver. These functions are realized by adjusting the capacitance of each loop. The optimal capacitance of each loop that corresponds to the three functions can be found using a genetic algorithm (GA); the three functions were verified by comparing simulations and measurements at a frequency of 6.78MHz. Finally, the beamforming mechanism of a near-field loop array was investigated using the relationship between the current magnitude and the resonance frequency of each loop, resulting in the findings that the magnitude and the resonance frequency are correlated. This focused current of the specified loop creates a strong magnetic field in front of that loop, resulting in near-field beamforming.

Quaternionic neural networks are extensions of neural networks using quaternion algebra. 3-D and 4-D quaternionic MLPs have been studied. 3-D quaternionic neural networks are useful for handling 3-D objects, such as Euclidean transformation. As for Hopfield neural networks, only 4-D quaternionic Hopfield neural networks (QHNNs) have been studied. In this work, we propose the 3-D QHNNs. Moreover, we define the energy, and prove that it converges.

Second-order sampling of 2-D frequency distributions is examined in this paper. When a figure in the frequency space can fill up the entire frequency space by tiling, we call this figure a tiling cluster. We also introduce the concept of pair regions. The results obtained for the second-order sampling of 1-D and 2-D frequency distributions are arranged using these two concepts. The sampling functions and sampling positions of second-order sampling of a 2-D rectangular-complement highpass frequency distribution, which have not been solved until now, are explicitly presented by using these two concepts.

A new class of visible light communication (VLC) systems, namely image sensor (IS) based VLC systems, has emerged. An IS consists of a two-dimensional (2D) array of photodetectors (PDs), and then VLC systems with an IS receiver are capable of exploiting the spatial dimensions invoked for transmitting information. This paper aims for providing a brief survey of topics related to the IS-based VLC, and then provides a matrix representation of how to map a series of one dimensional (1D) symbols onto a set of 2D symbols for efficiently exploit the associate grade of freedom offered by 2D VLC systems. As an example, the matrix representation is applied to the symbol mapping of layered space-time coding (L-STC), which is presented to enlarge the coverage of IS-based VLC that is limited by pixel resolution of ISs.

As a hot topic of speech signal processing, speech emotion recognition methods have been developed rapidly in recent years. Some satisfactory results have been achieved. However, it should be noted that most of these methods are trained and evaluated on the same corpus. In reality, the training data and testing data are often collected from different corpora, and the feature distributions of different datasets often follow different distributions. These discrepancies will greatly affect the recognition performance. To tackle this problem, a novel corpus-invariant discriminant feature representation algorithm, called transfer discriminant analysis (TDA), is presented for speech emotion recognition. The basic idea of TDA is to integrate the kernel LDA algorithm and the similarity measurement of distributions into one objective function. Experimental results under the cross-corpus conditions show that our proposed method can significantly improve the recognition rates.

Quick Response (QR) code is a two dimensional barcode widely used in many applications. A standard QR code consists of black and white square modules, and it appears randomized patterns. By modifying the modules using certain rule, it is possible to display a logo image on the QR code. Such a QR code is called an aesthetic QR code. In this paper, we change the encoding method of the Reed-Solomon (RS) code to produce an aesthetic QR code without sacrificing its error correcting capability. The proposed method randomly produces candidates of RS blocks and finds the best one during encoding. Considering an image to be displayed, we also introduce a weighting function during random selection that classifies the visually important regions in the image. We further investigate the shape of modules which represents the image and consider the trade-off between the visual quality and its readability. As a result, we can produce a beautiful aesthetic QR code, which still can be decoded by standard QR code reader.

It is expected that a large number of different objects, such as sensor devices and consumer electronics, will be connected to future networks. For such networks, we propose a name resolution method for directly specifying a condition on a set of attribute-value pairs of real-world information without needing prior knowledge of the uniquely assigned name of a target object, e.g., a URL. For name resolution, we need an algorithm to find the target object(s) satisfying a query condition on multiple attributes. To address the problem that multi-attribute searching algorithms may not work well when the number of attributes (i.e., dimensions) d increases, which is related to the curse of dimensionality, we also propose a probabilistic searching algorithm to reduce searching time at the expense of a small probability of false positives. With this algorithm, we choose permutation pattern(s) of d attributes to use the first K (K « d) ones to search objects so that they contain relevant attributes with a high probability. We argue that our algorithm can identify the target objects at a false positive rate less than 10-6 and a few percentages of tree-searching cost compared with a naive d-dimensional searching under a certain condition.

This paper proposes an effective signal processing scheme using a modulation code with two-dimensional (2D) run-length limited (RLL) constraints for bit-patterned media magnetic recording (BPMR). This 2D signal processing scheme is applied to be one of two-dimensional magnetic recording (TDMR) schemes for shingled magnetic recording on bit patterned media (BPM). A TDMR scheme has been pointed out an important key technology for increasing areal density toward 10Tb/in2. From the viewpoint of 2D signal processing for TDMR, multi-track joint decoding scheme is desirable to increase an effective transfer rate because this scheme gets readback signals from several adjacent parallel tracks and detect recorded data written in these tracks simultaneously. Actually, the proposed signal processing scheme for BPMR gets mixed readback signal sequences from the parallel tracks using a single reading head and these readback signal sequences are equalized to a frequency response given by a desired 2D generalized partial response system. In the decoding process, it leads to an increase in the effective transfer rate by using a single maximum likelihood (ML) sequence detector because the recorded data on the parallel tracks are decoded for each time slot. Furthermore, a new joint pattern-dependent noise-predictive (PDNP) sequence detection scheme is investigated for multi-track recording with media noise. This joint PDNP detection is embed in a ML detector and can be useful to eliminate media noise. Using computer simulation, it is shown that the joint PDNP detection scheme is able to compensate media noise in the equalizer output which is correlated and data-dependent.

The introduction of deep neural networks (DNNs) leads to a significant improvement of the automatic speech recognition (ASR) performance. However, the whole ASR system remains sophisticated due to the dependent on the hidden Markov model (HMM). Recently, a new end-to-end ASR framework, which utilizes recurrent neural networks (RNNs) to directly model context-independent targets with connectionist temporal classification (CTC) objective function, is proposed and achieves comparable results with the hybrid HMM/DNN system. In this paper, we investigate per-dimensional learning rate methods, ADAGRAD and ADADELTA included, to improve the recognition of the end-to-end system, based on the fact that the blank symbol used in CTC technique dominates the output and these methods give frequent features small learning rates. Experiment results show that more than 4% relative reduction of word error rate (WER) as well as 5% absolute improvement of label accuracy on the training set are achieved when using ADADELTA, and fewer epochs of training are needed.

Fine-grained visual categorization (FGVC) has drawn increasing attention as an emerging research field in recent years. In contrast to generic-domain visual recognition, FGVC is characterized by high intra-class and subtle inter-class variations. To distinguish conceptually and visually similar categories, highly discriminative visual features must be extracted. Moreover, FGVC has highly specialized and task-specific nature. It is not always easy to obtain a sufficiently large-scale training dataset. Therefore, the key to success in practical FGVC systems is to efficiently exploit discriminative features from a limited number of training examples. In this paper, we propose an efficient two-step dimensionality compression method to derive compact middle-level part-based features. To do this, we compare both space-first and feature-first convolution schemes and investigate their effectiveness. Our approach is based on simple linear algebra and analytic solutions, and is highly scalable compared with the current one-vs-one or one-vs-all approach, making it possible to quickly train middle-level features from a number of pairwise part regions. We experimentally show the effectiveness of our method using the standard Caltech-Birds and Stanford-Cars datasets.

As three dimensional (3D) discrete wavelet transform (DWT) is widely used for high resolution volumetric data compression, and to further improve the performance of lossless coding, the adaptive directional lifting (ADL) structure based on non-separable 3D DWT with a (5,3) filter is proposed in this paper. The proposed 3D DWT has less lifting steps and better prediction performance compared to the existing separable 3D DWT with fixed filter coefficients. It also has compatibility with the conventional DWT defined by the JPEG2000 international standard. The proposed method shows comparable and better results with the non-separable 3D DWT and separable 3D DWT and it is effective for lossless coding of high resolution volumetric data.

In this paper, we present a new encoding/decoding method for dynamic multidimensional datasets and its implementation scheme. Our method encodes an n-dimensional tuple into a pair of scalar values even if n is sufficiently large. The method also encodes and decodes tuples using only shift and and/or register instructions. One of the most serious problems in multidimensional array based tuple encoding is that the size of an encoded result may often exceed the machine word size for large-scale tuple sets. This problem is efficiently resolved in our scheme. We confirmed the advantages of our scheme by analytical and experimental evaluations. The experimental evaluations were conducted to compare our constructed prototype system with other systems; (1) a system based on a similar encoding scheme called history-offset encoding, and (2) PostgreSQL RDBMS. In most cases, both the storage and retrieval costs of our system significantly outperformed those of the other systems.

In this paper, we propose a communication-efficient top-k continuous query processing method on distributed local nodes where data are horizontally partitioned. A designated coordinator server takes the role of issuing queries from users to local nodes and delivering the results to users. The final results are requested via a top-k subscription which lets local nodes know which data and updates need to be returned to users. Our proposed method makes use of the active previously posed queries to identify a small set of needed top-k subscriptions. In addition, with the pre-indexed nodes' skylines, the number of local nodes to be subscribed can be significantly reduced. As a result, only a small number of subscriptions are informed to a small number of local nodes resulting in lower communication overhead. Furthermore, according to dynamic data updates, we also propose a method that prevents nodes from reporting needless updates and also maintenance procedures to preserve the consistency. The results of experiments that measure the volume of transferred data show that our proposed method significantly outperforms the previously proposed methods.