Squared-loss mutual information (SMI) is a robust measure of the statistical dependence between random variables. The sample-based SMI approximator called least-squares mutual information (LSMI) was demonstrated to be useful in performing various machine learning tasks such as dimension reduction, clustering, and causal inference. The original LSMI approximates the pointwise mutual information by using the kernel model, which is a linear combination of kernel basis functions located on paired data samples. Although LSMI was proved to achieve the optimal approximation accuracy asymptotically, its approximation capability is limited when the sample size is small due to an insufficient number of kernel basis functions. Increasing the number of kernel basis functions can mitigate this weakness, but a naive implementation of this idea significantly increases the computation costs. In this article, we show that the computational complexity of LSMI with the multiplicative kernel model, which locates kernel basis functions on unpaired data samples and thus the number of kernel basis functions is the sample size squared, is the same as that for the plain kernel model. We experimentally demonstrate that LSMI with the multiplicative kernel model is more accurate than that with plain kernel models in small sample cases, with only mild increase in computation time.

The estimation of the power spectral density (PSD) of noise is crucial for retrieving speech in noisy environments. In this study, we propose a novel method for estimating the non-white noise PSD from noisy speech on the basis of a generalized gamma distribution and the minimum mean square error (MMSE) approach. Because of the highly non-stationary nature of speech, deriving its actual spectral probability density function (PDF) using conventional modeling techniques is difficult. On the other hand, spectral components of noise are more stationary than those of speech and can be represented more accurately by a generalized gamma PDF. The generalized gamma PDF can be adapted to optimally match the actual distribution of the noise spectral amplitudes observed at each frequency bin utilizing two real-time updated parameters, which are calculated in each frame based on the moment matching method. The MMSE noise PSD estimator is derived on the basis of the generalized gamma PDF and Gaussian PDF models for noise and speech spectral amplitudes, respectively. Combined with an improved Weiner filter, the proposed noise PSD estimate method exhibits the best performance compared with the minimum statistics, weighted noise estimation, and MMSE-based noise PSD estimation methods in terms of both subjective and objective measures.

In this study, the influence of Gd composition on current-induced domain wall motion in a Gd-Co ferrimagnetic nanowire was theoretically investigated with taking into account of composition dependence of magnetic properties. As a result, the intrinsic critical density to move domain wall significantly reduces near the compensation composition, which is achieved to be less than 105A/cm2. Moreover, the intrinsic critical current density also significantly reduces near a certain Gd composition where the domain wall energies of Bloch and Néel walls are almost the same.

Spatially coupled (SC) low-density parity-check (LDPC) codes are defined by bipartite graphs that are obtained by assembling prototype graphs. The combination and connection of prototype graphs are designated by specifying some parameters, and Kudekar et al. showed that BP threshold of the ensemble of SC LDPC codes agrees with MAP threshold of the ensemble of regular LDPC codes when those parameters are grown up so that the code length tends to infinity. When we design SC LDPC codes with practical code length, however, it is not clear how to set those parameters to enhance the performance of SC LDPC codes. In this paper, we provide the result of numerical experiments that suggest the dependence of error performance of SC LDPC codes over BEC on their design parameters.

Recently, non-binary low-density parity-check (NB-LDPC) codes starts to show their superiority in achieving significant coding gains when moderate codeword lengths are adopted. However, the overwhelming decoding complexity keeps NB-LDPC codes from being widely employed in modern communication devices. This paper proposes a hybrid message-passing decoding algorithm which consumes very low computational complexity. It achieves competitive error performance compared with conventional Min-max algorithm. Simulation result on a (255,174) cyclic code shows that this algorithm obtains at least 0.5dB coding gain over other state-of-the-art low-complexity NB-LDPC decoding algorithms. A partial-parallel NB-LDPC decoder architecture for cyclic NB-LDPC codes is also developed based on this algorithm. Optimization schemes are employed to cut off hard decision symbols in RAMs and also to store only part of the reliability messages. In addition, the variable node units are redesigned especially for the proposed algorithm. Synthesis results demonstrate that about 24.3% gates and 12% memories can be saved over previous works.

In this paper, we propose a synthesizable LDPC decoder IP core for the WiMAX system with high parallelism and enhanced error-correcting performance. By taking the advantages of both layered scheduling and fully-parallel architecture, the decoder can fully support multi-mode decoding specified in WiMAX with the parallelism much higher than commonly used partial-parallel layered LDPC decoder architecture. 6-bit quantized messages are split into bit-serial style and 2bit-width serial processing lines work concurrently so that only 3 cycles are required to decode one layer. As a result, 12∼24 cycles are enough to process one iteration for all the code-rates specified in WiMAX. Compared to our previous bit-serial decoder, it doubles the parallelism and solves the message saturation problem of the bit-serial arithmetic, with minor gate count increase. Power synthesis result shows that the proposed decoder achieves 5.83pJ/bit/iteration energy efficiency which is 46.8% improvement compared to state-of-the-art work. Furthermore, an advanced dynamic quantization (ADQ) technique is proposed to enhance the error-correcting performance in layered decoder architecture. With about 2% area overhead, 6-bit ADQ can achieve the error-correcting performance close to 7-bit fixed quantization with improved error floor performance.

Bit-patterned medium (BPM) is one of the promising approaches for ultra-high density magnetic recording systems. However, BPM requires precise write synchronization, and exhibits write-errors due to insufficient write field gradient, medium switching field distribution (SFD), demagnetization field from adjacent islands, and island position variation. In this paper, an iterative decoding system using a non-binary low-density parity-check (LDPC) code is considered for a BPM R/W channel with write-errors at an areal recording density of 2Tbit/inch2 including the coding rate loss. The performance of the iterative decoding system using the non-binary LDPC code over the Galois field GF(28) is evaluated by computer simulation, and it is compared with the conventional iterative decoding system using a binary LDPC code. The results show that the non-binary LDPC system has a larger write margin than the binary LDPC system.

Mutual information (MI) is a standard measure of statistical dependence of random variables. However, due to the log function and the ratio of probability densities included in MI, it is sensitive to outliers. On the other hand, the L2-distance variant of MI called quadratic MI (QMI) tends to be robust against outliers because QMI is just the integral of the squared difference between the joint density and the product of marginals. In this paper, we propose a kernel least-squares QMI estimator called least-squares QMI (LSQMI) that directly estimates the density difference without estimating each density. A notable advantage of LSQMI is that its solution can be analytically and efficiently computed just by solving a system of linear equations. We then apply LSQMI to dependence-maximization clustering, and demonstrate its usefulness experimentally.

Feature selection is a technique to screen out less important features. Many existing supervised feature selection algorithms use redundancy and relevancy as the main criteria to select features. However, feature interaction, potentially a key characteristic in real-world problems, has not received much attention. As an attempt to take feature interaction into account, we propose 1-LSMI, an 1-regularization based algorithm that maximizes a squared-loss variant of mutual information between selected features and outputs. Numerical results show that 1-LSMI performs well in handling redundancy, detecting non-linear dependency, and considering feature interaction.

In this paper, an area-efficient decoder architecture is proposed for the quasi-cyclic low-density parity check (QC-LDPC) codes specified in the IEEE 802.16e WiMAX standard. The decoder supports all the code rates and codeword lengths defined in the standard. In order to achieve low area and maximize hardware utilization, the decoder utilizes 4 decoding function units, which is the greatest common divisor of the expansion factors. In addition, the decoder adopts a novel scheduling scheme named stride scheduling, which stores the extrinsic messages in non-sequential order to replace the conventional complex flexible permutation network with simple small-sized cyclic shifters and also minimize the number of memory accesses. To further minimize the complexity, the number of extrinsic memory instances for 24 block columns is reduced to 5 banks by identifying independent sets. All the memory instances used in the decoder are single-port memories which cost less area and price compared to dual-port ones. Finally, the decoding function units have partially parallel structure to make the decoding throughput sufficiently over the requirement of the WiMAX standard. The proposed decoder is synthesized with 49 K equivalent gates and 54,144 bits of memory, and the implementation occupies 0.40 mm2 in a 65 nm CMOS technology.

In this paper, we present a novel class of long quasi-cyclic low-density parity-check (QC-LDPC) codes. Each of the codes in this class has a structure formed by concatenating single-parity-check codes and QC-LDPC codes of shorter lengths, which allows for efficient, high throughput encoder/decoder implementations. Using a code in this class, we design a forward error correction (FEC) scheme for optical transmission systems and present its high throughput encoder/decoder architecture. In order to demonstrate its feasibility, we implement the architecture on a field programmable gate array (FPGA) platform. We show by both FPGA-based simulations and measurements of an optical transmission system that the FEC scheme can achieve excellent error performance and that there is no significant performance degradation due to the constraint on its structure while getting an efficient, high throughput implementation is feasible.

We propose a novel threshold-free salient object detection approach which integrates both saliency density and edge response. The salient object with a well-defined boundary can be automatically detected by our approach. Saliency density and edge response maximization is used as the quality function to direct the salient object discovery. The global optimal window containing a salient object is efficiently located through the proposed saliency density and edge response based branch-and-bound search. To extract the salient object with a well-defined boundary, the GrabCut method is applied, initialized by the located window. Experimental results show that our approach outperforms the methods only using saliency or edge response and achieves a comparable performance with the best state-of-the-art method, while being without any threshold or multiple iterations of GrabCut.

Turbo equalization is an iterative equalization and decoding technique that can achieve impressive performance gains for communication systems. In this letter, we investigate the turbo equalization method for the decoding of the Davey-MacKay (DM) construction over the IDS-AWGN channels, which indicates a cascaded insertion, deletion, substitution (IDS) channel and an additive white Gaussian noise (AWGN) channel. The inner decoder for the DM construction can be seen as an maximum a-posteriori (MAP) detector. It receives the beliefs generated by the outer LDPC decoder when turbo equalization is used. Two decoding schemes with different kinds of inner decoders, namely hard-input inner decoder and soft-input inner decoder, are investigated. Simulation results show that significant performance gains are obtained for both decoders with respect to the insertion/deletion probability at different SNR values.

In this paper, through extrinsic information transfer (EXIT) band chart analysis, an adaptive iterative decoding approach (AIDA) is proposed to reduce the iterative decoding complexity and delay for finite-length differentially encoded Low-density parity-check (DE-LDPC) coded systems with multiple-symbol differential detection (MSDD). The proposed AIDA can adaptively adjust the observation window size (OWS) of the MSDD soft-input soft-output demodulator (SISOD) and the outer iteration number of the iterative decoder (consisting of the MSDD SISOD and the LDPC decoder) instead of setting fixed values for the two parameters of the considered systems. The performance of AIDA depends on its stopping criterion (SC) which is used to terminate the iterative decoding before reaching the maximum outer iteration number. Many SCs have been proposed; however, these approaches focus on turbo coded systems, and it has been proven that they do not well suit for LDPC coded systems. To solve this problem, a new SC called differential mutual information (DMI) criterion, which can track the convergence status of the iterative decoding, is proposed; it is based on tracking the difference of the output mutual information of the LDPC decoder between two consecutive outer iterations of the considered systems. AIDA using the DMI criterion can adaptively adjust the out iteration number and OWS according to the convergence situation of the iterative decoding. Simulation results show that compared with using the existing SCs, AIDA using the DMI criterion can further reduce the decoding complexity and delay, and its performance is not affected by a change in the LDPC code and transmission channel parameters.

Existing salient object detection methods either simply use a threshold to detect desired salient objects from saliency map or search the most promising rectangular window covering salient objects on the saliency map. There are two problems in the existing methods: 1) The performance of threshold-dependent methods depends on a threshold selection and it is difficult to select an appropriate threshold value. 2) The rectangular window not only covers the salient object but also contains background pixels, which leads to imprecise salient object detection. For solving these problems, a novel saliency threshold-free method for detecting the salient object with a well-defined boundary is proposed in this paper. We propose a novel window search algorithm to locate a rectangular window on our saliency map, which contains as many as possible pixels belonging the salient object and as few as possible background pixels. Once the window is determined, GrabCut is applied to extract salient object with a well-defined boundary. Compared with existing methods, our approach doesn't need any threshold to binarize the saliency map and additional operations. Experimental results show that our approach outperforms 4 state-of-the-art salient object detection methods, yielding higher precision and better F-Measure.

This paper proposes an efficient broadcast scheme based on traffic density measurement to mitigate broadcast storms in Vehicular Ad Hoc Networks (VANETs). In a VANET, the number of vehicles that rebroadcasts a message is closely related with the collision ratio of the message, so a well-designed broadcast scheme should consider traffic density when rebroadcasting a message. The proposed scheme introduces a traffic density measurement scheme and broadcast scheme for VANET. It is based on the slotted p-persistence scheme, but the rebroadcast procedure is enhanced and the rebroadcast probability p is controlled dynamically according to the estimated traffic density. Simulation results demonstrate that the proposed scheme outperforms existing schemes in terms of the end-to-end delay and collision ratio.

When the Akebono (EXOS-D) satellite passed through the plasmasphere, a series of lightning whistlers was observed by its analog wideband receiver (WBA). Recently, we developed an intelligent algorithm to detect lightning whistlers from WBA data. In this study, we analyzed two typical events representing the clear dispersion characteristics of lightning whistlers along the trajectory of Akebono. The event on March 20, 1991 was observed at latitudes ranging from 47.83 (47,83N) to -11.09 (11.09S) and altitudes between ∼2232 and ∼7537 km. The other event on July 12, 1989 was observed at latitudes from 34.94 (34.94N) and -41.89 (41.89S) and altitudes ∼1420–∼7911 km. These events show systematic trends; hence, we can easily determine whether the wave packets of lightning whistlers originated from lightning strikes in the northern or the southern hemispheres. Finally, we approximated the path lengths of these lightning whistlers from the source to the observation points along the Akebono trajectory. In the calculations, we assumed the dipole model as a geomagnetic field and two types of simple electron density profiles in which the electron density is inversely proportional to the cube of the geocentric distance. By scrutinizing the dipole model we propose some models of dispersion characteristic that proportional to the electron density. It was demonstrated that the dispersion D theoretically agrees with observed dispersion trend. While our current estimation is simple, it shows that the difference between our estimation and observation data is mainly due to the electron density profile. Furthermore, the dispersion analysis of lightning whistlers is a useful technique for reconstructing the electron density profile in the Earth's plasmasphere.

Entangled states play crucial roles in quantum information theory and its applied technologies. In various protocols such as quantum teleportation and quantum key distribution, a good entangled state shared by a pair of distant players is indispensable. In this paper, we numerically examine entanglement sharing protocols using quantum LDPC CSS codes. The sum-product decoding method enables us to detect uncorrectable errors, and thus, two protocols, Detection and Resending (DR) protocol and Non-Detection (ND) protocol are considered. In DR protocol, the players abort the protocol and repeat it if they detect the uncorrectable errors, whereas in ND protocol they do not abort the protocol. We show that DR protocol yields smaller error rate than ND protocol. In addition, it is shown that rather high reliability can be achieved by DR protocol with quantum LDPC CSS codes.

We performed the (NH4)2S surface treatments before Al2O3 deposition to improve the Al2O3/III-Nitride interface quality in Al2O3/AlGaN/GaN metal-oxide-semiconductor heterostructure field-effect transistors (MOSHFETs). Interface state density at the Al2O3/GaN interface was decreased by the (NH4)2S treatment. The hysteresis width in ID-VGS and gm-VGS characteristics of the Al2O3/AlGaN MOSHFETs with the (NH4)2S treatment was smaller than that without the (NH4)2S treatment. In addition, transconductance (gm) decrease at a large gate voltage was relaxed by the (NH4)2S treatment. We also performed ultraviolet (UV) illumination during the (NH4)2S treatment for further improvement of the Al2O3/III-Nitride interface quality. Interface state density of the Al2O3/GaN MOS diodes with the UV illumination was smaller than that without the UV illumination.

In this paper, a fast and automated method of counting pedestrians in crowded areas is proposed along with three contributions. We firstly propose Local Empirical Templates (LET), which are able to outline the foregrounds, typically made by single pedestrians in a scene. LET are extracted by clustering foregrounds of single pedestrians with similar features in silhouettes. This process is done automatically for unknown scenes. Secondly, comparing the size of group foreground made by a group of pedestrians to that of appropriate LET captured in the same image patch with the group foreground produces the density ratio. Because of the local scale normalization between sizes, the density ratio appears to have a bound closely related to the number of pedestrians who induce the group foreground. Finally, to extract the bounds of density ratios for groups of different number of pedestrians, we propose a 3D human models based simulation in which camera viewpoints and pedestrians' proximity are easily manipulated. We collect hundreds of typical occluded-people patterns with distinct degrees of human proximity and under a variety of camera viewpoints. Distributions of density ratios with respect to the number of pedestrians are built based on the computed density ratios of these patterns for extracting density ratio bounds. The simulation is performed in the offline learning phase to extract the bounds from the distributions, which are used to count pedestrians in online settings. We reveal that the bounds seem to be invariant to camera viewpoints and humans' proximity. The performance of our proposed method is evaluated with our collected videos and PETS 2009's datasets. For our collected videos with the resolution of 320 × 240, our method runs in real-time with good accuracy and frame rate of around 30 fps, and consumes a small amount of computing resources. For PETS 2009's datasets, our proposed method achieves competitive results with other methods tested on the same datasets [1],[2].