Gang Wei

Exploiting Local Coherent Patterns for Unsupervised Feature Ranking

Prior to pattern recognition, feature selection is often used to identify relevant features and discard irrelevant ones for obtaining improved analysis results. In this paper, we aim to develop an unsupervised feature ranking algorithm that evaluates features using discovered local coherent patterns, which are known as biclusters. The biclusters (viewed as submatrices) are discovered from a data matrix. These submatrices are used for scoring relevant features from two aspects, i.e., the interdependence of features and the separability of instances. The features are thereby ranked with respect to their accumulated scores from the total discovered biclusters before the pattern classification. Experimental results show that this proposed method can yield comparable or even better performance in comparison with the well-known Fisher score, Laplacian score, and variance score using three UCI data sets, well improve the results of gene expression data analysis using gene ontology annotation, and finally demonstrate its advantage of unsupervised feature ranking for high-dimensional data.Prior to pattern recognition, feature selection is often used to identify relevant features and discard irrelevant ones for obtaining improved analysis results. In this paper, we aim to develop an unsupervised feature ranking algorithm that evaluates features using discovered local coherent patterns, which are known as biclusters. The biclusters (viewed as submatrices) are discovered from a data matrix. These submatrices are used for scoring relevant features from two aspects, i.e., the interdependence of features and the separability of instances. The features are thereby ranked with respect to their accumulated scores from the total discovered biclusters before the pattern classification. Experimental results show that this proposed method can yield comparable or even better performance in comparison with the well-known Fisher score, Laplacian score, and variance score using three UCI data sets, well improve the results of gene expression data analysis using gene ontology annotation, and finally demonstrate its advantage of unsupervised feature ranking for high-dimensional data.

Linear Tracking for 3-D Medical Ultrasound Imaging

As the clinical application grows, there is a rapid technical development of 3-D ultrasound imaging. Compared with 2-D ultrasound imaging, 3-D ultrasound imaging can provide improved qualitative and quantitative information for various clinical applications. In this paper, we proposed a novel tracking method for a freehand 3-D ultrasound imaging system with improved portability, reduced degree of freedom, and cost. We designed a sliding track with a linear position sensor attached, and it transmitted positional data via a wireless communication module based on Bluetooth, resulting in a wireless spatial tracking modality. A traditional 2-D ultrasound probe fixed to the position sensor on the sliding track was used to obtain real-time B-scans, and the positions of the B-scans were simultaneously acquired when moving the probe along the track in a freehand manner. In the experiments, the proposed method was applied to ultrasound phantoms and real human tissues. The results demonstrated that the new system outperformed a previously developed freehand system based on a traditional six-degree-of-freedom spatial sensor in phantom and in vivo studies, indicating its merit in clinical applications for human tissues and organs.

Blind linear channel estimation using genetic algorithm and SIMO model

In this paper, we propose to use genetic algorithm (GA) to solve the blind infinite-impulse-response (IIR) channel estimation problem. The contributions of this paper are three-fold: (1) We prove that by oversampling the output of a single-input-single-output IIR channel, one can build a single-input-multiple-output (SIMO) model in which the subchannels are IIR channels with the same Autoregressive (AR) order and coefficients. (2) Based on this SIMO model, we further develop a second-order statistics based objective function that includes the unknown model order and parameters whereas most of the existing work must assume the channel order is known in advance. (3) A GA is proposed to deal with this optimisation problem in that we encode the model order and parameters into one single chromosome. Therefore the order and parameters can be estimated simultaneously. Computer simulation results indicate the effectiveness of the proposed algorithms.

Fast communication: Adaptive beamforming based on covariance matrix reconstruction by exploiting interferences' cyclostationarity

Adaptive beamforming is known to be sensitive to array system mismatch, especially when the sample covariance matrix is used and the desired signal is present in the training snapshot. To alleviate the sensitivity, in this paper, the sample covariance matrix is replaced by the interference-plus-noise covariance matrix (INCM), which is reconstructed by exploiting the cyclostationarity of interference signals. In contrast to the existing INCM reconstruction methods, the proposed technique is based on the knowledge of the interferences cycle frequencies and needs no information of the array structure, thus it can deal with unknown perturbations in the array. The numerical simulations show that the proposed method improves the robustness of adaptive beamformers and has superior performance to the existing INCM reconstruction methods especially for strong interferences.

Fast communication: An exact analysis of Modified Covariance frequency estimation algorithm based on correlation of single-tone

The Modified Covariance (MC) algorithm, which exploits the time-domain correlation of sinusoidal signals, has been shown to be a simple but efficient method for the frequency estimation of single-tone. Recently MC method based on correlation sequence was investigated. Both simulation and theoretical analyses show that it is more robust to noise than the conventional MC method that is directly based on the recorded signal [6]. However, the analysis in [6] is only valid when lower-order lags of correlations are applied. In this communication we provide new analytical result to tackle this difficult.

Geometric moment in image watermarking

Digital image watermarking has become a popular technique for authentication and copyright protection. However, many proposed image watermarking techniques are sensitive to geometric distortions, such as rotation and scaling. Rotation and scaling, even by slight amount, can make the decoder disabled unless the corrupted image can be rescaled and rotated back to its original size and original orientation. So it is very important to estimate the rotation angle and scaling factor correctly before watermark detection. In our approach, we propose a new way to estimate the rotation angle and scaling factor after the watermarked image has been scaled or/and rotated by using geometric moments of original image. The experimental results show that our method has a good robustness to wide rotation angle and scaling factor ranges.

Accelerated reweighted nuclear norm minimization algorithm for low rank matrix recovery

In this paper we propose an accelerated reweighted nuclear norm minimization algorithm to recover a low rank matrix. Our approach differs from other iterative reweighted algorithms, as we design an accelerated procedure which makes the objective function descend further at every iteration. The proposed algorithm is the accelerated version of a state-of-the-art algorithm. We provide a new analysis of the original algorithm to derive our own accelerated version, and prove that our algorithm is guaranteed to converge to a stationary point of the reweighted nuclear norm minimization problem. Numerical results show that our algorithm requires distinctly fewer iterations and less computational time than the original one to achieve the same (or very close) accuracy, in some problem instances even require only about 50% computational time of the original one, and is also notably faster than several other state-of-the-art algorithms. HighlightsPropose an accelerated reweighted nuclear norm minimization algorithm to recover a low rank matrix.Provide a new analysis for reweighted nuclear norm minimization algorithm.Provide convergence analysis.Numerical results show that our algorithm requires distinctly fewer iterations and less computational time than the original one.

Quasi-clean Speech Construction Based Speech Quality Evaluation under Complex Environments

Objective evaluation of speech quality for complex environments is a main part of the quality of communications service. Usually, intrusive approaches outperform non-intrusive approaches because reference is used in the intrusive ones. This paper presented a new non-intrusive evaluation method for complex environments in which an improved noise tracking and subtraction algorithm is used to obtain the quasi-clean speech from the noisy speech. The quasi-clean speech is regarded as the reference speech to a perceptual model, which is a modified version of ITU-T PESQ. The perceptual model acquires the Mean Opinion Score (MOS) via the measurement of the distortions between the noisy speech and quasi clean speech. Experimental results demonstrate that the presented method got a correlation coefficient of 0.92 on NOIZEUS noisy dataset, which is 99% similar to that of PESQ, and 6.4% superior to ITUT P.563.

Blind IR Channel Equalization Based on Second-Order Statistics

Direct blind equalization of single-input-single-output (SISO) infinite-impulse-response (IIR) channels is investigated by oversampling the channel output. Only second-order statistics is required and the channel order need not to be known. We first transform the oversampled channel into a special single-input-multiple-output (SIMO) model. Then a linear prediction approach is applied to partially estimate the model parameters. Based on the estimated parameters, zero-forcing equalizers and MMSE equalizers are developed and their efficiency are evaluated by computer simulations

On the mapping capacity of feedforward neural networks

This paper addresses the mapping capacity of feedforward neural networks (FFNNs). We propose a new condition of the activation function for FFNNs to be capable of approximating any continuous maps defined on /spl Rfr//sup n/. Based on the result, we construct several FFNN based nonlinear predictors for speech signals, and study their performances.