Hwang-Ki Min

Complexity-Reduced Scheme for Feature Extraction With Linear Discriminant Analysis

Owing to the singularity of the within-class scatter, linear discriminant analysis (LDA) becomes ill-posed for small sample size (SSS) problems. Null-space-based LDA (NLDA), which is an extension of LDA, provides good discriminant performances for SSS problems. Yet, as the original scheme for the feature extractor (FE) of NLDA suffers from a complexity burden, a few modified schemes have since been proposed for complexity reduction. In this brief, by transforming the problem of finding the FE of NLDA into a linear equation problem, a novel scheme is derived, offering a further reduction of the complexity.

A computationally efficient scheme for feature extraction with kernel discriminant analysis

The kernel discriminant analysis (KDA), an extension of the linear discriminant analysis (LDA) and null space-based LDA into the kernel space, generally provides good pattern recognition (PR) performance for both small sample size (SSS) and non-SSS PR problems. Due to the eigen-decomposition technique adopted, however, the original scheme for the feature extraction with the KDA suffers from a high complexity burden. In this paper, we derive a transformation of the KDA into a linear equation problem, and propose a novel scheme for the feature extraction with the KDA. The proposed scheme is shown to provide us with a reduction of complexity without degradation of PR performance. In addition, to enhance the PR performance further, we address the incorporation of regularization into the proposed scheme. HighlightsWe propose a complexity-reduced scheme for the kernel discriminant analysis.The proposed scheme does not cause degradation of recognition accuracy performance.The core is the transformation of the original problem into a linear equation problem.We further address the incorporation of regularization into the proposed scheme.

Detection of Signals With Observations in Multiple Subbands: A Scheme of Wideband Spectrum Sensing for Cognitive Radio With Multiple Antennas

We address detection schemes of spectrum sensing for cognitive radio with multiple receive antennas operating over a wideband channel composed of a multitude of subbands. By taking the observations in all subbands into consideration in the likelihood functions for sensing a subband, the test statistics of the proposed schemes are functions of the sample covariance matrix in the subband under consideration and that in the subband exhibiting the lowest power spectral density. The false alarm and detection probabilities of the proposed schemes are analyzed theoretically and confirmed via simulations when the numbers of observations are the same for all the subbands. It is shown through computer simulations that the proposed schemes can provide considerable performance gains over conventional schemes for wideband spectrum sensing when the observations are spatially correlated and temporally independent/dependent.

Likelihood ratio test for wideband spectrum sensing

In this paper, we propose a novel detection scheme of spectrum sensing for cognitive radio with multiple receive antennas operating over a wideband channel composed of a multitude of subbands. By taking the observations in all subbands into consideration in the likelihood ratio test for sensing a subband, the proposed scheme can provide better performance than other conventional schemes. It is shown through computer simulations that the proposed scheme can provide considerable performance gains over conventional schemes for wideband spectrum sensing.

Abnormal Signal Detection in Gas Pipes Using Neural Networks

In this paper, we present a real-time system to detect abnormal events on gas pipes, based on the signals which are observed through the audio sensors attached on them. First, features are extracted from this signal so that they are robust to noise and invariant to the distance between a sensor and a spot at which an abnormal event like an attack on the gas pipes occurs. Then, a classifier is constructed to detect abnormal events using neural networks. It is a combination of two neural network models, a Gaussian mixture model and a multi-layer perceptron, for the reduction of miss and false alarms. The former works for miss alarm prevention and the latter for false alarm prevention. The experimental result with real data from the actual gas system shows that the propose system is effective in detecting the dangerous events in real-time having an accuracy of 92.9%.

Spectrum Sensing with Receive Diversity for Cognitive Radio Operating over Wideband Channel

A detection scheme of spectrum sensing is discussed for cognitive radio with multiple receive antennas operating over a wideband channel composed of a multitude of sub bands. By taking the observations in all sub bands into consideration in the likelihood ratio test for sensing a sub band, the proposed scheme can provide better performance than other conventional schemes.

Complexity reduction of kernel discriminant analysis

As an extension of the linear discriminant analysis (LDA), the kernel discriminant analysis (KDA) generally results in good pattern recognition performance for both small sample size (SSS) and non-SSS problems. Yet, the original scheme based on the eigen-decomposition technique suffers from a complexity burden. In this paper, by transforming the problem of finding the feature extractor (FE) of the KDA into a linear equation problem, reduction of the complexity is accomplished via a novel scheme for the FE of the KDA.

Non-intrusive Appliance Load Monitoring with Feature Extraction from Higher Order Moments

A pattern recognition (PR) system is addressed for nonintrusive appliance load monitoring. For the effective recognition of two home appliances (specifically, an electric iron and a cook top), we consider a novel feature extraction method employing higher order moments of power signals from the appliances. Through simulation results, we have confirmed that the PR system with the features from the proposed higher order moment technique and kernel discriminant analysis can effectively separate the two appliances.

Cooperative spectrum sensing based on generalized likelihood ratio test under impulsive noise circumstances

We propose nonlinear schemes for the spectrum sensing in cooperative cognitive radio networks under impulsive noise circumstances. By jointly employing the order statistics, generalized likelihood ratio test, and counting rule in the framework of spectrum sensing, the proposed scheme exhibits a better performance than the conventional counterparts. From simulation results, it is confirmed that the proposed scheme provides significant performance improvements over the conventional schemes.

Complexity reduction for null space-based linear discriminant analysis

In small sample size problems, the null space-based linear discriminant analysis (NLDA) provides a good discrimination performance but suffers from a complexity burden. Some schemes based on QR factorization and eigen-decomposition have been proposed for complexity reduction. In this paper, we propose a scheme based on Cholesky decomposition for a further reduction of the complexity.