Somchai Jitapunkul

A Lorentzian stochastic estimation for a robust iterative multiframe super-resolution reconstruction with Lorentzian-Tikhonov regularization

Recently, there has been a great deal of work developing super-resolution reconstruction (SRR) algorithms. While many such algorithms have been proposed, the almost SRR estimations are based on L1 or L2 statistical norm estimation, therefore these SRR algorithms are usually very sensitive to their assumed noise model that limits their utility. The real noise models that corrupt the measure sequence are unknown; consequently, SRR algorithm using L1 or L2 norm may degrade the image sequence rather than enhance it. Therefore, the robust norm applicable to several noise and data models is desired in SRR algorithms. This paper first comprehensively reviews the SRR algorithms in this last decade and addresses their shortcomings, and latter proposes a novel robust SRR algorithm that can be applied on several noise models. The proposed SRR algorithm is based on the stochastic regularization technique of Bayesian MAP estimation by minimizing a cost function. For removing outliers in the data, the Lorentzian error norm is used for measuring the difference between the projected estimate of the high-resolution image and each low-resolution image. Moreover, Tikhonov regularization and Lorentzian-Tikhonov regularization are used to remove artifacts from the final answer and improve the rate of convergence. The experimental results confirm the effectiveness of our method and demonstrate its superiority to other super-resolution methods based on L1 and L2 norms for several noise models such as noiseless, additive white Gaussian noise (AWGN), poisson noise, salt and pepper noise, and speckle noise.

TWO-Dimensional Linear Discriminant Analysis of Principle Component Vectors for Face Recognition

In this paper, we proposed a new two-dimensional linear discriminant analysis (2DLDA) method. Based on two-dimensional principle component analysis (2DPCA), face image matrices do not need to be previously transformed into a vector. In this way, the spatial information can be preserved. Moreover, the 2DLDA also allows avoiding the small sample size (SSS) problem, thus overcoming the traditional LDA. We combine 2DPCA and our proposed 2DLDA on the two-dimensional linear discriminant analysis of principle component vectors framework. Our framework consists of two steps: first we project an input face image into the family of projected vectors via 2DPCA-based technique, second we project from these space into the classification space via 2DLDA-based technique. This does not only allows further reducing of the dimension of the feature matrix but also improving the classification accuracy. Experimental results on ORL and Yale face database showed an improvement of 2DPCA-based technique over the conventional PCA technique

Adaptive Channel Estimation Using Pilot-Embedded Data-Bearing Approach for MIMO-OFDM Systems

Multiple-input multiple-output (MIMO) orthogonal-frequency-division-multiplexing (OFDM) systems employing coherent receivers crucially require channel state information (CSI). Since the multipath delay profile of channels is arbitrary in the MIMO-OFDM systems, an effective channel estimator is needed. In this paper, we first develop a pilot-embedded data-bearing (PEDB) approach for joint channel estimation and data detection, in which PEDB least-square (LS) channel estimator and maximum-likelihood (ML) data detection are employed. Then, we propose an LS fast Fourier transform (FFT)-based channel estimator by employing the concept of FFT-based channel estimation to improve the PEDB-LS one via choosing a certain number of significant taps for constructing a channel frequency response. The effects of model mismatch error inherent in the proposed LS FFT-based estimator when considering noninteger multipath delay profiles and its performance analysis are investigated. The relationship between the mean-squared error (MSE) and the number of chosen significant taps is revealed, and hence, the optimal criterion for obtaining the optimum number of significant taps is explored. Under the framework of pilot embedding, we further propose an adaptive LS FFT-based channel estimator employing the optimum number of significant taps to compensate the model mismatch error as well as minimize the corresponding noise effect. Simulation results reveal that the adaptive LS FFT-based estimator is superior to the LS FFT-based and PEDB-LS estimators under quasi-static channels or low Dopplers shift regimes

Face segmentation based on Hue-Cr components and morphological technique

This paper proposes a novel algorithm to dynamically define the region of interest (ROI) for videophone applications. The algorithm uses the color information Hue and Cr to find the skin-colored pixels and also uses the range of threshold obtained from red and blue components in normalized RGB color space to remove nonskin-colored pixels because the human skin tends to have a predominance of red and nonpredominance of blue. Post-processing is used to remove such noises by a morphological operator. Moreover, the algorithm performs temporal filtering to remove skin-color pixels that immediately appear from frame to frame by using an object tracking process to perform as memory for collecting skin-color objects obtained from the previous frame to guide the next frame. The experimental results confirm the effectiveness of the proposed algorithm.

A Robust Iterative Multiframe Super-Resolution Reconstruction using a Huber Bayesian Approach with Huber-Tikhonov Regularization

The traditional SRR (super-resolution reconstruction) estimations are based on L1 or L2 statistical norm estimation therefore these SRR methods are usually very sensitive to their assumed model of data and noise that limits their utility. This paper reviews some of these SRR methods and addresses their shortcomings. We propose a novel SRR approach based on the stochastic regularization technique of Bayesian MAP estimation by minimizing a cost function. The Huber norm is used for measuring the difference between the projected estimate of the high-resolution image and each low resolution image, removing outliers in the data and Tikhonov and Huber-Tikhonov regularization are used to remove artifacts from the final answer and improve the rate of convergence. The experimental results confirm the effectiveness of our methods and demonstrate its superiority to other super-resolution methods based on L1 and L2 norm for several noise models such as noiseless, AWGN, Poisson and salt & pepper noise

Thai syllable segmentation for connected speech based on energy

This paper proposes a novel technique based on local maximum and minimum energy contour to segment syllables of connected speech. Energy contour used in this technique was obtained from five energy algorithms: the absolute energy, the root mean square energy, the square energy, Teager energy and the modified Teager energy. The experimental result was conducted on 36 utterances, 11 speakers (7 male and 4 females) for evaluation empirical threshold and parameter values in each energy algorithm. The result from our technique that is based on the local maximum and minimum energy was better than the commonly used endpoint detection technique based on level equalizer proposed by Lamel L.F et al. [1981].

Online Thai handwritten character recognition using hidden Markov models and support vector machines

We propose a method for online Thai handwritten character recognition using HMMs and SVMs with score-space kernels. Score-space kernels are generalized Fisher kernels based on underlying generative models, such as Gaussian mixture models (GMMs), which are output distributions of each state in HMMs. Our system combines the advantages of both generative and discriminative classifiers. In the first phase, HMMs are used for multi-classification, then SVMs are applied to resolve any uncertainty remaining after the first-pass HMM-based recognizer, but they are not applied for all classes because the results of some classes are worse. We consider the HMM confusion matrix to find the confused candidates in each class. If there is one candidate, it means there is no confusion in this class, and HMMs alone are sufficient to classify. SVMs are applied if there is more than one candidate. If there are more than two, the multi-class method is applied. On account of the basic score-spaces, likelihood and likelihood ratio score-spaces are not symmetrical. In the case of likelihood score-space, the parameters refer to only one generative model from two class models. In the case of likelihood ratio score-space, the parameters refer to both of them, but in different positions; thus one observation sequence can map to two score-vectors. We propose a new symmetric score-space, called symmetric likelihood ratio score-space. In this way, one observation sequence is mapped to only one score-vector. Experimental results show the average recognition rate improved from 89.9%, using baseline HMM, to 92.5%, using our proposed method.

A Data-Bearing Approach for Pilot-Embedding Frameworks in Space-Time Coded MIMO Systems

Space-time (ST) coded MIMO systems employing coherent detectors crucially require channel state information. This paper presents a novel pilot-embedding framework for channel estimation and data detection by exploiting the null-space property and the orthogonality property of the data-bearer and pilot matrices. The ST data matrix is firstly projected onto the data bearer matrix, which is a null-space of the pilot matrix, and the resulting matrix and the pilot matrix are combined for transmitting. The data and pilot extractions are achieved independently through linear transformations by exploiting the null-space property. The unconstrained maximum-likelihood (ML) and linear minimum mean-squared error (lmmse) estimators are explored for channel estimation. Then the ML approach for data detection is developed by exploiting the orthogonality property. The mean-squared error (mse) of channel estimation, Cramer-Rao lower bound (CRLB), and the Chernoffs bound of a pair-wise error probability for ST codes are analyzed for examining the performance of the proposed scheme. The optimum power allocation scheme for data and pilot parts is also considered. Three data-bearer and pilot structures, including time-multiplexing (TM)-based, ST-block-code (STBC)-based, and code-multiplexing (CM)-based, are proposed. Simulation results show that the CM-based structure provides superior performance for nonquasi-static flat Rayleigh fading channels, while these three structures yield similar performances for quasi-static flat Rayleigh fading channels

A New Unified Lossless/Lossy Image Compression Based on a New Integer DCT

In this paper, we propose a new one-dimensional (1D) integer discrete cosine transform (Int-DCT) for unified lossless/lossy image compression. The proposed 1D Int-DCT is newly designed to reduce rounding effects by minimizing number of rounding operations. The proposed Int-DCT can be operated not only lossless coding for a high quality decoded image but also lossy coding for a compatibility with the conventional DCT-based coding system. Both theoretical analysis and simulation results confirm an effectiveness of the proposed Int-DCT.

A speaker-independent Thai polysyllabic word recognition using hidden Markov model

This correspondence presents a speech recognition system of speaker-independent Thai polysyllabic words. This development is based on the discrete hidden Markov model in conjunction with vector quantization algorithm, endpoint detection algorithm for syllable endpoint detection and separation, and time normalization algorithm. The 70-Thai word vocabulary is subdivided into four sets comprising single, double, and triple syllabled words, 20 words in each set, and the last set consists of 10-Thai numeric words, zero to nine. The separated speech training set and testing set are composed of both male and female speakers within the range of 18 to 25 years old. For the tonal characteristics of the Thai language, the algorithms and the model parameters are modified in order to be applicable to the Thai language. The experiments on the effects of model parameter variations on recognition rate are conducted. The model parameters are number of codebooks, number of model states, and number of training speakers. The results show that the increase in the number of codebook and the number of model states have the major effect on the recognition rates. Also, the number of training speakers has less effect than the others. The average recognition rate of this speaker-independent recognition system is 89.906 percent for 40 speakers testing set using 256 vector codebook of 10-order linear prediction coefficients and 15-state model parameters. The recognition rate of the four sets of words are 86.750 percent for single-syllabled words, 92.375 percent for double-syllabled words, 96.250 percent for triple-syllabled words, and 84.250 percent for the numeric words.