Chih-Yu Hsu

Optimization-based image watermarking with integrated quantization embedding in the wavelet-domain

This study presents an optimization-based image watermarking scheme with integrated quantization embedding. First, peak signal to noise ratio (PSNR) is rewritten as a performance index in matrix form. In order to guarantee the robustness, this study embeds a watermark into the low-frequency coefficients of discrete wavelet transform (DWT). Unlike traditional way of single-coefficient quantization, this study applies amplitude quantization to embed the watermark and then rewrite this amplitude quantization as a constraint with embedding state. Then, an optimization-based equation connecting the performance index and amplitude-quantization constraint is obtained. Second, Lagrange Principle is used to solve the equation and then the optimal results are applied to embed the watermark. In detection, the hidden watermark can be extracted without original image. Finally, the performance of the proposed scheme is evaluated by PSNR and BER.

Optimization-Based Embedding for Wavelet-Domain Audio Watermarking

This work proposes a new blind digital audio watermarking system by using optimization-based modification on low-frequency amplitude of discrete wavelet transform (DWT). To modify the low-frequency amplitude under the best signal-to-noise ratio (SNR), the proposed embedding system minimizes the difference between the original and the embedded coefficients. Accordingly, an optimization-based embedding formula connecting the SNR and the embedding system is derived. The formula is then applied to embed the synchronization codes and the watermarks. Based on this formula, the number of DWT coefficients for embedding a binary bit can be increased to enhance the robustness without decreasing the audio quality. Consequently, the embedded audio has good quality and good robustness under high capacity. In addition, the system can extract the hidden data without the knowledge of original audio signal. Finally, the performance of the proposed watermarking method is tested. Experimental results indicate that the performance of proposed system is better than other DWT amplitude modification method.

Biometric Electrocardiogram Card for Access Control System

With the rapid development of access control system based on biometric technologies, the conventional identification exposed more and more weakness while they are easy to be counterfeited and imitated. In this paper, we bring the access control systems with electrocardiogram identification into a practical application. Moreover, a new prompt based ECG algorithm are proposed to seek a more secure and accurate identification. In our evaluation, a hardware board were designed to verified its feasibility, then the analysis and comparison among various biometric identification and previous ECG identification researches were further disused as well. The result shown our proposed design could provide a more secure, low cost and convenient identification for access control system.

A PSO- SVM Lips Recognition Method Based on Active Basis Model

The paper proposed a novel method for lip recognition based on Active Basis Model (ABM). There are four stages in a flowchart of this novel method. At the first stage the deformable templates of lip images is obtained. The lip images of deformable templates are obvious open or closed. The second stage is to obtain the deformed template of each testing images. The third stage, the difference between the deformable template and deformed template is calculated and used as a feature vector. Finally, the support vector machine (SVM) is use to classify the feature vector. SVM is a supervised learning to be a classifier for lip recognition. It is necessary to set and parameters, which are the two factors affecting quality of the model, when establishing SVM method. Hence, in terms of parameters selections, the researcher adopted PSO (Particle Swarm Optimization, PSO) algorithm in this research to set the best parameters combination and then incorporated into the SVM to obtain the classified results. In this paper, the novel method which utilizes PSO algorithm to select the parameters and automatically is called PSO-SVM method. There are 1000 face images in BioID face database used for the experiment. The experimental results show that PSO-SVM method can be a more accurate model to recognize the lip images.

Computer-assisted system with multiple feature fused support vector machine for sperm morphology diagnosis

Sperm morphology is an important technique in identifying the health of sperms. In this paper we present a new system and novel approaches to classify different kinds of sperm images in order to assess their health. Our approach mainly relies on a one-dimensional feature which is extracted from the sperms contour with gray level information. Our approach can handle rotation and scaling of the image. Moreover, it is fused with SVM classification to improve its accuracy. In our evaluation, our method has better performance than the existing approaches to sperm classification.

QRS Complex Detection Using Cubic Splines

The electrocardiogram is widely used for diagnosis of heart diseases. Several methods have been developed for ECG enhancement. This work presents a new ECG enhancement method for QRS complex detection. The moving average and cubic splines are successfully applied to detect the QRS complex through experiments on the MIT-BIH databases. The results show that the proposed method achieves good detection performances.

Optimization-Based Audio Watermarking Using Low-Frequency Amplitude Modification

Based on Karush-Kuhn-Tucker Theorem, a new blind digital audio watermarking scheme is proposed. In order to guarantee the robustness of watermarks, this scheme embeds information into low-frequency coefficients of discrete wavelet transform. For the modification of low-frequency amplitude, this work uses Karush-Kuhn-Tucker Theorem to minimize the difference between the original and the modified coefficients. Consequently, the watermarked audio has strong robustness under sufficient watermarked-audio quality and high embedding capacity. In addition, the system can extract the hidden data without the knowledge of original audio signal. Experimental results indicate that the performance of proposed system is mostly better than other amplitude modification method.

Symmetry Auto-detection Based on Contour and Corner Models

In this paper we present a novel symmetry auto-detection approach which is used for the symmetry detection of the objects in images. Rather than relying on gradients and pixel matching of an image, symmetry can be automatically detected by using corner detection on a symmetry function which is proposed in Hsus paper, and a result will be shown to demonstrate the probability that the object in the image is bilateral symmetry. In this paper, we first use modified active contour model to get the contour of the object in an image. Then we use a symmetries evaluation function to get the symmetry function of the contour. Last, we detect corner in the proposed function. By matching corner points, the symmetry function shows us the symmetry-probability of the object. We apply our method in several natural or man-made objects, and they all exhibit significant performance.

Wavelet-Based Entropy for Digital Audio Watermarking

Unlike traditional entropy in information theory, this work uses the normalized energy instead of probability to obtain a low-frequency amplitude transform (LAT) on coefficients of discrete wavelet transform (DWT). The watermark is embedded based on the properties and characteristics of this transform. Finally, performance of the proposed scheme is assessed by signal-to-watermark (SWR) and bit error rate (BER). Experimental results demonstrate that the embedded data are robust against most signal processing and attacks, such as re-sampling, low-pass filtering, and amplitude-scaling.

Corner Detection Based on Normal Vector of Boundary Fitting Line

this paper shows a novel and low complexity approach for corner detection which is based on a normal vector of boundary fitting line. It avoids wrong detection of superfluous corners on no-corner arcs. Our proposed method is superior to Sun’s k-cosine corner detection in detection time and has a better performance in localization. Our experiment results confirmed that the proposed approach of corner detection has reached our goal. It is free from rotation and able to locate the corner correctly. In addition, it also performs well for scaling images with the adjustable thresholds.