Jang-Woo Kwon

Face detection and recognition using PCA

In this paper we developed a computer system that can locate and track a subjects head in a complex background and then recognize the person by comparing characteristics of the face to those of known individuals. The computational approach taken in this system is motivated by color and motion Information and PCA (principal component analysis). Our approach treats the face recognition problem as a two-dimensional (2-D) problem rather than three-dimensional geometry. So, the problem is easier to treat. The system functions by two steps, first, extracting face image in a complex background using difference image and color model, and second, projecting pre-extracted face images onto a feature space that represents the significant variations among known face images. We use this weight vector to recognize each individual. Several evaluation methods of this weight vector are attempted in this paper.

Signal hybrid HMM-GA-MLP classifier for continuous EMG classification purpose

This paper describes an approach for classifying electromyographic (EMG) signals using a multilayer perceptron (MLP) with genetic algorithm (GA) and hidden Markov models (HMMs) hybrid classifier. Instead of using MLP as probability generators for HMMs we propose to use MLP with GA as the second classifiers to increase discrimination rates of myoelectric patterns. The GA for MLP was driven to boost the learning time when it applied to backpropagation (BP) algorithm. This strategy is proposed to overcome weak discrimination and to consider dynamic properties of EMG signals. Four discrimination strategies (HMM-MLP, HMM-GA-MLP, HMM-counter propagation network (CPN), and HMM-GA-CPN) for discriminating signals representative of 6 primitive class of motions are described and compared. The proposed strategy increase the discrimination results considerably. Results are presented to support this approach.

Development of neural network based plasma monitoring system for laser welding quality analysis

Neural networks are shown to be effective in being able to distinguish incomplete penetration-like weld defects by directly analyzing the plasma which is generated on each impingement of the laser on the materials. The performance is similar to that of existing methods based on extracted feature parameters. In each case around 93% of the defects in a database derived from 100 artificially produced defects of known types can be placed into one of two classes: incomplete penetration and bubbling. The present method based on classification using plasma is faster, and the speed is sufficient to allow on-line classification during data collection.

A Study on EEG based Concentration Power Index Transmission and Brain Computer Interface Application

This research measures EEG signals which are generating on the head skin and extracts brain concentration level related with brain activity. We have developed wireless transmission system of concentration power index for controlling hardware by using this signal. Two channel electrodes was used for measuring EEG signal on front head and Biopac system with EEG100C was also used for measuring EEG signal, amplifying and filtering the signal. LabVIEW 8.5 was used for FFT transformation,frequency and spectrum analysis of the measured EEG signal. As a result, several kinds of brain waves,such as α waves, β waves, θ waves and δ have been classified power index by adapting concentration extraction algorithm. This concentration power index was transferred into an automobile Lego device by wireless module for BCI application test.

Remote monitoring based on the nonpersonal computer with electrocardiogram

Remote monitoring for medical diagnosis is widespread from personal computer to wireless communication nowadays. In this paper we propose an ECG (electrocardiogram) monitoring system which is based on the nonpersonal computer of a subject. Using the network device of small size, the detected biomedical signals are transmitted from a subject to a medical doctor. The designed system are consisted of three parts which are biomedical signals acquisition part, network device having only one IP address and the intermediate AVR-microcontroller unit, which is placed between two parts. In programming, some observer can monitor the cardiogram of a subject, if they know the IP address of a subject, which is displayed by the applet of the Java program. Consequently, with this small size and freedom from the restriction of the place can enable a subjects biomedical signals to be monitored easily from the remote area through the Ethernet.

Hybrid hidden Markov model-neural network system for EMG signals recognition

Describes an approach for classifying electromyographic (EMG) signals using a multilayer perceptrons (MLPs) and hidden Markov models (HMMs) hybrid classifier. Instead of using MLPs as probability generators for HMMs the authors propose to use MLPs as the second classifiers to increase discrimination rates of myoelectric patterns. This strategy is proposed to overcome weak discrimination and to consider dynamic properties of EMG signals. Two discrimination strategies (HMM, and HMM with three subnet MLPs) for discriminating signals representative of 6 primitive class of motions are described and compared. The proposed strategy increase the discrimination results considerably. Results are presented to support this approach.

Classification of the EMG Signal using Cascaded Classifier

This paper describes an approach for classifying electromyographic (EMG) signals using the multilayer perceptrons (MLP’s) and the hidden markov models (HMM’s). In this paper, we propose the combination of MLP’s and HMM’s cascaded structure to increase discrimination rates of myoelectric patterns. Feature sets based upon the short-time Fourier transform, the wavelet transform, and the wavelet packet transform for discriminating signals represent active of motions are applied and compared. This strategy is proposed to overcome weak discrimination and to consider dynamic properties of EMG signlas.

Development of leather quality discrimination system by texture analysis

The process of a leather quality decision-making process with unaided eyes can be unreliable because of the lack of consistency caused by tiredness due to continued exposure. Therefore, an objective standard for considering the quality of leather, and based on this standard, the automation of a quality decision-making process are required The automatic discrimination system of leather which is applied in this dissertation consists of the process of achieving information about leather and evaluating grades from the former process. The quality of leather depends on the chosen density extraction. In this dissertation, an algorithm which decides the grade of leather by extracting density and defects from the black image taken by a digital camera is suggested. Density is calculated by the area, width and height of the area exited in Fourier spectrum. Also, defects are extracted by the distinction of histogram distribution of pixel considered as a window from the image proceeded the pre-operation process using a searching window. The distinctions of whole leather are used as the standard for evaluating the grade and it is considered that it can substitute the eye sight test in other fields.

AdaBoost-based Real-Time Face Detection & Tracking System

This paper presents a method for real-time face detection and tracking which combined Adaboost and Camshift algorithm. Adaboost algorithm is a method which selects an important feature called weak classifier among many possible image features by tuning weight of each feature from learning candidates. Even though excellent performance extracting the object, computing time of the algorithm is very high with window size of multi-scale to search image region. So direct application of the method is not easy for real-time tasks such as multi-task OS, robot, and mobile environment. But CAMshift method is an improvement of Mean-shift algorithm for the video streaming environment and track the interesting object at high speed based on hue value of the target region. The detection efficiency of the method is not good for environment of dynamic illumination. We propose a combined method of Adaboost and CAMshift to improve the computing speed with good face detection performance. The method was proved for real image sequences including single and more faces.

A study on the development of gas measurement system in shoes mold and automatic gas-vent exchange machine with computer vision

This paper presents a gas measurement system for deciding hole positions on a PU middle-sole mold from computed gas amount. The optimal number of holes and their positions on the shoe mold are decided from statistical experiment results to overcome the problem of excessive expenses in gas vent exchange. This paper also describes a gas vent exchange mechanism using computer vision system. The gas hole detecting process is based on computer vision algorithms represented as a simple Pattern Matching. The experimental result showed us that the system was useful to calculate the number of holes and their positions on the shoes mold.