• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.10
• /
• pp.2121-2128
• /
• 2012

This paper proposed the real time face area detection using Wavelet transform and the strong detection algorithm that satisfies the efficiency of computation and detection performance at the same time was proposed. The detected face image recognizes the face by configuring the low-dimensional face symbol through the principal component analysis. The proposed method is well suited for real-time system construction because it doesn't require a lot of computation compared to the existing geometric feature-based method or appearance-based method and it can maintain high recognition rate using the minimum amount of information. In addition, in order to reduce the wrong recognition or recognition error occurred during face recognition, the input symbol of Hidden Markov Model is used by configuring the feature values projected to the unique space as a certain symbol through clustering algorithm. By doing so, any input face will be recognized as a face model that has the highest probability. As a result of experiment, when comparing the existing method Euclidean and Mahananobis, the proposed method showed superior recognition performance in incorrect matching or matching error.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.8
• /
• pp.1537-1545
• /
• 2016

In real time video sequence, object segmentation and tracking method are actively applied in various application tasks, such as surveillance system, mobile robots, augmented reality. This paper propose a robust object tracking method. The background models are constructed by learning the initial part of each video sequences. After that, the moving objects are detected via object segmentation by using background subtraction method. The region of detected objects are continuously tracked by using the HSV color histogram with particle filter. The proposed segmentation method is superior to average background model in term of moving object detection. In addition, the proposed tracking method provide a continuous tracking result even in the case that multiple objects are existed with similar color, and severe occlusion are occurred with multiple objects. The experiment results provided with 85.9 % of average object overlapping rate and 96.3% of average object tracking rate using two video sequences.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.05a
• /
• pp.801-804
• /
• 2013

We propose a new tongue examination model according to the taste division of tongue. The proposed sytem consists of image acquisition, region segmentation, color distribution analysis and abnormality decision of tongue. Tongue DB which is classified into abnormality is constructed with tongue images captured from oriental medicine hospital inpatients. We divided 4 basic taste(bitter, sweet, salty and sour) regions and performed color distribution analysis targeting each region under HSI(Hue Saturation Intensity) color model. To minimize the influence of illumination, the histograms of H and S components only except I are utilized. The abnormality of taste regions each by comparing the proposed diagnosis model with diagnosis results by a doctor of oriental medicine. We confirmed the 87.5% of classification results of abnormality by proposed algorithm is coincide with the doctor's results.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• v.9 no.1
• /
• pp.1149-1152
• /
• 2005

Detecting faces in images with complex backgrounds is a difficult task. Our approach, which obtains state of the art results, is based on a generative neural network model: the Constrained Generative Model (CGM). To detect side view faces and to decrease the number of false alarms, a conditional mixture of networks is used. To decrease the computational time cost, a fast search algorithm is proposed. The level of performance reached, in terms of detection accuracy and processing time, allows to apply this detector to a real word application: the indexation of face images on the Web.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.7
• /
• pp.319-324
• /
• 2006

We Propose two new methods for non-keyword modeling to improve the performance of speaker- and vocabulary-independent keyword spotting system. The first method is decision tree clustering of monophone at the state level instead of monophone clustering method based on K-means algorithm. The second method is multi-state multiple mixture modeling at the syllable level rather than single state multiple mixture model for the non-keyword. To evaluate our method, we used the ETRI speech DB for training and keyword spotting test (closed test) . We also conduct an open test to spot 100 keywords with 400 sentences uttered by 4 speakers in an of fce environment. The experimental results showed that the decision tree-based state clustering method improve 28%/29% (closed/open test) than the monophone clustering method based K-means algorithm in keyword spotting. And multi-state non-keyword modeling at the syllable level improve 22%/2% (closed/open test) than single state model for the non-keyword. These results show that two proposed methods achieve the improvement of keyword spotting performance.

• Journal of Korea Multimedia Society
• /
• v.8 no.3
• /
• pp.336-344
• /
• 2005

In this paper, we extracted the contour of lung parenchyma on EBT images with the improved active contour model. The objects boundary in conventional active contour model can be extracted by controlling internal energy and external energy as energy minimizing form. However, there are a number of problems such as initialization and the poor convergence about concave part. Expecially, contour can not enter the concave region by discouraging characteristic about stretching and bending in internal energy. We controlled internal energy by moving local perpendicular bisector point of each control point in the contour and implemented the object boundary by minimizing energy with external energy The convergence of concave part could be efficiently implemented toward lung parenchyma region by this internal energy and both lung images for initial contour could also be detected by multi-detection method. We were sure this method could be applied detection of lung parenchyma region in medical image.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.5
• /
• pp.85-89
• /
• 2017

In this papers, we apply high performance hardware and machine learning algorithm to build an advanced VAD algorithm for speech enhancement. Since speech is made of series of phoneme, using recurrent neural network (RNN) which consider previous data is proper method to build a speech model. It is impossible to study every noise in real world. So our algorithm is builded by phoneme based study. we detect voice present frames in noisy speech signal and make enhancement of the speech signal. Phoneme based RNN model shows advanced performance in speech signal which has high correlation among each frames. To verify the performance of proposed algorithm, we compare VAD result with label data and speech enhancement result in various noise environments with previous speech enhancement algorithm.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.10
• /
• pp.419-426
• /
• 2022

In this paper, we propose a method to improve the accuracy of 3D human pose estimation model in various move motions. Existing human pose estimation models have some problems of jitter, inversion, swap, miss that cause miss coordinates when estimating human poses. These problems cause low accuracy of pose estimation models to detect exact coordinates of human poses. We propose a method that consists of detection and correction methods to handle with these problems. Deep learning-based outlier detection method detects outlier of human pose coordinates in move motion effectively and rule-based correction method corrects the outlier according to a simple rule. We have shown that the proposed method is effective in various motions with the experiments using 2D golf swing motion data and have shown the possibility of expansion from 2D to 3D coordinates.

• Journal of Aerospace System Engineering
• /
• v.17 no.6
• /
• pp.100-107
• /
• 2023

Open-phase fault in a Permanent Magnet Synchronous Motor (PMSM) occurs due to disconnection of phases of motor windings or inverter switch failures. When an open-phase occurs, it leads to the generation of torque ripples and vibrations in the motor, which can have a critical impact on the safety of the vehicle (including aircraft) using a PMSM as an actuator. Therefore, rapid fault detection and classification are essential. This paper proposes a classification method for detecting open-phase faults and locating fault positions in a PMSM used in aircraft applications. The proposed approach uses an Extended Kalman Filter for fault diagnosis, and it subsequently classifies faults using a Multiple Model filter.

• Journal of KIISE:Software and Applications
• /
• v.31 no.3
• /
• pp.379-386
• /
• 2004

Video segmentation is a fundamental task in video indexing and it includes two kinds of shot change detections such as the abrupt transition and the gradual transition. The abrupt shot boundaries are detected by computing the image-based distance between adjacent frames and comparing this distance with a pre-determined threshold value. However, the gradual shot boundaries are difficult to detect with this approach. To overcome this difficulty, we propose the method that detects gradual transition in the MPEG compressed video using the HMM (Hidden Markov Model). We take two different HMMs such as a discrete HMM and a continuous HMM with a Gaussian mixture model. As image features for HMM's observations, we use two distinct features such as the difference of histogram of DC images between two adjacent frames and the difference of each individual macroblock's deviations at the corresponding macroblock's between two adjacent frames, where deviation means an arithmetic difference of each macroblock's DC value from the mean of DC values in the given frame. Furthermore, we obtain the DC sequences of P and B frame by the first order approximation for a fast and effective computation. Experiment results show that we obtain the best detection and classification performance of gradual transitions when a continuous HMM with one Gaussian model is taken and two image features are used together.