• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2003.05a
• /
• pp.165-170
• /
• 2003

This paper presents an integrated approach to unconstrained face recognition in arbitrary scenes. The front end of the system comprises of a scale and pose tolerant face detector. Scale normalization is achieved through novel combination of a skin color segmentation and log-polar mapping procedure. Principal component analysis is used with the multi-view approach proposed in[10] to handle the pose variations. For a given color input image, the detector encloses a face in a complex scene within a circular boundary and indicates the position of the nose. Next, for recognition, a radial grid mapping centered on the nose yields a feature vector within the circular boundary. As the width of the color segmented region provides an estimated size for the face, the extracted feature vector is scale normalized by the estimated size. The feature vector is input to a trained neural network classifier for face identification. The system was evaluated using a database of 20 person's faces with varying scale and pose obtained on different complex backgrounds. The performance of the face recognizer was also quite good except for sensitivity to small scale face images. The integrated system achieved average recognition rates of 87% to 92%.

• Journal of Korea Multimedia Society
• /
• v.21 no.8
• /
• pp.837-848
• /
• 2018

Monitoring technology of machining has a long history since unmanned machining was introduced. Despite the long history, many researchers have presented new approaches continuously in this area. Sound based machine fault diagnosis is the process consisting of detecting automatically the damages that affect the machines by analyzing the sounds they produce during their operating time. The collected sound is corrupted by the surrounding work environment. Therefore, the most important part of the diagnosis is to find hidden elements inside the data that can represent the error pattern. This paper presents a feature extraction methodology that combines various digital signal processing and pattern recognition methods for the analysis of the sounds produced by tools. The magnitude spectrum of the sound is extracted using the Fourier analysis and the band-pass filter is applied to further characterize the data. Statistical functions are also used as input to the nonlinear classifier for the final response. The results prove that the proposed feature extraction method accurately captures the hidden patterns of the sound generated by the tool, unlike the conventional features. Therefore, it is shown that the proposed method can be applied to a sound based automatic diagnosis system.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.6
• /
• pp.1312-1317
• /
• 2003

This paper presents an integrated approach to unconstrained face recognition in arbitrary scenes. The front end of the system comprises of a scale and pose tolerant face detector. Scale normalization is achieved through novel combination of a skin color segmentation and log-polar mapping procedure. Principal component analysis is used with the multi-view approach proposed in[10] to handle the pose variations. For a given color input image, the detector encloses a face in a complex scene within a circular boundary and indicates the position of the nose. Next, for recognition, a radial grid mapping centered on the nose yields a feature vector within the circular boundary. As the width of the color segmented region provides an estimated size for the face, the extracted feature vector is scale normalized by the estimated size. The feature vector is input to a trained neural network classifier for face identification. The system was evaluated using a database of 20 person's faces with varying scale and pose obtained on different complex backgrounds. The performance of the face recognizer was also quite good except for sensitivity to small scale face images. The integrated system achieved average recognition rates of 87% to 92%.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.47 no.5
• /
• pp.62-68
• /
• 2010

This study is to identify target locations with low false alarms on thermal infrared images obtained from natural environment. The proposed method is different from the previous researches because it uses morphology filters for Gabor response images instead of an intensity image in initial detection stage. This method does not need precise extracting a target silhouette to distinguish true targets or clutters. It comprises three distinct stages. First, morphological operations and adaptive thresholding are applied to the summation image of four Gabor responses of an input image to find out salient regions. The locations of extracted regions can be classified into targets or clutters. Second, local texture features are computed from salient regions of an input image. Finally, the local texture features are compared with the training data to distinguish between true targets and clutters. The multi-layer perceptron having three layers is used as a classifier. The performance of the proposed method is proved by using natural infrared images. Therefore it can be applied to real automatic target detection systems.

• IEIE Transactions on Smart Processing and Computing
• /
• v.5 no.4
• /
• pp.256-266
• /
• 2016

Concurrently detected and annotated abnormal events can have a significant impact on surveillance systems. By considering the specific domain of pedestrian trajectories, this paper presents two main contributions. First, as introduced in much of the work on trajectory-based anomaly detection in the literature, only information about pedestrian paths, such as direction and speed, is considered. Differing from previous work, this paper proposes a framework that deals with additional types of trajectory-based anomalies. These abnormal events take places when a person enters prohibited areas. Those restricted regions are constructed by an online learning algorithm that uses surrounding information, including detected pedestrians and background scenes. Second, a simple data-boosting technique is introduced to overcome a lack of training data; such a problem particularly challenges all previous work, owing to the significantly low frequency of abnormal events. This technique only requires normal trajectories and fundamental information about scenes to increase the amount of training data for both normal and abnormal trajectories. With the increased amount of training data, the conventional abnormal trajectory classifier is able to achieve better prediction accuracy without falling into the over-fitting problem caused by complex learning models. Finally, the proposed framework (which annotates tracks that enter prohibited areas) and a conventional abnormal trajectory detector (using the data-boosting technique) are integrated to form a united detector. Such a detector deals with different types of anomalous trajectories in a hierarchical order. The experimental results show that all proposed detectors can effectively detect anomalous trajectories in the test phase.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.6
• /
• pp.162-169
• /
• 2008

Film restoration is to detect the location and extent of defected regions from a given movie film, and if present, to reconstruct the lost information of each regions. It has gained increasing attention by many researchers, to support multimedia service of high quality. Among artifacts, scratch is the most frequent degradation. In this paper, an automatic film line scratch removal system is developed that can detect and restore all kind of scratches. For this we use the spatial information of scratches: The scratch in old films has lower or higher brightness than neighboring pixels in its vicinity and usually appears as a vertically long thin line. Our systems consists of scratch detection and scratch restoration. The scratches of various types are detected by neural network based texture classifier and morphology-based shape filter and then the degraded regions are restored using bilinear interpolation. To assess the validity of the Proposed method, it has been tested with all kinds of scratches, and then experimental results show that the proposed approach is robust to various scratches and efficient to apply a real film removal system.

• Journal of Korea Multimedia Society
• /
• v.6 no.5
• /
• pp.788-796
• /
• 2003

This paper proposes using dynamic momentum for squential learning method. Using The dynamic momentum improves convergence speed and performance by the variable momentum, also can identify it in the RMSE(root mean squared error). The proposed method is reflected using variable momentum according to current state. While static momentum is equally influenced on the whole, dynamic momentum algorithm can control the convergence rate and performance. According to the variable change of momentum by training. Unlike former classification and regression problems, this paper confirms both performance and regression rate of the dynamic momentum. Using RMSE(root mean square error ), which is one of the regression methods. The proposed dynamic momentum has been applied to the kernel adatron and kernel relaxation as the new sequential learning method of support vector machine presented recently. In order to show the efficiency of the proposed algorithm, SONAR data, the neural network classifier standard evaluation data, are used. The simulation result using the dynamic momentum has a better convergence rate, performance and RMSE than those using the static moment, respectively.

• Journal of Korea Multimedia Society
• /
• v.21 no.4
• /
• pp.450-463
• /
• 2018

Sound based machine fault diagnosis is the process consisting of detecting automatically the damages that affect the machines by analyzing the sounds they produce during their operating time. The collected sounds being inevitably corrupted by random disturbance, the most important part of the diagnosis consists of discovering the hidden elements inside the data that can reveal the faulty patterns. This paper presents a novel feature extraction methodology that combines various digital signal processing and pattern recognition methods for the analysis of the sounds produced by the drills. Using the Fourier analysis, the magnitude spectrum of the sounds are extracted, converted into two-dimensional vectors and uniformly normalized in such a way that they can be represented as 8-bit grayscale images. Histogram equalization is then performed over the obtained images in order to adjust their very poor contrast. The obtained contrast enhanced images will be used as the features of our diagnosis system. Finally, principal component analysis is performed over the image features for reducing their dimensions and a nonlinear classifier is adopted to produce the final response. Unlike the conventional features, the results demonstrate that the proposed feature extraction method manages to capture the hidden health patterns of the sound.

• Journal of IKEEE
• /
• v.24 no.3
• /
• pp.777-782
• /
• 2020

In this paper, we propose CNN-based HMI system using Doppler radar and voice sensor, and present hardware design and implementation results. To overcome the limitation of single sensor monitoring, the proposed HMI system combines data from two sensors to improve performance. The proposed system exhibits improved performance by 3.5% and 12% compared to a single radar and voice sensor-based classifier in noisy environment. In addition, hardware to accelerate the complex computational unit of CNN is implemented and verified on the FPGA test system. As a result of performance evaluation, the proposed HMI acceleration platform can be processed with 95% reduction in computation time compared to a single software-based design.

• Journal of the Institute of Convergence Signal Processing
• /
• v.20 no.4
• /
• pp.205-211
• /
• 2019

Steganalysis is a technique that aims to detect and recover data hidden by steganography. Steganalytic methods detect hidden data by analyzing visual and statistical distortions caused during data embedding. However, for recovering the hidden data, they need to know which steganographic methods the hidden data has been embedded by. Therefore, we propose a hierarchical convolutional neural network (CNN) structure that identifies a steganographic method applied to an input image through multi-level classification. We trained four base CNNs (each is a binary classifier that determines whether or not a steganographic method has been applied to an input image or which of two different steganographic methods has been applied to an input image) and connected them hierarchically. Experimental results demonstrate that the proposed hierarchical CNN structure can identify four different steganographic methods (LSB, PVD, WOW, and UNIWARD) with an accuracy of 79%.