• KIPS Transactions on Software and Data Engineering
• /
• v.3 no.9
• /
• pp.369-374
• /
• 2014

Path generation methods are required for safe and efficient driving in unmanned autonomous vehicles. There are two kinds of paths: global and local. A global path consists of all the way points including the source and the destination. A local path is the trajectory that a vehicle needs to follow from a way point to the next in the global path. In this paper, we propose a novel method for local path generation through machine learning, with an effective curve function used for initializing the trajectory. First, reinforcement learning is applied to a set of candidate paths to produce the best trajectory with maximal reward. Then the optimal steering angle with respect to the trajectory is determined by training an artificial neural network. Our method outperformed existing approaches and successfully found quality paths in various experimental settings, including the cases with obstacles.

• The Journal of Bigdata
• /
• v.5 no.1
• /
• pp.89-96
• /
• 2020

Deep learning using an artificial neural network has been recently researched and developed in various fields such as image recognition, big data and data analysis. However, federated learning has emerged to solve issues of data privacy invasion and problems that increase the cost and time required to learn. Federated learning presented learning techniques that would bring the benefits of distributed processing system while solving the problems of existing deep learning, but there were still problems with server-client system and motivations for providing learning data. So, we replaced the role of the server with a blockchain system in federated learning, and conducted research to solve the privacy and security problems that are associated with federated learning. In addition, we have implemented a blockchain-based system that motivates users by paying compensation for data provided by users, and requires less maintenance costs while maintaining the same accuracy as existing learning. In this paper, we present the experimental results to show the validity of the blockchain-based system, and compare the results of the existing federated learning with the blockchain-based federated learning. In addition, as a future study, we ended the thesis by presenting solutions to security problems and applicable business fields.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.5
• /
• pp.110-116
• /
• 2008

This paper presents a new postprocessing method to eliminate the false minutiae, that caused by the skelectonization of fingerprint image, and an image compression method using Isometric Self Organizing Map(ISOSOM). Since the SOM has simple structure, fast encoding time, and relatively good classification characteristics, many image processing areas adopt this such as image compression and pattern classification, etc. But, the SOM shows limited performances in pattern classification because of it's single layer structure. To maximize the performance of the pattern classification with small code book, we a lied the Isometric SOM with the isometry of the fractal theory. The proposed Isometric SOM postprocessing and compression algorithm of fingerprint image showed good performances in the elimination of false minutiae and the image compression simultaneously.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.11 no.4
• /
• pp.91-97
• /
• 2011

In this paper we present an approach to the structure identification based on genetic algorithm and to the parameter identification by hybrid learning method in neuro-fuzzy-genetic hybrid system in order to predicate the Mackey-Glass Chaotic time series. In this scheme the basic idea consists of two steps. One is the construction of a fuzzy rule base for the partitioned input space via genetic algorithm, the other is the corresponding parameters of the fuzzy control rules adapted by the backpropagation algorithm. In an attempt to test the performance the proposed system, three patterns, x(t-3), x(t-6) and x(t-9), was prepared according to time interval. It was through lots of simulation proved that the initial small error of learning owed to the good structural identification via genetic algorithm. The performance was showed in Table 2.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.4
• /
• pp.286-291
• /
• 2020

In this paper, we propose a system that recommends music through tempo-oriented music classification and sensor-based human activity recognition. The proposed method indexes music files using tempo-oriented music classification and recommends suitable music according to the recognized user's activity. For accurate music classification, a dynamic classification based on a modulation spectrum and a sequence classification based on a Mel-spectrogram are used in combination. In addition, simple accelerometer and gyroscope sensor data of the smartphone are applied to deep spiking neural networks to improve activity recognition performance. Finally, music recommendation is performed through a mapping table considering the relationship between the recognized activity and the indexed music file. The experimental results show that the proposed system is suitable for use in any practical mobile device with a music player.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.4
• /
• pp.263-269
• /
• 2020

In this study, a rotating machine that can reproduce normal condition and 8 fault conditions were produced, and vibration data was acquired. Feature is calculated from the acquired data, and accuracy is analyzed through fault diagnosis using artificial neural networks and genetic algorithms. In order to achieve optimal timing and higher accuracy, features by three domains were applied to the fault diagnosis. The learning number was selected as a setting variable. As a result of the rotating machine fault diagnosis, high precision was found in the frequency domain than in others, and precise fault diagnoses were accomplished through all of 10 operations, at the learning number of 5000 and 8000. Given the efficiency of time, it was estimated to be the most efficient when the number of learning was 5000.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.12 no.2
• /
• pp.137-142
• /
• 2012

The temporal subtraction technique as one of computer aided diagnosis has been introduced in medical fields to enhance the interval changes such as formation of new lesions and changes in existing abnormalities on deference image. With the temporal subtraction technique radiologists can easily detect lung nodules on visual screening. Until now, two-dimensional temporal subtraction imaging technique has been introduced for the clinical test. We have developed new temporal subtraction method to remove the subtraction artifacts which is caused by mis-registration on temporal subtraction images of lungs on MDCT images. In this paper, we propose a new computer aided diagnosis scheme for automatic enhancing the lung nodules from the temporal subtraction of thoracic MDCT images. At first, the candidates regions included nodules are detected by the multiple threshold technique in terms of the pixel value on the temporal subtraction images. Then, a rule-base method and artificial neural networks is utilized to remove the false positives of nodule candidates which is obtained temporal subtraction images. We have applied our detection of lung nodules to 30 thoracic MDCT image sets including lung nodules. With the detection method, satisfactory experimental results are obtained. Some experimental results are shown with discussion.

• Journal of the Korea Convergence Society
• /
• v.11 no.3
• /
• pp.181-186
• /
• 2020

The purpose of this study is to empirically examine the factors affecting the binge-watching behaviors of OTT service users by using a multi-layer perceptron (MLP) artificial neural network. All samples (n=1,000) were collected from 'A survey on user awareness in OTT service' published by a Media Research Center of the Korea Press Foundation in 2018. Our research model includes one dependent variable which is binge-watching behaviors on OTT service and five independent variables such as gender, age, frequency of service usage, users' satisfaction with content recommendation algorithm, and content types mainly consumed. Our findings demonstrate that age, frequency of service usage, users' satisfaction with content recommendation algorithms, and certain types of contents (e.g., Korean dramas, Korean films, and foreign dramas) were found to be highly related to binge-watching behavior on OTT services.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2008.08a
• /
• pp.155-159
• /
• 2008

In this thesis, in order to develop a new classification model of Sasang Constitutional medical types, which is helpful for improving the accuracy of diagnosis of medical types. various data-mining classification models such as discriminant analysis. decision trees analysis, neural networks analysis, logistics regression analysis, clustering analysis which are main classification methods were applied to the questionnaires of medical type classification. In this manner, a model which scientifically classifies constitutional medical types in the field of Sasang Constitutional Medicine, one of a traditional Korean medicine, has been developed. Also, the above-mentioned analysis models were systematically compared and analyzed. In this study, a classification of Sasang constitutional medical types was developed based on the discriminate analysis model and decision trees analysis model of which accuracy is relatively high, of which analysis procedure is easy to understand and to explain and which are easy to implement. Also, a diagnosis system of Sasang constitution was implemented applying the two analysis models.

• Journal of the Institute of Convergence Signal Processing
• /
• v.2 no.1
• /
• pp.1-11
• /
• 2001

This paper presents a learning-based approach for the construction of license Plate recognition system. The system consist of three modules. They are respectively, car detection module, license plate recognition module and recognition module. Car detection module detects a car in the given image sequence obtained from the camera with simple color-based approach. Segmentation module extracts the license plate in detect car image using neural network as filters for analyzing the color and texture properties of license plate. Recognition module then reads characters in detected license plate with support vector machine (SVM)-based characters recognizer. The system has been tested from parking lot and tollgate, etc. and have show the following performances on average: Car detect rate 100%, segmentation rate 97.5%, and character recognition rate about 97.2%. Overall system performances is 94.7% and processing time is one sec. Then our propose system does well using real world.