• The Journal of the Korea Contents Association
• /
• v.5 no.5
• /
• pp.33-42
• /
• 2005

In this paper, we propose a new training algorithm for improving pattern classification performance of ART neural network. The proposed train algorithm restricts unnecessary cluster generation and transition, applies the location extraction algorithm, and operates the reset system based on the agreement between the present learning pattern and the initial pattern. As a result, repetitive input of a pattern does not generate a new cluster and mis-recognition rate decreases.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06b
• /
• pp.321-323
• /
• 2012

스마트폰 사용의 보편화와 센서기술의 발달로 이를 응용하는 다양한 연구가 진행되고 있다. 특히 가속도, GPS, 조도, 방향센서 등의 센서들이 스마트폰에 부착되어 출시되고 있어서, 이를 이용한 상황인지, 행동인식 등의 관련 연구들이 활발하다. 하지만 다양한 클래스를 분류하면서 높은 인식률을 유지하는 것은 어려운 문제이다. 본 논문에서는 인식률 향상을 위해 계층적 구조의 순환 신경망을 이용하여 제스처를 인식한다. 스마트폰의 가속도 센서를 이용하여 사용자의 제스처 데이터를 수집하고 BLSTM(Bidirectional Long Short-Term Memory) 구조의 순환신경망을 계층적으로 사용하여, 20가지 사용자의 제스처와 비제스처를 분류한다. 약 24,850개의 시퀀스 데이터를 사용하여 실험한 결과, 기존 BLSTM은 평균 89.17%의 인식률을 기록한 반면 계층적 BLSTM은 평균 91.11%의 인식률을 나타내었다.

• Journal of information and communication convergence engineering
• /
• v.2 no.2
• /
• pp.132-137
• /
• 2004

Most of the current computer vision theories are based on hypotheses that are difficult to apply to the real world, and they simply imitate a coarse form of the human visual system. As a result, they have not been showing satisfying results. In the human visual system, there is a mechanism that processes information due to memory degradation with time and limited storage space. Starting from research on the human visual system, this study analyzes a mechanism that processes input information when information is transferred from the retina to ganglion cells. In this study, a model for the characteristics of ganglion cells in the retina is proposed after considering the structure of the retina and the efficiency of storage space. The MNIST database of handwritten letters is used as data for this research, and ART2 and SOM as recognizers. The results of this study show that the proposed recognition model is not much different from the general recognition model in terms of recognition rate, but the efficiency of storage space can be improved by constructing a mechanism that processes input information.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.10a
• /
• pp.613-616
• /
• 2018

Sequence tagging is an important task in Natural Language Processing (NLP), in which the Named Entity Recognition (NER) is the key issue. So far the most widely adopted model for NER in NLP is that of combining the neural network of bidirectional long short-term memory (BiLSTM) and the statistical sequence prediction method of Conditional Random Field (CRF). In this work, we improve the prediction accuracy of the BiLSTM by supporting an aligned word representation mechanism. We have performed experiments on multilingual (English, Spanish and Dutch) datasets and confirmed that our proposed model outperformed the existing state-of-the-art models.

• Journal of Institute of Control, Robotics and Systems
• /
• v.3 no.5
• /
• pp.511-519
• /
• 1997

This paper presents the automatic recognition algorithm of the license number in on vehicle image. The proposed algorithm uses the correlation coefficient and Hough transform to detect license plate. The m/n ratio reduction is performed to save time and memory. By the correlation coefficient between the standard pattern and the target pattern, licence plate area is roughly extracted. On the extracted local area, preprocessing and binarization is performed. The Hough transform is applied to find the extract outline of the plate. If the detection fails, a smaller or a larger standard pattern is used to compute the correlation coefficient. Through this process, the license plate of different size can be extracted. Two algorithms to each separate number are proposed. One segments each number with projection-histogram, and the other segments each number with the label. After each character is separated, it is recognized by the neural network. This research overlomes the problems in conventional methods, such as the time requirement or failure in extraction of outlines which are due to the processing of the entire image, and by processing in real time, the practical application is possible.

• The KIPS Transactions:PartA
• /
• v.9A no.2
• /
• pp.147-162
• /
• 2002

Multi-processor system with fast I/O devices improves processing performance and reduces the bottleneck by I/O concentration. In the system, the Performance influenced by shared memory used for exchanging data between processors varies with configuration and utilization. This paper suggests a prediction model for buffer and shared memory optimization under interrupt recognition method using mailbox. Ethernet (IEEE 802.3) packets are used as the input of system and the amount of utilized memory is measured for different network bandwidth and burstiness. Some empirical studies show that the amount of buffer and shared memory varies with packet concentration rate as well as I/O bandwidth. And the studies also show the correlation between two memories.

• The Journal of Korean Institute of Information Technology
• /
• v.16 no.12
• /
• pp.75-83
• /
• 2018

Recently, autonomous vehicles have been actively studied. Pedestrian detection and recognition technology is important in autonomous vehicles. Pedestrian detection using CNN(Convolutional Neural Netwrok), which is mainly used recently, generally shows good performance, but there is a performance degradation depending on the environment of the image. In this paper, we propose a pedestrian detection system applying long-term memory structure of hippocampal neural network based on CNN network with LGP-FL (Local Gradient Pattern-Feature Layer) added. First, change the input image to a size of $227{\times}227$. Then, the feature is extracted through a total of 5 layers of convolution layer. In the process, LGP-FL adds the LGP feature pattern and stores the high-frequency pattern in the long-term memory. In the detection process, it is possible to detect the pedestrian more accurately by detecting using the LGP feature pattern information robust to brightness and color change. A comparison of the existing methods and the proposed method confirmed the increase of detection rate of about 1~4%.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.30B no.10
• /
• pp.95-104
• /
• 1993

IPA(interpattern association) model is a method of feature extraction using a neural network. Even in the case that the reference patterns are simuklar to one another, this model can recover the reference patterns effectively. However, when the pattern whose feature pixels are lost is used as input, this model can not guarantee perfect recovery of the reference pattern. It is proposed a improved interpattern association(IPA) model for the feature extraction using neural network. The improved IPA model that combines the first interconnection weight matrix of the IPA model with the second additional weight matrix is proposed here to overcome the recovery problem of the original IPA model. The results of computer simulation and optical experiment are advanced.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.16 no.5
• /
• pp.403-412
• /
• 1991

This paper proposes a type of self organizing neural network which recognizes arbitrary symbols as well as numerical or alphabetic characters. The proposed algorithm autonomically organizes and classifies similar patterns on the basis of the distribution types of characteristics in the input images. Thus it can be appliced for the recognition of arbitrary images when it is difficult to establish a learning rule. It performs a stale recognition process with in the limit of the memory capacity. The cheme was applied and tested to 50 different image patterns with increased noise level up to 44%(SNR 2dB). The implementation results demonstrate that the proposed algorithm successfully recognizes the image patterns changed due to the various noise levels and thus proves excellent antinoise characteristics.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4412-4428
• /
• 2018

The purpose of this work is to solve the problem of representing an entire video using Convolutional Neural Network (CNN) features for human action recognition. Recently, due to insufficient GPU memory, it has been difficult to take the whole video as the input of the CNN for end-to-end learning. A typical method is to use sampled video frames as inputs and corresponding labels as supervision. One major issue of this popular approach is that the local samples may not contain the information indicated by the global labels and sufficient motion information. To address this issue, we propose a binary hashing method to enhance the local feature extractors. First, we extract the local features and aggregate them into global features using maximum/minimum pooling. Second, we use the binary hashing method to capture the motion features. Finally, we concatenate the hashing features with global features using different normalization methods to train the classifier. Experimental results on the JHMDB and MPII-Cooking datasets show that, for these new local features, binary hashing mapping on the sparsely sampled features led to significant performance improvements.