• Smart Media Journal
• /
• v.12 no.11
• /
• pp.48-56
• /
• 2023

The technology of Three-dimensional human posture estimation is used in sports, motion recognition, and special effects of video media. Among various methods for this, multi-view 3D human pose estimation is essential for precise estimation even in complex real-world environments. But Existing models for multi-view 3D human posture estimation have the disadvantage of high order of time complexity as they use 3D feature maps. This paper proposes a method to extend an existing monocular viewpoint multi-frame model based on Transformer with lower time complexity to 3D human posture estimation for multi-viewpoints. To expand to multi-viewpoints our proposed method first generates an 8-dimensional joint coordinate that connects 2-dimensional joint coordinates for 17 joints at 4-vieiwpoints acquired using the 2-dimensional human posture detector, CPN(Cascaded Pyramid Network). This paper then converts them into 17×32 data with patch embedding, and enters the data into a transformer model, finally. Consequently, the MLP(Multi-Layer Perceptron) block that outputs the 3D-human posture simultaneously updates the 3D human posture estimation for 4-viewpoints at every iteration. Compared to Zheng[5]'s method the number of model parameters of the proposed method was 48.9%, MPJPE(Mean Per Joint Position Error) was reduced by 20.6 mm (43.8%) and the average learning time per epoch was more than 20 times faster.

• Journal of KIISE:Software and Applications
• /
• v.36 no.11
• /
• pp.875-890
• /
• 2009

Metabolimics interprets an organism as a network of functional units and an organism is represented by a metabolic pathway i.e., well-displayed graph. So a software tool for drawing pathway is necessary to understand it comprehensively. These tools have a problem that edge-crossings exponentially increase as the number of nodes grows. To apply automatic graph layout techniques to the genome-scale metabolic flow, it is very important to reduce unnecessary edge-crossing on a metabolic pathway layout. In this paper, we design and implement 2.5D metabolic pathway layout modules. Metabolic pathways are represented hierarchically by making use of the '2-layered layout algorithm' in 3D. It enhances the readability and reduces unnecessary edge-crossings by using 3D layout modules instead of 2D layout algorithms.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.1
• /
• pp.48-53
• /
• 2003

Eigenvoice method is known to be adequate for fast speaker adaptation, but it hardly shows additional improvement with increased amount of adaptation data. In this paper, to deal with this problem, we propose a modified method estimating the weights of eigenvoices in each feature vector dimension. We also propose an adaptation mode selection scheme that one method with higher performance among several adaptation methods is selected according to the amount of adaptation data. We used POW DB to construct the speaker independent model and eigenvoices, and utterances(ranging from 1 to 50) from PBW 452 DB and the remaining 400 utterances were used for adaptation and evaluation, respectively. With the increased amount of adaptation data, proposed dimensional eigenvoice method showed higher performance than both conventional eigenvoice method and MLLR. Up to 26% of word error rate was reduced by the adaptation mode selection between eigenvoice and dimensional eigenvoice methods in comparison with conventional eigenvoice method.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.7
• /
• pp.1667-1679
• /
• 2010

This thesis proposes an adaptive variable-sized matching window method using the characteristic points of the image and a method to increase the reliability of the cross-consistency check to raise the correctness of the final disparity image. The proposed adaptive variable-sized window method segments the image with the color information, finds the characteristic points in each segmented image, and varies the size of the matching window according to the existence of the characteristic points inside the window. Also the proposed cross-consistency check method processes the two cases with the cost values corresponding to the best disparity and the second-best disparity: when the cost values themselves are too large and when the difference between the two cost values are too small. The two proposed methods were experimented with the four test images provided by the Middleburry site. As the results from the experiments, the proposed adaptive variable-sized matching window method decreased up to 18.2% of error ratio and the proposed cross-consistency check method increased up to 7.4% of reliability.

• Journal of Digital Contents Society
• /
• v.11 no.1
• /
• pp.17-24
• /
• 2010

The purpose of this study is to design and implement a kernel for the support vector machine(SVM) to improve the performance of face recognition. Local feature analysis(LFA) has been well known for its good performance. SVM kernel plays a limited role of mapping low dimensional face features to high dimensional feature space but the proposed kernel using LFA is designed for face recognition purpose. Because of the novel method that local face information is extracted from training set and combined into the kernel, this method is expected to apply to various object recognition/detection tasks. The experimental results shows its improved performance.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.5
• /
• pp.141-147
• /
• 2021

In this paper, in order to improve the utilization of hyperspectral large-capacity data feature information by reducing complex computations by dimension reduction of neural network inputs in embedded systems, the band selection algorithm is applied in each subset. Among feature extraction and feature selection techniques, the feature selection aim to improve the optimal number of bands suitable for datasets, regardless of wavelength range, and the time and performance, more than others algorithms. Through this experiment, although the time required was reduced by 1/3 to 1/9 times compared to the others band selection technique, meaningful results were improved by more than 4% in terms of performance through the K-neighbor classifier. Although it is difficult to utilize real-time hyperspectral data analysis now, it has confirmed the possibility of improvement.

• The KIPS Transactions:PartB
• /
• v.12B no.6 s.102
• /
• pp.715-720
• /
• 2005

This paper presents a hybrid recognition method of first moment of face image and principal component analysis(PCA). First moment is applied to reduce the dimension by shifting to the centroid of image, which is to exclude the needless backgrounds in the face recognitions. PCA is implemented by single layer neural network which has a teaming rule of Foldiak algorithm. It has been used as an alternative method for numerical PCA. PCA is to derive an orthonormal basis which directly leads to dimensionality reduction and possibly to feature extraction of face image. The proposed method has been applied to the problems for recognizing the 48 face images(12 Persons $\ast$ 4 scenes) of 64$\ast$64 pixels. The 3 distances such as city-block, Euclidean, negative angle are used as measures when match the probe images to the nearest gallery images. The experimental results show that the proposed method has a superior recognition performances(speed, rate). The negative angle has been relatively achieved more an accurate similarity than city-block or Euclidean.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.4
• /
• pp.381-385
• /
• 2003

We suggest a method which reduces the search space of a model-base on multiple-view approach for polyhedral object recognition using the ART-1 neural network. In this approach, the model-base is consisted of extracted features from two-dimensional projections observed at the predetermined viewpoints of a viewing sphere enclosing the object.

• Proceedings of the KAIS Fall Conference
• /
• 2009.12a
• /
• pp.850-853
• /
• 2009

반도체 조립공정에서 사용되는 Pick and Placement장비는 반도체패키지를 컴퓨터 비젼을 이용하여 위치 정렬하고 Placement Tray에 적재하는 장비로서 고속,고정밀도가 요구된다. 다변량 통계적 분석방법인 주성분 분석법은 주어진 데이터에서 특징이 되는 일정한 패턴을 찾는 방법으로 영상의 차원감소를 위해 최근 많이 사용되어지고 있다. 본 논문에서는 반도체패키지의 기하학적 형태를 이용하여 위치정렬을 하도록 한 후 성능을 검증하도록 하였다. 패키지 원영상에서 밝기값의 차이에 따른 윤곽선을 인식한 후, 각 위치값들을 주성분 분석법을 이용해 직선을 추출한 방법으로 위치정렬한 결과 신뢰할만한 위치정렬 성능을 보였다.

• Journal of the Korean Society for Precision Engineering
• /
• v.22 no.10 s.175
• /
• pp.107-113
• /
• 2005

This paper presents a method to reduce errors in registration, which is used in transformation coordinate system of the multiple measuring data. In general, the ICP algorithms and feature-based approaches are used for registration. In order to measure wrap-around object, it is necessary to change the scanning direction or set-up of the object. A fixture is made to reduce registration errors caused by inaccurate center point of tooling balls, providing the more accurate registration method. And, the motorized fixture controls rotation and tilting to get precise the measuring data and registration. The proposed motorized fixture and registration method have advantages in accurate registration and precise measurement, compared with the conventional methods.