• IEMEK Journal of Embedded Systems and Applications
• /
• v.16 no.5
• /
• pp.179-186
• /
• 2021

For stereo matching based on deep learning, the design of network structure is crucial to the calculation of matching cost, and the time-consuming problem of convolutional neural network in image processing also needs to be solved urgently. In this paper, a method of stereo matching using sparse loss volume in parallax dimension is proposed. A sparse 3D loss volume is constructed by using a wide step length translation of the right view feature map, which reduces the video memory and computing resources required by the 3D convolution module by several times. In order to improve the accuracy of the algorithm, the nonlinear up-sampling of the matching loss in the parallax dimension is carried out by using the method of multi-category output, and the training model is combined with two kinds of loss functions. Compared with the benchmark algorithm, the proposed algorithm not only improves the accuracy but also shortens the running time by about 30%.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.6
• /
• pp.347-353
• /
• 2022

In this paper, we propose a method that performs a text-video retrieval model by replacing video properties using captions. In general, the exisiting embedding-based models consist of both joint embedding space construction and the CNN-based video encoding process, which requires a lot of computation in the training as well as the inference process. To overcome this problem, we introduce a video-captioning module to replace the visual property of video with captions generated by the video-captioning module. To be specific, we adopt the caption generator that converts candidate videos into captions in the inference process, thereby enabling direct comparison between the text given as a query and candidate videos without joint embedding space. Through the experiment, the proposed model successfully reduces the amount of computation and inference time by skipping the visual processing process and joint embedding space construction on two benchmark dataset, MSR-VTT and VATEX.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2009.11a
• /
• pp.147-148
• /
• 2009

In this paper, we describe a telemetry stimulation system that controls animal-robots. In our system, we send the main control command from PC to the controller embedded in the pigeon based animal-robots. Once the controller receives the control signal, it makes biphasic stimulation pulses to medial forebrain bundle neurons to control the pigeon behavior as we want. We design the embedded controller using CUBLOC, which is lightweight for attaching on the pigeon.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.51 no.3
• /
• pp.112-121
• /
• 2002

This paper describes a new bike system for the postural balance rehabilitation training. Virtual environment and three dimensional graphic model is designed with CAD tools such as 3D Studio Max and World Up. For the real time bike simulation, the optimized WorldToolKit graphic library is embedded with the dynamic geometry generation method, multi-thread method, and portal generation method. In this experiment, 20 normal adults were tested to investigate the influencing factors of balancing posture. We evaluated the system by measuring the parameters such as path deviation, driving velocity, COP(center for pressure), and average weight shift. Also, we investigated the usefulness of visual feedback information by weight shift. The results showed that continuous visual feedback by weight shift was more effective than no visual feedback in the postural balance control It is concluded this system might be applied to clinical use as a new postural balance training system.

• The Journal of Korea Robotics Society
• /
• v.12 no.4
• /
• pp.425-431
• /
• 2017

As drones gain more popularity these days, drone detection becomes more important part of the drone systems for safety, privacy, crime prevention and etc. However, existing drone detection systems are expensive and heavy so that they are only suitable for industrial or military purpose. This paper proposes a novel approach for training Convolutional Neural Networks to detect drones from images that can be used in embedded systems. Unlike previous works that consider the class probability of the image areas where the class object exists, the proposed approach takes account of all areas in the image for robust classification and object detection. Moreover, a novel loss function is proposed for the CNN to learn more effectively from limited amount of training data. The experimental results with various drone images show that the proposed approach performs efficiently in real drone detection scenarios.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.52 no.9
• /
• pp.526-532
• /
• 2003

This paper describes a recognition method of a road direction for an autonomous vehicle based on magnetic sensors. Using the sensors mounted on a vehicle and the magnetic markers embedded along the center of road, the autonomous vehicle can recognize a road direction and control a steering angle. Using the front lateral deviation of a vehicle and the rear one, the road direction is calculated. The analysis of magnetic field, the acquisition technique of training data, the training method of neural network and the computer simulation are presented. According to the computer simulation, the proposed method is simulated, and its performance is verified. Also, the experimental test is confirmed its reliability.

• Journal of The Korean Association For Science Education
• /
• v.22 no.5
• /
• pp.1082-1102
• /
• 2002

Most of comparative studies in science teacher education so far have been conducted in terms of teacher education policy, pre- and in-service training system and curriculum, and certificate system. While such superficial information can be readily obtainable, it does not necessarily enable us to make access to reality of science teachers' professional development in respective countries, because practice in professional development among science teachers is deeply embedded into respective socio-cultural environment or climate. In order to get information on reality in science teachers' professional development, alternative approaches of research should be developed. This paper aims at pursuing an alternative way to approach reality of Japanese science teachers' professional development. An email survey of free description method with 29 in-service science teachers with a variety of years of experience in Ibaraki Prefecture, Japan, revealed that Japanese science teachers have developed their expertise through very close daily-based communication with their peer science teachers. At least, within their consciousness, neither formal in-service training programs, nor pre-service training programs have had much stronger effects on their professional development than such non-formal, daily-based, deep, apprenticeship-typed or in some sense, family-typed communication. The results suggest that in order to conduct meaningful comparative studies, we should take much more attention to how to make access to reality of science teachers' professional development.

• The Journal of the Acoustical Society of Korea
• /
• v.30 no.8
• /
• pp.461-467
• /
• 2011

Accurate classification of input signals is the key prerequisite for variable bit-rate coding, which has been introduced in order to effectively utilize limited communication bandwidth. Especially, recent surge of multimedia services elevate the importance of speech/music classification. Among many speech/music classifier, the ones based on support vector machine (SVM) have a strong selling point, high classification accuracy, but their computational complexity and memory requirement hinder their way into actual implementations. Therefore, techniques that reduce the computational complexity and the memory requirement is inevitable, particularly for embedded systems. We first analyze implementation of an SVM-based classifier on embedded systems in terms of execution time and energy consumption, and then propose two techniques that alleviate the implementation requirements: One is a technique that removes support vectors that have insignificant contribution to the final classification, and the other is to skip processing some of input signals by virtue of strong correlations in speech/music frames. These are post-processing techniques that can work with any other optimization techniques applied during the training phase of SVM. With experiments, we validate the proposed algorithms from the perspectives of classification accuracy, execution time, and energy consumption.

• Speech Sciences
• /
• v.5 no.1
• /
• pp.7-21
• /
• 1999

This paper proposes an improved training procedure in speech recognition based on the continuous density of the Hidden Markov Model (CDHMM). Of the three parameters (initial state distribution probability, state transition probability, output probability density function (p.d.f.) of state) governing the CDHMM model, we focus on the third parameter and propose an efficient algorithm that determines the p.d.f. of each state. It is known that the resulting CDHMM model converges to a local maximum point of parameter estimation via the iterative Expectation Maximization procedure. Specifically, we propose two independent algorithms that can be embedded in the segmental K -means training procedure by replacing relevant key steps; the adaptation of the number of mixture Gaussian p.d.f. and the initialization using the CDHMM parameters previously estimated. The proposed adaptation algorithm searches for the optimal number of mixture Gaussian humps to ensure that the p.d.f. is consistently re-estimated, enabling the model to converge toward the global maximum point. By applying an appropriate threshold value, which measures the amount of collective changes of weighted variances, the optimized number of mixture Gaussian branch is determined. The initialization algorithm essentially exploits the CDHMM parameters previously estimated and uses them as the basis for the current initial segmentation subroutine. It captures the trend of previous training history whereas the uniform segmentation decimates it. The recognition performance of the proposed adaptation procedures along with the suggested initialization is verified to be always better than that of existing training procedure using fixed number of mixture Gaussian p.d.f.

• Journal of Digital Contents Society
• /
• v.18 no.2
• /
• pp.267-274
• /
• 2017

Recently, there is a new trend about internet of things (IoT) such as shops with smart mirror around the fashion and beauty industry. Since smart mirror can display a content through a monitor which is attached to back of mirror system while looking through a mirror, it can be applied to various industries such as fashion, beauty and health care. This paper proposes an efficient learning system requiring no assistance from others for the hearing deaf who atrophy verbal skill and are inaccurate in pronunciation by using features of smart mirror. Also, this system proposes an efficient and simple lip reading method which can be applied to an embedded system and improves a learning efficiency by employing previously verified pronunciation training data.