• Journal of Broadcast Engineering
• /
• v.21 no.1
• /
• pp.76-86
• /
• 2016

This study was investigated whether the perceived depth was changed depending on the measurement methods. In the method of direct comparison, virtual object with one of the various binocular disparities was presented in the frontal space with LEDs which were used for depth estimation for a binocular stimulus, while in the method of indirect comparison, visual object was presented in the frontal space but the LEDs were placed rightward at the angle of 45 degree from the mid-sagittal line. In these experimental setup, the depth of binocular stimulus was directly matched that of LED in direct comparison condition. In indirect comparison condition, however, observer estimated the depth of binocular stimulus, turned one's head rightward to the array of LEDs and turned on the LED which was supposed to be the same depth as binocular stimulus. Additionally, it was investigated whether the perceived depth was different depending on observer's stereo acuity. The results showed that perceived depths measured in the direct comparison were more similar to the depth predicted from geometry than those in the indirect comparison, and that the perceived depths from observers with high stereo acuity were similar to the predicted depth from geometry those from observers with low stereo acuity. These results indicated that stereoscopic depths of the binocular stimuli would vivid and compelling when binocular stimuli was simultaneously presented with real objects in the same visual space, like a mixed reality.

• Journal of Broadcast Engineering
• /
• v.21 no.6
• /
• pp.929-943
• /
• 2016

This paper proposes the rate control algorithm based on the S-JND (Saliency-Just Noticeable Difference) model for considering perceptual visual quality. The proposed rate control algorithm employs the S-JND model to simultaneously reflect human visual sensitivity and human visual attention for considering characteristics of human visual system. During allocating bits for CTU (Coding Tree Unit) level in a rate control, the bit allocation model calculates the S-JND threshold of each CTU in a picture. The threshold of each CTU is used for adaptively allocating a proper number of bits; thus, the proposed bit allocation model can improve perceptual visual quality. For performance evaluation of the proposed algorithm, the proposed algorithm was implemented on HM 16.9 and tested for sequences in Class B and Class C under the CTC (Common Test Condition) RA (Random Access), Low-delay B and Low-delay P case. Experimental results show that the proposed method reduces the bit-rate of 2.3%, and improves BD-PSNR of 0.07dB and bit-rate accuracy of 0.06% on average. We achieved MOS improvement of 0.03 with the proposed method, compared with the conventional method based on DSCQS (Double Stimulus Continuous Quality Scale).

• Journal of Broadcast Engineering
• /
• v.23 no.3
• /
• pp.351-360
• /
• 2018

Human emotion recognition is a research topic that is receiving continuous attention in computer vision and artificial intelligence domains. This paper proposes a method for classifying human emotions through multiple neural networks based on multi-modal signals which consist of image, landmark, and audio in a wild environment. The proposed method has the following features. First, the learning performance of the image-based network is greatly improved by employing both multi-task learning and semi-supervised learning using the spatio-temporal characteristic of videos. Second, a model for converting 1-dimensional (1D) landmark information of face into two-dimensional (2D) images, is newly proposed, and a CNN-LSTM network based on the model is proposed for better emotion recognition. Third, based on an observation that audio signals are often very effective for specific emotions, we propose an audio deep learning mechanism robust to the specific emotions. Finally, so-called emotion adaptive fusion is applied to enable synergy of multiple networks. The proposed network improves emotion classification performance by appropriately integrating existing supervised learning and semi-supervised learning networks. In the fifth attempt on the given test set in the EmotiW2017 challenge, the proposed method achieved a classification accuracy of 57.12%.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.6
• /
• pp.203-212
• /
• 2019

Along with the recent fourth industrial revolution, the world's manufacturing powerhouses are pushing for national strategies to revive the sluggish manufacturing industry. Moon Jae-in, the government is in accordance with the trend, called 'advancement of science and technology is leading the fourth round of the Industrial Revolution' strategy. Intelligent information technology such as IoT, Cloud, Big Data, Mobile, and AI, which are key technologies that lead the fourth industrial revolution, is promoting the emergence of new industries such as robots and 3D printing and the smarting of existing major manufacturing industries. Advances in technologies such as smart factories have enabled IoT-based sensing technology to measure various data that could not be collected before, and data generated by each process has also exploded. Thus, this paper uses data generators to generate virtual data that can occur in smart factories, and uses them to analyze the cause of the defect in real time and to predict the replacement cycle of the facility.

• Journal of Broadcast Engineering
• /
• v.12 no.5
• /
• pp.391-400
• /
• 2007

The ways of expression are infinite to make animated film. When we compare to live action film, that fact is the actual power that animation only has and is one of the big weapons to appeal to the audience. Nonetheless, techniques to be chosen are monolithic like drawing, 3D computer, and clay when animation maker makes feature animation to screen in movie theatre. It's probably because feature animation must have popularity of visual style and they need to standardize technique about side of making. But popularity in expression does not mean monolithic visual style and making technique. It should be understood about side of artistic value to fascinate public audience. Audience is always eager for new subject, new directing style and new visual style. Monotonous and boring animation cannot but turn away from the audience. So if production does not have artistic value, we cannot expect commercial value and success also even it plans and makes for the feature animation. To create new visual style fnr animation is not only limited to artwork itself but also creation, which is included story and tone of the film and acting style about character. Also making process must be calculated and experimented to actualize that visual style about side of making.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.2
• /
• pp.21-27
• /
• 2020

The purpose of this study is to investigate the ecosystem structure underlying the development of artificial intelligence technology and related industries. In addition, the research on the AI business ecosystem based on AI technology and the ways to activate them was discussed. Ecosystems play a role in organically connecting producers, consumers, and decomposers. In the AI ecosystem, we classified the AI service producers, producers of AI services using the produced services, and data and related infrastructure services that are the basis of AI services. Stakeholders in the AI business ecosystem are the government and various private organizations that have a direct or indirect influence on AI service production, consumption, and operation. In Korea, in particular, the government plays a role as the most influential stakeholders. For example, the company contributes to the increase of producers, which are related to human resource development, and plays a catalyst role in the increase of services produced by R & D funding. In this study, the policy for revitalizing the AI business ecosystem includes (1) securing the environment for increasing producers, (2) spreading AI awareness among consumers, (3) securing data exchange and supply infrastructure, and (4) supporting services and related laws. Secure the system. This study is meaningful in that it contributes to and contributes to the construction of domestic AI-based environment and related research.

• Journal of Broadcast Engineering
• /
• v.15 no.3
• /
• pp.426-433
• /
• 2010

The H.264/AVC video coding standard performs inter prediction using variable block sizes to improve coding efficiency. Since we predict not only the motion of homogeneous regions but also the motion of non-homogeneous regions accurately using variable block sizes, we can reduce residual information effectively. However, each motion vector should be transmitted to the decoder. In low bit rate environments, motion vector information takes approximately 40% of the total bitstream. Thus, motion vector competition was proposed to reduce the amount of motion vector information. Since the size of the motion vector difference is reduced by motion vector competition, it requires only a small number of bits for motion vector information. However, we need to send the corresponding index of the best motion vector predictor for decoding. In this paper, we propose a new codeword table based on the phased-in code to encode the index of motion vector predictor efficiently. Experimental results show that the proposed algorithm reduces the average bit rate by 7.24% for similar PSNR values, and it improves the average image quality by 0.36dB at similar bit rates.

• Journal of Broadcast Engineering
• /
• v.16 no.3
• /
• pp.510-519
• /
• 2011

Intra mode decision is a useful coding tool in Distributed Video Coding (DVC) for improving DVC coding efficiency for video sequences having fast motion. A major limitation associated with the existing intra mode decision methods, however, is that its efficiency highly depends on user-specified thresholds or modeling parameters. This paper proposes an entropy-based method to address this problem. The probabilities of intra and Wyner?Ziv (WZ) modes are determined firstly by examining correlation of pixels in spatial and temporal directions. Based on these probabilities, entropy of the intra and the WZ modes are computed. A comparison based on the entropy values decides a coding mode between intra coding and WZ coding without relying on any user-specified thresholds or modeling parameters. Experimental results show its superior rate-distortion performance of improvements of PSNR up to 2 dB against a conventional Wyner?Ziv coding without intra mode decision. Furthermore, since the proposed method does not require any thresholds or modeling parameters from users, it is very attractive for real life applications.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.2
• /
• pp.141-146
• /
• 2017

New infectious diseases such as MERS have been in need of many measures such as initial discovery, isolation, and crisis response. In addition, the culture of hospitals is changing, such as the general public 's visiting and Nursing Care Integration Services. However, as the qualifications and regulations of medical personnel in hospitals become rigid, overseas such as linens, wastes movements are replacing possible works with robots. we have developed a hospital logistics robot that can carry out various goods delivery within a hospital, and can move various kinds of objects safely to a desired location. In this thesis, we have studied a hospital logistics robot that can carry out various kinds of goods delivery within the hospital, and can move various kinds of objects such as waste, and linen safely to a desired location. The movement of a robot in a hospital may cause a collision between a person and an object, so that the collision must be prevented. In order to prevent collision, it is necessary to recognize whether or not an object exists in the movement path of the robot. And if there is an object, it should recognize whether it moves or not. In order to recognize human beings and objects, we recognize the person with face/body recognition technology and generate the context awareness of the object using 3D Vision image segmentation technology. We use the generated information to create a map that considers objects and person in the robot moving range. Thus, the robot can be operated safely and efficiently.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.16 no.4 s.95
• /
• pp.427-437
• /
• 2005

In this paper, the perceived audio level metering algorithm of digital audio sound to be able to operate in real-time is proposed. Through analyzing a conventional recommendation ITU-RBS1387-I for objective audio quality analysis, FFT-based loudness metering algorithm is implemented and the real-time method of that algorithm was advised and proved. The proposed method is based on look-up table. In order to prove the proved method, using 23 pure tones and 30 preselected digital audio samples, its performance and operation time is evaluated. Its performance, compared with an original algorithm's, have a good figure of less than $2\;\%$ error even if look-up table related with spectral spreading have large level resolution of $10\;\cal{dB}$. The proposed algorithm take only 1/21 of original algorithm's measuring time. Also, in the proposed algorithm auditory pitch group energy calculation take 1/450 of original algorithm's and excitation calculation take 1/3.57. In conclusion, the proposed algorithm is expected to be implemented into DSP-based real-time loudness meter.