Keynote Speeches, Reserch Presentations, and Symposia

5/25 (Thu) Research Presentations at NHK STRL Auditorium

Research Presentation

1:00 pm ~ 1:30 pm

Face Recognition Technique for TV Program Video

Yoshihiko KAWAI
Senior Research Engineer, Internet Service Systems Research Division

A video analysis technique that uses artificial intelligence (AI), machine learning and big data analysis has undergone dramatic development in recent years, raising expectations for its application to broadcasting and Internet services. We are studying an image analysis technique for automatically annotating content-based information (metadata) of TV program, aiming at sophisticated support of the production of TV programs and effective use of archived video. The metadata will enable TV producers to search for desired video scenes from a large amount of video materials quickly and to produce a shorter summary video automatically, for example.

This report introduces a face recognition technique that can identify a person in a TV program video. Compared with authentication at a security gate, it is more diffi cult to achieve high recognition accuracy on TV video because the luminance condition, face direction and facial expression vary signifi cantly on a TV program video. In this report, we introduce the technique of using image features considering the positional relationship of facial parts such as the eyes, nose and mouth to improve the accuracy of face detection. We also introduce the technique of integrating image features calculated using a small-region step-by-step method to reduce the negative effects of video variation and to recognize individual differences accurately.

Research Presentation

1:30 pm ~ 2:00 pm

R&D on Live Production System Running on IP Networks

Tomofumi KOYAMA
Advanced Transmission Systems Research Division

As the IP network becomes faster and broader, broadcasters have begun using it in their program production facilities. In the beginning, the IP network was adopted in a video editing system. Recently, live production systems running on the IP network are gaining attention. The IP network makes the live production system simple and effi cient because it can transfer different types of media (e.g., video, audio, and synchronization) and video streams that are of different resolutions (e.g., high defi nition and ultrahigh defi nition).
In addition, it has the potential to provide new features such as remote production and automatic production system control.

In this report, we explain what broadcasters expect of an IP-based production system. Next, we describe challenges in the realization of the system and related works. Then, we introduce three technologies to realize the above advantages: “Resource Provider”, which is used to share resources among studios and extend the functions of studios; “8K signal convertor from SDI to IP”, which converts 8K from SDI to IP and from IP to SDI to transmit 8K signals with low delay on IP networks; and “Smart FPU”, which transmits return video, communication audio, and control signals as well as recorded video footage on a dedicated wireless link.

Research Presentation

2:00 pm ~ 2:30 pm

Research and Development towards Practical-Use Full-featured 8K Super Hi-Vision Cameras

Advanced Television Systems Research Division

Full-featured 8K Super Hi-Vision (SHV) is a next-generation television system that meets the superlative video parameters for 8K video standardized as ITU-R Rec. BT. 2020, and BT. 2100, 33-megapixel full resolution for RGB colors, 120 Hz frame frequency, 12- bit depth, wide color gamut, and high dynamic range (HDR). NHK STRL have already developed 33-megapixel image sensors that
can operate at 120 Hz and a three-chip Full-featured 8K SHV camera that has the reference imaging performance as a Full-featured 8K SHV camera.

In this report, first, an overview of conventional 8K camera developments is given. Next, a road map of future research and development toward practical Full-featured 8K cameras is shown. Finally, a newly developed portable 8K full-resolution single-chip camera system using a 133-megapixel CMOS image sensor is described. This camera system realized the world's first 8K fullresolution capture with 60-Hz operation by single-chip imaging without a color separation prism. Moreover, the camera system can shoot interpolated 8K 120-Hz video by applying a 2:1 interline and double-frame-frequency scanning technique and adaptive line interpolation with motion estimation. In this report, the outline of this camera system and the experimental results are explained.

Research Presentation

3:00 pm ~ 3:30 pm

Automatic Generation of Audio Descriptions for Sports Programs

Tadashi KUMANO
Principal Research Engineer, Human Interface Research Division

Audio descriptions, which provide vocal explanations of screen contents via audio subchannels, are important services for people with visual impairments to enjoy TV programs. However, only about 10 percent of all programs in NHK General TV are currently provided with audio descriptions owing to the cost of manual production. Furthermore, it is quite diffi cult to produce them for live programs. In order to solve these problems, NHK STRL has been researching a technology for the automatic generation of audio descriptions for sports programs in real time from sport-related live data including current scores or play-by-play records of events distributed by, for example, the organizers of the events. In this report, we explain our technology of the automatic generation.
First, a textual explanation of events or the current status of the game is generated, and then it is vocally conveyed using a speech synthesizer. We indicate some technical issues that must be addressed to enable practical services, and present our plans for research and development toward the 2020 Tokyo Olympic and Paralympic Games.

Research Presentation

3:30 pm ~ 4:00 pm

Natural 3D Visualization Technology for Integral 3D Displays Based on the Characteristics of Space Perception and Cognition

Three-Dimensional Image Research Division

We have been developing integral three-dimensional television (3D TV) systems as promising media for next-generation broadcasting services. An integral 3D TV has the potential to provide glasses-free natural 3D visualization as if the 3D objects actually exist in the viewer’s space. However, displayable 3D scenes are limited in actual implementations because their depth reconstruction range, in which 3D objects can be reconstructed with the best spatial resolution, is restricted to within a certain range. In fact, the images reconstructed outside the depth reconstruction range are inevitably blurred.

By taking into account the characteristics of human vision, we are investigating a method for enabling integral 3D TVs to display a large variety of scenes with more depth than their depth reconstruction range without inducing the feeling of unnaturalness in viewers.
Unlike measuring equipment, human vision does not always perceive the physical size and/or length of objects accurately. In this report, we briefl y introduce the characteristics of human vision related to space perception and cognition. Then, we show that our method can expand the depth reconstruction range in integral 3D TVs virtually, and that 3D scenes with substantial depth can be shown with perceptually appealing qualities.

Research Presentation

4:00 pm ~ 4:30 pm

Development of Green Organic Light-Emitting Diode with High Color Purity

Advanced Functional Devices Research Division

In the Super Hi-Vison video format, a wide color gamut system that can faithfully reproduce more vivid colors has been adopted, and the realization of a display suitable for the wide color gamut system is required. Organic electroluminescence (EL) displays have the advantages of excellent response speed and high contrast ratio. However, their color purity is not suffi cient owing to the broad emission spectrum of organic compounds. In particular, improving the color purity of the green organic light-emitting diode (OLED) has been a serious challenge.

In this report, two topics related to the demonstration of an effi cient green OLED with high color purity are described. One is a major change in the organic materials used in the OLED, enabling a drastic reduction in the spread of the emission wavelength compared with current OLEDs. At the same time, an internal quantum effi ciency of approximately 100% is demonstrated. The other topic is the improvement of color purity by changing the device structure to utilize the interference of light between the electrodes, which is based on the results of the optical simulation. The color purity of the green OLED demonstrated in this work is higher than that of previously reported OLEDs as a result of suppressing the luminescent component that hinders high color purity.


Please enter the exhibition booth number.

1 2 3
4 5 6
7 8 9
Del 0 Enter