September 2021 Vol. 19 No. 9

View from the Top

• Hidehiro Tsukano, Senior Vice President, Head of NTT IOWN Integrated Innovation Center

  

  Overview
  In addition to the three laboratory groups that have been the cornerstones of NTT’s research and development (i.e., NTT Service Innovation Laboratory Group, NTT Information Network Laboratory Group, and NTT Science and Core Technology Laboratory Group), the NTT Innovative Optical and Wireless Network (IOWN) Integrated Innovation Center (IIC) was established on July 1, 2021 to extend technology development closer to the commercial implementation stage. IIC is striving to create and implement photonics-electronics convergence technology, which fuses optical and electrical signals and is key to enable IOWN. We interviewed Hidehiro Tsukano, head of IIC, about the purpose of the establishment and mission of IIC as well as the qualities required of top management.

Front-line Researchers

• Shinji Matsuo, Senior Distinguished Researcher, NTT Device Technology Laboratories and NTT Basic Research Laboratories

  

  Overview
  The annual power consumption of datacenters in Japan accounted for 1% of Japan’s total power consumption in 2015. As the speed and capacity of data processing and transmission increases, power consumption is steadily increasing, and reducing that consumption is becoming a serious issue. To address this issue, Dr. Shinji Matsuo, a senior distinguished researcher at NTT Device Technology Laboratories and NTT Basic Research Laboratories, is researching and developing innovative technologies for high-density integration of compound semiconductors on silicon substrates to enable photonics-electronics converged integrated circuits. We interviewed him about the progress of his research and his attitude as a researcher.

• Masayuki Terada, Senior Manager, X-Tech Development Department, NTT DOCOMO

  

  Overview
  Technological innovation has made it possible to generate, collect, and store a vast and diverse range of data. There are efforts being made to analyze data that have been overlooked in the past and use them for business. NTT DOCOMO is also using big data from a new perspective to solve social issues. We interviewed Masayuki Terada, who is engaged in the research and practical application of population statistics using mobile network data called Mobile Spatial Statistics and traffic-jam prediction using artificial intelligence called Traffic Congestion Forecasting AI, about the current progress of research and development and the thrill of being a researcher and developer.

Rising Researchers

• Kenta Niwa, Distinguished Researcher, NTT Communication Science Laboratories

  

  Overview
  While modern deep learning often requires aggregating data into a single datacenter to train models, in the near future data will be distributed due to increased data volume and privacy protection concern. In this article, we spoke to Kenta Niwa, a distinguished researcher working on asynchronous distributed deep learning technology. This technology allows us to optimize machine learning models as if the data was aggregated in a single datacenter, even in the modern era of distributed data.

Feature Articles: Creativity and Technology—Designing for an Unknown Future

• Reach Out and Touch Someone’s Heart: Exploring the Essence of Communication to Create a Spiritually Rich Society

  

  Abstract
  NTT Communication Science Laboratories has been exploring the essence of communication since its founding 30 years ago. With the aim of achieving communication that reaches the heart, its researchers have been creating innovative technologies that approach and exceed human abilities in fields such as media processing and data science. They have also been discovering basic principles that lead to a deeper understanding of humans in fields such as cognitive neuroscience and brain science. This article introduces key activities at NTT Communication Science Laboratories in pursuit of the essence of communication with a look back at past research.

• The Day a System Becomes a Conversation Partner—Exploring New Horizons in Social Dialogue Systems with Large-scale Deep Learning

  

  Abstract
  People live their lives by casually talking with others on a daily basis. Such “social” dialogue contributes to building trust among people and satisfying their desire to talk with others. There has been a growing interest in social dialogue systems to satisfy the human desire for chatting with others, and we have been working on a wide range of research projects to develop such systems. With the rapid progress in deep learning, high-performance social dialogue systems using deep learning have been proposed. In this article, we introduce NTT’s social dialogue system using the latest deep-learning models as well as the current achievements obtained and challenges with this system.

• Looking More, Acting Better

  

  Abstract
  One key issue in developing user-friendly information and communication technology is to understand the behavior or action of users. Humans readily exhibit natural and complex movements, and such motor control is enabled by sophisticated brain mechanisms, including the control of eye movements to obtain target information and generation of limb movements. In this article, we address the question, ”Why is the eye important for skilled motor actions?” by introducing explanations from literature and propose an interpretation based on our recent findings.

• Developing AI that Pays Attention to Who You Want to Listen to: Deep-learning-based Selective Hearing with SpeakerBeam

  

  Abstract
  In a noisy environment such as a cocktail party, humans can focus on listening to a desired speaker, an ability known as selective hearing. In this article, we discuss approaches to achieve computational selective hearing. We first introduce SpeakerBeam, which is a neural-network-based method for extracting speech of a desired target speaker in a mixture of speakers, by exploiting a few seconds of pre-recorded audio data of the target speaker. We then present our recent research, which includes (1) the extension to multi-modal processing, in which we exploit video of the lip movements of the target speaker in addition to the audio pre-recording, (2) integration with automatic speech recognition, and (3) generalization to the extraction of arbitrary sounds.

• Technique for Modulating the Tactile Sensation of Objects Using an Illusion

  

  Abstract
  The human ability to determine the tactile textures of objects seems to be very stable: When someone rubs, holds, hits, or touches a stone, for example, she/he never confuses its tactile texture with that of fur or a sponge. However, the phenomenon known as the velvet hand illusion indicates that human perception of tactile texture can be easily distorted. This article describes a psychological study that examined the nature of this tactile illusion and presents a technique with which the perception of texture can be modulated using this phenomenon.

Regular Articles

• Routing and Spectrum Assignment Using Deep Reinforcement Learning in Optical Networks

  

  Abstract
  In future optical networks, effective use of spectral resources will be an issue as complexity increases due to diversified and dynamic requirements. In this article, we first give an overview of the routing and spectrum assignment (RSA) problem in elastic optical networks. We then introduce our novel RSA algorithm called Mask RSA, which allows for efficient route and spectral resource selection by deep reinforcement learning.

Global Standardization Activities

• Next-generation Metallic Access Technologies and Their Standardization Activities

  

  Abstract
  This article introduces certain next-generation metallic access technologies and the standardization activities relating to them under study in the International Telecommunication Union’s Telecommunication Standardization Sector (ITU-T) Study Group 15 with a strong focus on G.fast (“G” stands for ITU-T G series of recommendations and “fast” stands for fast access to subscriber terminals) and MGfast (multi-gigabit fast access to subscriber terminals) that can provide optical-fiber-grade ultrahigh-speed transmission services using pre-installed metallic cables in the existing infrastructure. MGfast, which is expected to be a next-generation metallic access technology, targets a transmission rate of 5 to 10 Gbit/s (uplink and downlink combined) over twisted-pair cables and coaxial cables.

External Awards/Papers Published in Technical Journals and Conference Proceedings

• External Awards/Papers Published in Technical Journals and Conference Proceedings