About Us
Yamagishi Lab
Message
Acoustic Research
and Technology Development that Benefits Society
Our laboratory is more than just a venue for acoustic research.
We are an acoustics laboratory that evolves together with society. We are committed to mid- and long-term scientific exploration of sound and developing advanced technologies that address the complex acoustics-related challenges and societal needs facing Japan.
Our vision is to contribute to the betterment of Japan's future through cutting-edge sound research. Aside from presenting academic papers at top conferences, we also collaborate with Japanese companies on initiatives to translate the fruits of our research into concrete applications for society.
By publishing research data sets and undertaking projects that pioneer innovation, we help drive the advancement of Japanese university acoustics research that paves the way for the future.
Mission
Contributing to Society Through Acoustic Research
and State-of-the-Art Technology
Our laboratory is committed to its mission to resolve complex sound-related issues affecting Japanese society through academic research and advanced acoustic technologies.
We continuously endeavor to contribute to society, pioneering efforts to tackle issues that cannot be addressed by conventional speech information processing
and develop scientific and technological solutions to address them. Some of the problems we have focused on include deep fakes,
voice cloning, and privacy implications associated with large-scale speech data used to train generative models.
We also provide support for Japanese university research by publishing research data sets and organizing academic challenges as an Inter-University Research Institutional Corporation.
Vision
Acoustic Research
that Paves the Way to the Future and Giving Back to Society
At the National Institute of Informatics's Yamagishi Lab (a national research institute), our vision is to conduct acoustic research and develop state-of-the-art technology rooted in society's mid- and long-term needs.
These R&D initiatives are intended to give back to society by producing seed technologies to better Japan's future and enhance the technological capabilities of Japanese universities and companies.
We aim to be an acoustic research laboratory that evolves together with society for the future of Japan.
Research Environment
Research Environment
The National Institute of Informatics' Yamagishi Lab, located in Hitotsubashi, Tokyo, provides a suitable environment for conducting a wide range of research on sound and language.
The laboratory provides the ideal space to conduct a wide range of research on sound and language, from acoustic experiments, machine-learning research on sound generation and identification models based on large-scale data, and applied science research to solve societal issues such as combating deepfakes.
for Companies
Yamagishi Lab
specializes in research on acoustics and linguistics.
-
01
Sound generation and conversion research
Speech Generation and Transformation
Our research includes text-to-speech technology that automatically synthesizes natural sounds from textual information in various languages, and sound generation models that automatically generate natural instrumental sounds from musical score information such as MIDI. We also conduct research on integrating speech signal processing and machine learning, voice conversion techniques such as intelligibility enhancement to improve the clarity of voices in crowded places, and making computers capable of understanding sound quality as heard by humans.
-
02
Research on automated speaker recognition from speech
Speaker Recognition from Speech
Speech carries information beyond language. It can also help computers understand non-verbal details such as who the speaker is, their gender, dialect, emotions, health status, among others. Our laboratory is also working on research concerning speaker recognition (voice biometric technology). These efforts aim to improve the matching accuracy of speaker recognition and system robustness against noise, as well as develop technology to guard against voice impersonation.
-
03
Research on Discerning Between AI-synthesized Speech
and Human Speech.Distinguishing AI-Generated Speech from Human Speech
Generative AI can automatically generate audio and video media remarkably similar to the real thing. While this technology is expected to be useful for various business applications, its misuse for malicious purposes may cause serious social harm. Our laboratory is working on how to accurately and automatically detect deepfakes to prevent the proliferation of misinformation.
-
04
Voice Privacy Research,
Such as Those on Voice Anonymization.Voice Privacy and Anonymization
With the current ease of identifying individuals from audio on the Web and the creation of deepfake audio through voice cloning, speaker anonymization technology is becoming essential to protect the speaker's privacy by modifying speech data before releasing it to the public. Our laboratory is conducting research to transform voices so that they sound natural while keeping the speaker anonymous. This research aims to create technologies that can be used for broadcasting.
Joint Research
Collaborations Between Corporations and Yamagishi Lab
The National Institute of Informatics' Yamagishi Laboratory can work with businesses in three ways: joint research, technology provision (patent and software licensing), and technical support and advisory services.
Three Key Advantages of Collaborative Research
-
( 01 )
Presentations at Top International Conferences
The findings of collaborative research with Yamagishi Lab can be presented at top international conferences and journal papers on speech information processing.
-
( 02 )
Global Technology Recognition
Many of these speech information processing conferences are also attended by Big Tech, providing an opportunity for our research and technological achievements to gain worldwide recognition.
-
( 03 )
Enhancing the Research Expertise of Corporate Researchers
We provide a range of support and advice via collaborative research to studies being conducted by corporate researchers. This aid helps grow the skills of these corporate researchers.
Collaboration Project Members
Lab Members
Providing Technology and Technical Support
-
Patents and Software Licenses
We offer our technology to companies
that want to use our research in their business. -
Technical Support and Advisory
Using our scientific and technical expertise
to help companies overcome problems.
Achievements
Track Record of Collaboration with Companies
-
Collaborative Research
Fujitsu Limited
Fujitsu, along with nine industry-academic organizations, is working on a collaborative research project aimed at evaluating audio and video content to determine whether they are deepfakes. Information obtained from these sources, along with other information, will then be used to counter the spread of misinformation.
-
Collaborative Research
DataGrid Inc.
We developed a deepfake detection model for Japanese audio as part of a joint initiative with DATAGRID Inc. to create technology for countermeasures against false and misleading information on the Internet.
-
Provided Technologies
Japan Broadcasting Corporation (NHK)
Speaker anonymization technology developed by Yamagishi Lab is being used by NHK (Japan Broadcasting Corporation) in their TV broadcasts.
Provided Technologies
CyberAgent, Inc.
We provided CyberAgent, Inc. with SYNTHETIQ VISION, a program that automatically determines the authenticity of AI-generated faces.
-
Technical Guidance
HOYA Corporation Readspeaker Division
Using Yamagishi Lab's speech intelligibility enhancement technology, we improved the ReadSpeaker text-to-speech system for Tokaido Shinkansen station broadcasts and provided technical support on enhancing speech intelligibility in noisy environments.
-
