The Future of AI Speech — ObEN’s State-of-the-Art Research at Interspeech 2018
Innovation has always been in ObEN’s DNA, which is why so much of our team consists of researchers, engineers, and developers working on developing state-of-the-art technology for everything from voice synthesis to motion tracking. Every year, their work leads to publications in leading scientific journals and magazines as well as presentations at major technology conferences.
One such conference was last months Interspeech 2018, held in Hyderabad, India, where two of ObEN’s researchers to invited to present their work to a global audience of their peers and pillars within the speech technology. Interspeech is the world’s largest and most respected conference focused on the science and technology of spoken language processing. The conference is a medium of members of the speech industry to present fundamental theories, cutting-edge research, and industry applications. Competition to present is fierce and only the best of research projects are invited to present on the Interspeech stage — which is why ObEN is even more proud that two of our researchers were invited to present at the conference.
ObEN Senior Research Scientist Taehwan presented his paper titled “INVESTIGATION OF USING DISENTANGLED AND INTERPRETABLE REPRESENTATIONS FOR ONE-SHOT CROSS-LINGUAL VOICE CONVERSION”. It’s a mouthful, but the paper, published along with two other ObEN research scientists, details a solution for converting a speaker’s voice to different languages. This is part of the research that enables ObEN’s Personal AI avatars to speak in multiple languages in the user’s own voice, even if the only language the user can actually speak in English. Not only are ObEN’s researchers solving this advanced speech technology problem with our AI, they are also developing ways to do it with less training time for the AI, and with smaller samples of the speaker’s data — as little as just one sentence!
In addition, our Speech Research Engineer Carol presented her Master’s dissertation at a special showcase of exceptional female researchers. One of only 24 presenters from top universities worldwide, Carol not only spoke about her doctoral work but also participated in discussions on the importance of developing female leadership in STEM fields. Her participation at the conference was a highlight for ObEN, as we pride ourselves in the gender diversity of our research teams and senior leadership. Carol also returned with insight into the work done in automated speech recognition and how ObEN’s application of Deep Neural Networks can continue to help our speech technology reach new heights.
Our research presentations attracted many of the attendees, with internationally renowned scholars and industry experts alike visiting our team to discuss their work. The conference also afforded our researchers an opportunity to exchange ideas, techniques, and challenges with the best in their field, creating new avenues of research and potential partnerships. Our researchers will continue presenting their work at the IEEE Spoken Language Technology Workshop in Greece, so stayed tuned for more innovations from the ObEN team!