**Deep Brain AI virtual human technology recognized at the most prestigious academic conference**
Deep Brain AI, a company specializing in artificial intelligence (AI), announced on the 2nd that a paper submitted by its artificial intelligence research institute was accepted at the world’s most prestigious International Conference on Acoustics, Voice and Signal Processing (ICASSP).
ICASSP, which celebrates its 48th anniversary this year, is the world’s largest international conference in the field of sound, voice, and signal processing hosted by the Signal Processing Society of the International Association of Electrical and Electronics Engineers. The theme of ICASSP this year is ‘Signal Processing in the Age of Artificial Intelligence’, and more than 6,000 papers were submitted, almost twice as many as last year.
The papers of Deep Brain AI adopted by ICASSP this time are two studies dealing with video synthesis and voice synthesis, respectively. Deep Brain AI will participate in ICASSP 2023 to be held in Rhodes Island, Greece in June of this year and introduce its papers to famous scholars and corporate researchers around the world.
The video synthesis thesis deals with ‘DisCoHead’, a technology that separates and controls head movement, lip sync, and expression through a geometric transformation bottleneck. We introduce a method of subdividing the body parts of the head and face so that the upper body image naturally follows the desired voice and video. This technology, which succeeded for the first time in the world in Deep Brain AI, can freely adjust the posture of the upper body, speech motion, and non-speech expression, making it possible to create a high-quality virtual human with a much higher degree of freedom than conventional methods.
In the speech synthesis paper, we dealt with context modeling techniques for Chinese grapheme-phoneme conversion. This study proposes a ‘context reinforcement technique (Reinforcer)’ that enhances the accuracy of Pinyin conversion of the next character by emphasizing the relationship between adjacent characters during Chinese Pinyin conversion.
In addition, Deep Brain AI has applied for or registered 148 domestic and foreign intellectual property rights in the field of AI image synthesis, and papers related to deepfake detection and speech synthesis have been listed in the International Computer Vision Society (ICCV) 2021 and Interspeech 2021.
https://www.sedaily.com/NewsView/29MVC13Z87
https://www.sedaily.com/NewsView/29MVC13Z87