报告人简介：

Sadaoki Furui，芝加哥丰田技术研究院院长、日本东京工业大学教授、亚太信号与信息处理协会主席。在语音信号处理与通信技术领域做出杰出贡献，发表论文700余篇，出版数字语音处理等著6部，在信号处理与通信领域有重大的国际影响力。先后获得IEEE ASSP协会最高奖、ICICE成就奖、IEEE信号处理协会最高奖。

报告摘要：

This talk overviews how DNN-based technology has changed human-computer interactions, mainly focusing on speech recognition and speech synthesis technology, and explains why DNN is helpful.  Using various DNN-based technology, state-of-the-art speech recognition systems can achieve almost the same performance as human beings, and speech synthesis systems can produce natural voice with various voice characteristics.  These days, not only the voice, but also a face image that looks exactly like someone, a natural face picture of a person who does not exist, and a video as if President Obama is talking, can be made.  It gets harder to distinguish which is the natural one.  It is getting harder for us to trust our eyes and ears.  Although there are still a lot of things that current AI cannot yet do or is not yet good at, AI definitely will develop rapidly from now.  It will be extremely important to demonstrate human wisdom and intelligence to control the usage and ensure responsible applications, avoiding the harmful or even unintended outcomes.

This talk also includes some recent speech recognition research which has been conducted at TTIC.