Outline

We are studying speech recognition as a means of creating closed-captioning in live information programs and transcribing comments in video footage. This exhibit demonstrates our technology for recognizing information programs containing many indistinct utterances in real time, and for efficiently producing annotations in video footage.

Features

Real-time speech recognition in information programs international standards

In information programs that cover a broad range of subjects and feature guest speakers offering commentaries on these subjects, the speech often includes background noise and inarticulate utterances, which can cause problems in speech recognition. Using a deep neural network^※ , we can perform accurate speech recognition to generate subtitles to this kind of information program in real time.

Rapid production of transcripts from video footage international standards

For program production based on a wide range of video footage, it is essential that comments included in the video footage can be transcribed rapidly. We are developing a system that uses speech recognition technology to produce transcripts from video footage both quickly and efficiently.