It’s surprising that no one has created auto-subtitles for video calls as a plugin for Skype or other conference call software. Besides the huge market of hard of hearing or deaf people who cannot perceive audio but can read subtitles, there is also a need to have transcripts for calls to facilitate quick searching. It is necessary to decipher the voice from each participant separately, so that simultaneous or overlapping dialogues do not interfere with each other. In theory, this could also be done after the call, but by then the voices are mixed together, and the quality would be lower.
For corporations, it’s possible to train the recognizer on specific terms so that it does not stumble over them. The accumulated database for more accurate recognition could even be shared with other companies in the same market, if the client agrees to share.
The next step would be recognizing sign language and lip movements by the same Skype to improve the quality of speech recognition (if there is speech) or gestures (if they are present).
Hi, I'm Rauf Aliev. 