Wednesday, April 9, 2008

The Science of Speech - 1913 Dictation Machine

This is about as steam punk as you can get.

In 1913 John B Flowers invented a machine that could convert speech to typed text. has a great write up of the invention, and it's limitations.
There were "serious limitations which must be considered" with Flowers's voice-operated typewriter, chief among them its inability to distinguish between homophones like "to," "too," and "two"...
Expanding on IO9's post, I found a paper published by Mr. Flowers in 1916. It's a dry read, but pretty incredible reading through his analysis of the human voice. With all of our fancy software algorithms and powerful computers, we're really not that much farther along.
If you're interested at all in the science behind how we make noise (recording engineers I'm looking in your general direction), then I would highly recommend reading through some of this. It's already given me some EQ ideas for my next demo production session.

You can read his paper at the link below:
Transactions of the American Institute of Electrical Engineers @ Google Books
The link should take you right to John B. Flowers' paper, but i it doesn't scroll down to page 213.

And here's IO9's post:
A Machine That Types What Is Spoken To It, 1913
someaudioguy some audio guy speech text dictation voice over production voice acting demo animation audiobook speech to text narration recording audio engineering
Picture Courtesy of

No comments:

Post a Comment