Speech Device Gives Hope To Those With No Voice Researchers in Germany are developing a device that translates the movements of the face into computer-generated speech. It was demonstrated recently at the meeting of the American Association for the Advancement of Science.
NPR logo

Speech Device Gives Hope To Those With No Voice

  • Download
  • <iframe src="https://www.npr.org/player/embed/134068740/134123056" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
Speech Device Gives Hope To Those With No Voice

Speech Device Gives Hope To Those With No Voice

  • Download
  • <iframe src="https://www.npr.org/player/embed/134068740/134123056" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

STEVE INSKEEP, host:

Now we're going to hear about a group of researchers from Germany who are developing a way to turn facial movements into computer speech. Their system could make it possible for people who have lost their voices to speak, just by silently mouthing the words. Reporter Gretchen Cuda-Kroen has the story.

GRETCHEN CUDA-KROEN: Sixteen years ago, Al Keneda lost his vocal chords, and his voice, to cancer. In the beginning, he communicated with a white board and a pen, then with something called an electrolarynx.

Mr. AL KENEDA (Speaking with electrolarynx): I used this for three years before I got my voice prosthesis. It was a barrel of laughs.

(Soundbite of laughter)

CUDA-KROEN: Today, Keneda speaks a lot better. He has a voice prosthesis a small valve surgically inserted into an opening in his throat. When he speaks, he holds his thumb across the opening, and uses his breath to force air upwards through his throat and mouth to form speech.

But even this prosthesis isn't perfect. His voice is raspy, and he speaks a bit slower than he used to because he has to take a breath in between each phrase. And certain sounds still elude him, like the letter H.

Mr. KENEDA: L's, O's, I (unintelligible). Everything else you can pretty well do, but the letter H is a bother.

CUDA-KROEN: A couple of weeks ago at a scientific conference in Washington, D.C., German researchers showed off something that might work even better for Keneda - once it's perfected, that is.

German researcher Michael Wand demonstrated the system for me.

Mr. MICHAEL WAND (Karlsruhe Institute of Technology): Let us have dinner at the pub.

Mechanical Voice: Let us have dinner at the pub.

Mr. WAND: A good time would be 7 o'clock.

Mechanical Voice: A good time would be 7 o'clock.

CUDA-KROEN: For his Ph.D. thesis at the Karlsruhe Institute of Technology, Wand developed a way to translate the movements of the face into speech. Right now, the device is a half-dozen or so wires connecting electrodes on Wand's face to a computer. The idea is that a person can silently form the words they want to say - not actually speak them, as Wand had to do when demonstrating for radio -and the computer interprets the muscle movements and turns them into sound. But there still a lot to do. For example, what girl would say yes to a pub dinner with a guy who has electrodes all over his face?

Wand agrees and says one day, the electrodes and the software will be completely integrated into a smartphone that could be held discretely to the side of the face and produce fluent speech. Wand says the system currently recognizes about 2,000 words and operates with 90 percent accuracy. Good enough for everyday speech, he says - well, at least most of the time.

Mr. WAND: My name is Michael Wand.

Mechanical Voice: I would be glad to.

(Soundbite of laughter)

CUDA-KROEN: Some of this works.

(Soundbite of laughter)

Mr. WAND: Yeah. Let's - another try. So - the recognizer interprets my muscle movements.

Mechanical Voice: The recognizer interprets my muscle.

Mr. WAND: Therefore, I can talk to you by simply mouthing words.

Mechanical Voice: Sure you can get my data.

(Soundbite of laughter)

CUDA-KROEN: OK, so not perfect.

(Soundbite of laughter)

Mr. WAND: I have no idea what's happening. Maybe it's getting a bit late - electrodes falling off.

(Soundbite of laughter)

CUDA-KROEN: Wand tells me that sometimes, smiling and other non-verbal facial movements can confuse the system. But who knows? Maybe it's good enough to be used by reporters with stuffy noses.

Mechanical Voice: For NPR News, I'm Gretchen Cuda-Kroen in Cleveland.

Copyright © 2011 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.