4 ms·

Several years ago I was trying to solve the problem of voice recognition in a noisy environment. After brainstorming with some colleagues, we came up with the

Several years ago I was trying to solve the problem of voice recognition in a noisy environment. After brainstorming with some colleagues, we came up with the idea of using a throat mike. The only problem was that we couldn't find reasonably priced hardware that would work with commodity computers.

jerryr 10y ago
Jawbone Bluetooth headsets actually contact your jaw to use the conducted vibrations to cancel background noise, similar to your idea. Disclosure: I worked on their first generation Bluetooth headset. I'm not certain the current generations still work this way.
- Karunamon 10y ago
  Noiseassassin - yup, the later gen ones do as well, but as I understand it, Jawbone is basically dead at this point.
  - Already__Taken 10y ago
    Could get a second wind with voice recognition because the problem with comms is everyone sounds the same through that style of mic.
- abricot 10y ago
  I had that one. It was really great in noisy environments... unless there were also wind. Wind outright killed the sound.
M_Grey 10y ago
What really needs to happen is for someone to figure out how to decode the electrical signals generated by subvocalization, train a DNN to interpret them, and use that.
- xgen 10y ago
  I don't think that is practical without invasive surgery. Afaik the best we can do with external monitoring is roughly tell whether a person is trying to move or thinking about moving. Reading thoughts doesn't seem realistic with current technology.
  - M_Grey 10y ago
    If you're trying to tap the brain that's true, but if you could readout the muscular contractions around the vocal chords, you could translate that instead. It's definitely easier to understand the simple output at the level of muscles, than reading "thoughts" in the brain. There are already some prosthetic limbs which used implants to read muscular contractions to control the prosthetic, but those are far too crude to interpret speech.
hatsunearu 10y ago
What the heck? I found cheap ones on Amazon designed for the Baofeng ham radios. Surely Shenzhen has figured it out.

jerryr 10y ago

Jawbone Bluetooth headsets actually contact your jaw to use the conducted vibrations to cancel background noise, similar to your idea. Disclosure: I worked on their first generation Bluetooth headset. I'm not certain the current generations still work this way.

Karunamon 10y ago
Noiseassassin - yup, the later gen ones do as well, but as I understand it, Jawbone is basically dead at this point.
- Already__Taken 10y ago
  Could get a second wind with voice recognition because the problem with comms is everyone sounds the same through that style of mic.
abricot 10y ago
I had that one. It was really great in noisy environments... unless there were also wind. Wind outright killed the sound.

M_Grey 10y ago

What really needs to happen is for someone to figure out how to decode the electrical signals generated by subvocalization, train a DNN to interpret them, and use that.

xgen 10y ago
I don't think that is practical without invasive surgery. Afaik the best we can do with external monitoring is roughly tell whether a person is trying to move or thinking about moving. Reading thoughts doesn't seem realistic with current technology.
- M_Grey 10y ago
  If you're trying to tap the brain that's true, but if you could readout the muscular contractions around the vocal chords, you could translate that instead. It's definitely easier to understand the simple output at the level of muscles, than reading "thoughts" in the brain. There are already some prosthetic limbs which used implants to read muscular contractions to control the prosthetic, but those are far too crude to interpret speech.

hatsunearu 10y ago

What the heck? I found cheap ones on Amazon designed for the Baofeng ham radios. Surely Shenzhen has figured it out.