4 ms·

I applied to YC with a speech interface for drones, one application of which would be hands-free operation of selfie drones like this. E.g., if you're mountain

I applied to YC with a speech interface for drones, one application of which would be hands-free operation of selfie drones like this. E.g., if you're mountain biking you probably don't want to take one hand off the handlebars and try to press the button on the remote that means "Switch to leading shot". Instead, you'd just talk into your earbud: "Switch to leading shot" or "Climb to 100 feet", or more complex cinematographic instructions like "Pan from me to tracker B" Especially in the extreme sports selfie video genre, hands-free operation would be a big advantage.

crimsonalucard 11y ago
Reliability needs to be way higher then what we're currently seeing with google now and siri due to the inherit danger of what ur controlling.
- jjwiseman 11y ago
  For specific domains, like drone control, you can achieve much better accuracy then you typically see in "dictation"-style speech recognition. You can use a statistical language model that represents the things you're most likely to hear. For example, Google & Siri kind of need to be able to handle anything I throw at them: "What is Ke$ha's new album?" "What year was the Hardy Boys book 'Hunting For Hidden Gold' written?" They may use a language model that favors grammatical language, that is: "What is Ke$ha" is a more likely speech recognition hypothesis than "What hiss kush ball", but they still need a big model to represent that. For drone control, you have much more constrained language, which helps recognition accuracy significantly. The model can tell the recognizer that if it heard "Go <unsure> 100 feet" that the <unsure> word is most likely to be a direction like forward/back/left/right/up/down, and not "neutrino". It's a lot like the way that Norvig uses n-grams to illustrate writing a spelling corrector: http://norvig.com/ngrams/ch14.pdf http://norvig.com/ngrams/ch14.pdf Having a model lets you fix errors in the input. Having constrained language and a good model is often critical to creating a successful speech interface.
jdiez17 11y ago
Interesting idea. I think the biggest challenge would be to filter out noise (i.e wind blowing on the microphone). But in a limited domain such as this you can do aggressive filtering and get fairly good results in very harsh environments. Really cool idea. Now I have a strange urge to break out the ArduCopter.
- jjwiseman 11y ago
  I have an early test at https://www.youtube.com/watch?v=uhBa11gdbeU https://www.youtube.com/watch?v=uhBa11gdbeU

crimsonalucard 11y ago

Reliability needs to be way higher then what we're currently seeing with google now and siri due to the inherit danger of what ur controlling.

jjwiseman 11y ago
For specific domains, like drone control, you can achieve much better accuracy then you typically see in "dictation"-style speech recognition. You can use a statistical language model that represents the things you're most likely to hear. For example, Google & Siri kind of need to be able to handle anything I throw at them: "What is Ke$ha's new album?" "What year was the Hardy Boys book 'Hunting For Hidden Gold' written?" They may use a language model that favors grammatical language, that is: "What is Ke$ha" is a more likely speech recognition hypothesis than "What hiss kush ball", but they still need a big model to represent that. For drone control, you have much more constrained language, which helps recognition accuracy significantly. The model can tell the recognizer that if it heard "Go <unsure> 100 feet" that the <unsure> word is most likely to be a direction like forward/back/left/right/up/down, and not "neutrino". It's a lot like the way that Norvig uses n-grams to illustrate writing a spelling corrector: http://norvig.com/ngrams/ch14.pdf http://norvig.com/ngrams/ch14.pdf Having a model lets you fix errors in the input. Having constrained language and a good model is often critical to creating a successful speech interface.

jdiez17 11y ago

Interesting idea. I think the biggest challenge would be to filter out noise (i.e wind blowing on the microphone). But in a limited domain such as this you can do aggressive filtering and get fairly good results in very harsh environments. Really cool idea. Now I have a strange urge to break out the ArduCopter.

jjwiseman 11y ago
I have an early test at https://www.youtube.com/watch?v=uhBa11gdbeU https://www.youtube.com/watch?v=uhBa11gdbeU