Voice controlled HTPC ?

bretttp1

Does anyone know of any voice activation software for use in a HTPC setup? I was looking at this and was just wondering would it be possible to build something similar, I know the case is a zalman.

Anyone got any ideas? or is voice recognition still not a very good interface?

kaisersose77

check this out, saw it posted somewhere else months ago, think theres a demo video on the site

http://www.sayittoplayit.com/

mathias

They are a great idea , just like voice dictation into a word processor , but like those , they still dont work too well , background noise triggers when you dont want it too , and unless you speak diction perfect american english it fails half the time , the problem is accents , it just cant handle them , you have to speak like a robot to them to get good results.
So unfortunately , though it looks cool enough , it doesnt work too well.

Tenshot

Brettp, I'm one of the guys involved with Amulet Devices, which you linked to above. Possibly you visited our stand at Toys 4 Big Boys over the weekend?

You're quite right that we use the Zalman case (HD160XT to be exact); it's the best case we've found that has both decent air flow and a respectable home cinema appearance.

The case itself isn't involved directly in the voice recognition. However, its low-noise cooling capabilities let us use a high-end CPU and graphics card (Intel Dual Core & Radeon X1600) while keeping the noise from the system at a minimum (it's essentially silent). The built-in 7" touchscreen is also useful, since it lets us provide visual feedback to voice commands when the main TV is turned off.

Having a powerful CPU is important, since voice recognition is very CPU hungry; when you give a speech command, you don't want to be waiting 4-5 seconds while the system tries to figure out what you said.

The speech recognition itself is based on Vista's built-in speech recogition system, tweaked to recognise UK english. We've been tracking the various PC speech recognition engines for a while, and Vista is the first systems we've seen that does a good job of speaker-independent recognition over a wide range of voices and input.

The other key is having a really good quality audio input source. This makes a huge difference to the accuracy of the recognition. We experimented with Bluetooth for a while (and built a number of prototype Bluetooth remote mike systems) but the quality of the speech received was too compressed to get the quality of recognition we wanted. We eventually switched to direct radio transmission, and immediately saw a big improvement.

If you want to build a system yourself, I'd recommend:

- Find some software that interfaces with Media Center, or write your own; the standard Vista recogniser will only control the normal GUI out of the box
- Make sure you have a decent input mike
- Make sure you have a decent CPU
- For bonus marks, upgrade Vista's built-in text-to-speech software to provide a more natural sounding voice, when giving vocal feedback

Results in our trials have been very good, and even with the ridiculous amount of background noise from other stands at the show, we were still getting excellent results. (Again, using the right microphone here makes all the difference.)

Eddy

subway

the whole things seems like massive overkill.
remote controls have been doing the job for years and a device that cant be used easily by anyone who picks it up doesnt really belong in front of a tv

GEORGEHenry

I was at T4BB, saw this amulet thing. It was really cool. They got people in the crowd to talk to it and it worked. I seriously want one -now just have to find the money.