Question about Voice Commands

I did a search and it returned lots and lots of pages and I’m not really sure where best to start. I figure someone can read my question that has experience in this area and pretty quickly tell me if this is possible and point me in the right direction.

I’d like to have a device that can be plugged into power that is constantly “listening” for a key word (like “Vera”, or “Hippopotamus”, or whatever) and then a follow up command (like “living room main lights on”). I don’t want to have to push any button or manually do anything for the device to start listening. Is this currently possible? I have iOS devices but I’m fine with having to buy an Android device or something else.

If this is possible, can someone point me in the right direction with a thread or device/app suggestions?

Much appreciated!

Well, the answer is yes. AutHomation (Android) can do this. The current alpha version has this functionality provided it is the app in focus at the time. You need to say “OK Vera” to fire up the SR functionality. But the usability is hit or miss with SR technology at this point in time. Fairly warned be ye.

Thanks S-F. I’ve done a lot of reading and watching videos since posting the questions. It looks like there are a few apps that have this ability (kinda) (AutHomation, autovera/autovoice, HAL, etc…).

A few more questions. What kind of range do these apps have for the most part? Can I put the tablet on a table in my living room and loudly speak/shout commands at it from my kitchen 30 feet away? Or do I need to have the device right in front of me for the most part?

Also, it looks like there aren’t really any iOS apps with this function so I’d need to get an Android device. Being completely new to Android I have a couple more questions. It looks like most of these apps have to have the app open in order for it this to work (voice commands without pushing a button). Can I just go into the settings somewhere and set the tablet to never go to sleep/shut off? Open up the app and it would just stay on it forever? I’d have it permanently plugged into power obviously because the battery wouldn’t last long. Maybe go to a display setting and turn the brightness all the way down or even turn the display off if possible so I don’t burn out the display too quickly. Does anyone have a device they have set up like this?

Will just about any Android device work? Do they all have mics for the most part and access to the Google App Store? Or are there some I need to avoid that don’t let you install 3rd party apps? Any reason if I’m only using it for this single purpose that I shouldn’t go with a very cheap $60 to $100 Android tablet?

Whew, lots of questions…sorry.

OK, to answer some of my own questions in case anyone is following this topic or for future searches…

I went and purchased a cheap ($100) HP 7" tablet and Android 4.1 running on it.

There is not, at least on my device, an option for the device to never go to sleap (the longest setting is 30 minutes). I’ve seen apps that will allow devices to never sleep so I assume this is an easily fixable “issue” but I have not tried them out yet.

Thus far, I’ve only tried out HAL and AuthomationHD. HAL doesn’t seen to have a “hotword” feature so it’s not terribly relevant to my needs at the moment. Authomation does have this feature and it works fairly well if the device is right in front of me. I can not use the voice feature if I am more than about 5 feet away from the device and certainly not from another room. This could be specific to my device’s mic but I suspect it is the same for pretty much any device. A few possible solutions:

  1. Get more tablets and have one or two for each room that I want voice control.
  2. Get an external higher quality mic that can be plugged into the tablet that may be able to pick up sound from much further away.
  3. Get a bluetooth mic that can connect wirelessly to the tablet and clip it onto my shirt while at home.

Haven’t tested any of these possibilities out yet but I will and will update as I do.

If anyone is currently doing something similar to what I am trying to do I’d love to hear from you about your setup.

I’m very interested in this functionality as well. I believe the voice recognition in AuthomationHD (which is my favorite Android app), relies on the built in Google voice recognition engine.

Ultimately, I’d like to have a wall-mounted tablet in my living room to control my Vera. I hadn’t though about a bluetooth mic, I’ll have to try that out.

@ justauser,

Sorry about the delay in responding. It looks like you’ve figured out most of what I’d have told you any how.

Some thoughts though.
Phones have better microphones than tablets. The accuracy of my phone blows my tablet out of the water!
I think that if there were more interest in HAL RTS would add more functionality. I think that ATM not many people are taking SR too seriously so using it in AutHomation is a nice side benefit.
I can’t really comment on the BT mic idea so much. I do myself use something like that (Clip on to shirt with attached ear buds) with my phone and SR seems OK but not as good as with the built in mic on the phone. Other than a very limited amount of usage though I’m in the dark there.
As for the external mic idea: I think that this is the way to go for something like you’re interested in. I have seen people spend a small fortune and hundreds of hours trying to get this kind of thing working. There are several wall mounted mics that people spread throughout their houses, run through some kind of mixer and interface to their input device. If you’re interested in any of the specifics of the hardware just let me know. I myself have toyed with the idea of going this route as I’m a really big fan of SR but I’ve decided to put this project on hold until the SR technology in general matures a bit more.

All that being said, I’ve been using SR for writing papers for school, instant messaging (now text messaging), writing emails, writing forum responses and so forth for about 8 years. I don’t think that the accuracy has improved very much when working with a well trained setup. My experience is that with good conditions and good hardware I can dictate much more accurately than I can type. I haven’t tried using top notch hardware with Google’s SR but it seems to work very well with no training at all so I imagine that if you give it really good material to analyze you will get phenomenal results.

If you have any interest in any of the points I have brought up just let me know and I will gladly elucidate.

Oh, and one more thing. Android for x86… I would love to be able to run an Android VM to do things like run Vera Alerts for my whole house audio announcements and for centralized SR, like you’re thinking of.

Both HAL and AutHomationHD rely on Google’s speech recognition. The best results are achieved when you are close to the device. This is due to the mic and your surrounding environment. Voice Recognition is more of a added bonus in AutHomationHD. I am not expert in this field and it can be very complex.

  • Garrett

[quote=“S-F, post:6, topic:178266”]…
As for the external mic idea: I think that this is the way to go for something like you’re interested in. I have seen people spend a small fortune and hundreds of hours trying to get this kind of thing working. There are several wall mounted mics that people spread throughout their houses, run through some kind of mixer and interface to their input device. If you’re interested in any of the specifics of the hardware just let me know.
…[/quote]

Hi S-F,

I’d be interested if you could expand a bit on wall mounted mics you mentioned and what kind of input device would be used in that case. Are these mics wireless? Or would wires need to be run through the walls? I actually have a mixer and a nice shotgun mic that I used to use for some voice over work. I’m just not sure if there are other device other than a computer I should be looking into for input.

Most (or all) tablets or phones I’ve looked at only have a headphone jack so I’m not sure how to use an external mic with these. When I mentioned the idea I thought I’d be able to find some devices that may take a mic but haven’t been successful. I did buy a bluetooth headset with mic but I can’t seem to get that to work with my tablet. Not sure if there’s a setting other than simply pairing with my tablet but I don’t know how to tell Authomation to listen to my bluetooth headset and not the tablet’s mic.

Android for x86 sounds interesting. I’m looking into installing android via virtualbox right now so I’ll see how that works out. I don’t really want my computer running constantly (which is why I wanted to go a tablet/phone route) but we’ll see what happens.

Speech recognition (SR) needs a large Signal to Noise Ratio (SNR) for accurate recognition.
And even good recognition, without user training, is no where near 100 %.
Training techniques are used by various software packages and can increase recognition for a given SNR.
High accuracy is obtained when only a limited grammar needs to be recognized, in that case the speech recognition engine can do a best guess given the voice record and the allowed grammar. With Android/Google recognition we are using a grammar-less recognition.

You are not going to get the good SNR on anything other than a PHONE at this time.
The phone companies have done a lot of work to put active noise canceling into the phone. (This is on the Digital Signal Processing (DSP) chips associated with the Phone part of the Smart Phone. Their motivation was clear sound on the telephone. But has been very helpful for SR.

When you are close to the phone … The front and rear microphones can differentiate between a Voice source and background noise. This is used by the DSP to actively subtract the noise from the voice source. (Sounds easy but it’s sophisticated DSP processing).

With a microphone at a distance … their is no real way to differentiate voice source from noise source. There are some active beam forming (multiple microphones) mikes that track the largest sound, and try to subtract the sound from the other mikes from this. They are an improvement. But sometimes they can track the noise source if it is louder than the speaker.

A blue-tooth headset is virtually use-less unless it has the noise cancelling in the Headset. I do know know of any that cancel the noise while speaking. I have a nice pair of Sennheiser MM-550X blue tooth headset that cancels the background noise for the listening to music. But it does not help with voice enhancement when I talk. I often have to disconnect the head set and talk directly on the phone. The technology is in the head set for the receive channel … they just did not anticipate the need on the voice channel.

So unless you use a phone - I think you will be disappointed.

Just as Scottie was:

Thanks for all the info Richard! I’ll continue playing around with different devices and see where it gets me. I know now to lower my expectations for the time being though which is a bit disappointing but it’s reality.

I think I’ll scour Craigslist for some used cell phones and experiment with distances so that when I speak I don’t get multiple phones picking it up.

The mixer I was talking about is the ClearOne Genter AP800. They’re all over Ebay. Check 'em out. In this case it would just run to an Android device. You’d probably want to use a phone as I think most tablets don’t have the second ring to accommodate a mic. The mics are Crown PMZ wall mounted units. They requite that you run wires. I don’t think there is currently anything wireless that will fit the bill. Like I said, people spend a lot of time and money trying to get a Star Trek like experience with their systems and usually end up frustrated to one degree or another.