[singer36]

Hi Kempniu !
thanks for your answer. I indeed tried running it in standard user, and not root. Same error message.
arecord did work...

maybe my pocketsphinx isnot well installed.

Do you know where are the log files of pocketsphinx ?

[Kempniu]

How did you install xbmcvc and pocketsphinx? From PPA or did you compile something manually?

The error message you are getting is due to a bad return value from a call to a pocketsphinx function. pocketsphinx is a library and it doesn't write any logs, it just spits out a lot of messages, which xbmcvc suppresses for readability. If you can compile xbmcvc from source after commenting out the relevant part (lines 1327-1328 of xbmcvc.c from current master; let me know if you don't understand what I've written), you should be able to make pocketsphinx flood you with its messages - which are probably the only thing that might shed some light on the matter.

[Enoran]

I have installed xbmcvc and it's working pretty well for me. However, there is quite some delay (~2 seconds) between my voice input command and xbmc showing the result. Program is supposedly detecting the silence.
Is there some way for me to reduce this delay?

[Kempniu]

Thank you for trying out xbmcvc. Have you tried tweaking recording levels using a mixer? From what you're saying, it sounds like pocketsphinx might be unable to start processing immediately after you stop speaking because it still receives samples loud enough to suppress silence detection - try lowering microphone sensitivity.

To rule out other possible problems, you can also try using xbmcvc in test mode - if a command gets executed by your XBMC instance immediately after it is entered in test mode, it's most likely an audio issue indeed. If command execution is delayed in test mode as well, your audio setup might not be the culprit here.

BTW, which XBMC version are you using?

[Enoran]

I'm using 12.3.

Indeed, it seems like lowering the microphone does indeed shorten the delay.

Another issue I'm facing here is that I'm using a Press-To-Talk(PTT) wireless mic, so each time I input a voice command I have to keep holding down the PTT button for the silence detection.
If I release the PPT button before the command executes, the command will not be executed at all.

Any settings I can change to get by this issue? Thanks.

[Kempniu]

I sense this one's going to be tough. I have no wireless microphone to test what happens "behind the scenes" when it is used in Press-To-Talk mode. The way I'd expect such a feature to work is simply to provide the sound card with zero samples as long as recording is in progress and the PTT button is not pressed, but from what you're saying, it seems that is not exactly what's happening (in fact, if it was working that way, silence detection would work perfectly for you). Maybe your microphone doesn't produce any samples at all until PTT is pressed (power saving?) and that causes pocketsphinx to go wild. It's hard to say without getting my hands on a similar device.

Sorry to say this, but your problem is probably a dead end unless I have a similar device to test my program with. Even if that was the case, xbmcvc was specifically designed with continuous listening in mind, while your microphone seems to be following a different philosophy. It's implementable, of course, but it's not an easy fix and I'd still need a similar device in the first place.

Just out of curiosity, is it a Bluetooth microphone or does it ship with a dongle of some kind that you plug into a USB port / sound card jack?

[Enoran]

Hi Kempniu,

The mic is indeed a BT mic with a dongle and detects as USB Audio/Mic. As with all mobile devices, power management is critical so when the PTT button is not pressed, nothing is sent.

Thanks for your prompt reply.

[Kempniu]

singer36, I know it's been half a year since you've posted in this thread, but this might be useful information for others as well, so here I go: while testing xbmcvc with Gotham, I've most likely found the cause of your problems - PulseAudio. See, Ubuntu 13.04+ has an official package for pocketsphinx version 0.8 that is built with PulseAudio support and without ALSA support. What that means is if you've installed xbmcvc and pocketsphinx using apt, then behind the scenes pocketsphinx will be trying to connect to a PulseAudio server, which is normally running on a regular Ubuntu installation/live CD, but is disabled on XBMCbuntu by default. I've updated the README with information in this regard.

Based on the tests I've been doing lately with xbmcvc and Gotham, I can tell that the two seem to play nicely together, so Gotham is now supported without displaying any warning messages about experimental support. XBMCbuntu/Ubuntu packages should be ready later today.

As usual, if you come across any issues, please let me know.

[efegue]

Did anyone manage to get this working on OpenELEC?

[Kempniu]

I have never messed with OpenELEC before, but packaging xbmcvc for it will be tougher than packaging xbmcvc for Ubuntu as one also has to take care of pocketsphinx packaging, which is a mess. Though I see no reason for xbmcvc itself not to work fine with OpenELEC. If there is interest in using xbmcvc with OpenELEC among forum users, I might research the topic, but I'd have to hear from the people interested beforehand. I have a lot less spare time right now than I had back when I started developing xbmcvc and, unfortunately, I'm not that comfortable with going through all the hassle if just a handful of people will benefit from it. Yet contributions are welcome, of course

[efegue]

Thanks for your answer Kempniu.

I can have a look of the needed libraries/stuff for OpenELEC, can you give me some directions/hints? I'm not sure what's required to "run" xbmcvc.

Thank you!

[Kempniu]

It's all laid out in xbmcvc's README.

[Kempniu]

Just a heads up for everyone interested - in an effort to steer away from legacy naming, xbmcvc will be renamed to kodivc before Jan 17, 2015. While I will do my best to make the transition as painless as possible, current PPA URLs will probably become broken at some point during the next week. Shall that be the case, replacing all instances of "xbmcvc" with "kodivc" in your APT configuration files should do the trick. Needless to say, the documentation will be updated to match the state of things after the migration.

As usual, feel free to post in this thread if you encounter any problems and I'll do my best to help.

[stuCONNERS]

Did anyone ever get it working with openelec? Is it still not possible to open movies directly? Also is it possible to switch tv stations via voice control?

[Kempniu]

Did anyone ever get it working with openelec?

I really hope you have at least tried looking five posts above before asking. Assuming you have, the answer is I'm not aware of anyone achieving success in this field. Which, of course, is far from saying it's not possible.

Is it still not possible to open movies directly?

If by "opening movies directly" you mean using commands like Play The Godfather, then no, it is not possible nor will it ever be. This project is aiming for simplicity, which comes with reasonable speech recognition accuracy. It is not aspiring to become the next Siri. You should rather think of it as a voice-controlled keyboard than your personal assistant. Though if you'd like that to change, nothing stops you from forking the project - in fact, you're encouraged to do it

Also is it possible to switch tv stations via voice control?

Yes, just use NEXT and PREVIOUS.