Coming soon to an Android near you! The first demo of using an Android as a microphone for VoxCommando with our special app that taps into the accelerometer to turn the mic on when you lift it.
http://www.youtube.com/watch?v=vfd1btCsOmo&feature=youtu.be
EDIT: get the code, report bugs, request features etc. here: http://code.google.com/p/voxwav/
Cool James, congratulations, well done.
Thanks. So far so good. Though it has been a lot of work. It is always fun to learn new stuff, especially when you are able to accomplish something.
So now I am thinking I may either have two apps, or possibly two modes in a single app.
one for sending audio, and one that does recognition using google and then sends the text to voxcommando (or does other things directly)
Quote from: jitterjames on July 24, 2012, 10:01:41 PM
Coming soon to an Android near you! The first demo of using an Android as a microphone for VoxCommando with our special app that taps into the accelerometer to turn the mic on when you lift it.
http://www.youtube.com/watch?v=vfd1btCsOmo&feature=youtu.be
congratulations, seems better than "Mobilio headset" for iOS. :clap
Thank You.
Cant wait to test this some day.
Cool! :clap
That's great James!
I'm interested in your comment about 2 modes. Are you thinking that one would use Google for recognition and the other just sending audio to VC? Have you figured how to overcome the WiFi quality issues?
I have both a ICS phone and tablet so I'm hoping for an early XMAS present ;D
Quote from: DHHJ on July 25, 2012, 10:46:04 AM
I'm interested in your comment about 2 modes. Are you thinking that one would use Google for recognition and the other just sending audio to VC? Have you figured how to overcome the WiFi quality issues?
Yes. In some cases you may want to automate some tasks directly from the tablet. Perhaps, just as an example, your VC machine is sleeping. It would be nice to use a voice command to wake it up. Or maybe you want to adjust the screen brightness of your tablet. There is really no need to use VC on the PC for this. There may also be situations where you want to use multiple languages. Maybe you use English engine on the PC-VC but you want to be able to do web searches in German. In that case you could use the German recognizer from Google and send text results. There are a lot of possibilities and of course it will take a long time to sort them all out into something that makes sense and can adapt to different users's needs. My head is flooded with ideas right now so I am just trying to concentrate on the basics to get something out the door that people can test.
There is no wifi issue now. We are transferring an uncompressed 16 bit 44KHz mono wav from the android to the PC using Samba file sharing. That means that you will need to enter a network path, and credentials to allow you to access the files on the PC that is running VC.
I am having some sound issues that I can't explain but they are more related to my lack of understanding about Wav files (and perhaps the android hardware) than to WIFI. Sound quality seems to be excellent on my tablet, but there is often static at the beginning of the wav file and sometimes at the end. I'm not sure where it is coming from, or how to get rid of it. I imagine I will sort it out eventually, but as you can see from the demo it may not be an issue for some users. it seems to be working reliably from my tablet.
The only actual WIFI related issue that I am having, which may vary from one device (or android OS) to another is that (on my ICS tab anyway), I can't send wifi when the screen is off and I have no way of turning the screen on either. This stuff seems to be locked down by google, at least on ICS. Actually I can send over wifi when the screen is off, but it is
super slow. It is possible that this is due to the CPU being in a low power state, but I am pretty sure it is because the WIFI is in a lower power state. I have the exact same problem trying to stream audio from my android to my Onkyo. Works fine with the screen on, but when the screen goes off, the audio immediately starts to stutter.
Quote from: DHHJ on July 25, 2012, 10:46:04 AM
I have both a ICS phone and tablet so I'm hoping for an early XMAS present ;D
It will be a very early present, though probably a buggy one. Stay tuned because I am hoping to release some tests very soon. :)
Cool! Can't wait to try this :)
check your emails. If you did not receive the app contact me, and make sure to install 0.956
:D
Will test VoxMic as soon as I receive the app.
I'm guessing that this solution will not have any "wake-up delay" issues like the Amulet remote. That one irks me everytime I lift the remote and speak into it, only to find out that my command hasn't fully been transmitted to VoX!
Dude. You really are a genius of epic proportions. I will be trying this tonight if I get back from Band Practice with enough time.
Will report back.
If it's helpful I can try it on Galaxy s2 with Resurection remix ROM, HTC DesireHD with MIUI ROM, and HTC desire (Bravo) with MIUI ROM. is that helpful or have you got it covered?
-P
Quote from: ajith on July 26, 2012, 08:23:59 AM
Will test VoxMic as soon as I receive the app.
I'm guessing that this solution will not have any "wake-up delay" issues like the Amulet remote. That one irks me everytime I lift the remote and speak into it, only to find out that my command hasn't fully been transmitted to VoX!
OK, you should have it. In my tests so far the response is pretty much instant from lifting to speaking. After you speak and lower the android, there is a small delay, while it copies the file across the network. The time it takes is proportional to the the amount of time that the mic was on. I am using wireless g and it is still very fast. N might be faster.
I will probably include an option to record at either 22khz or 44khz. For now it's 44.
Quote from: MrWolf on July 26, 2012, 09:19:38 AM
If it's helpful I can try it on Galaxy s2 with Resurection remix ROM, HTC DesireHD with MIUI ROM, and HTC desire (Bravo) with MIUI ROM. is that helpful or have you got it covered?
I am new to android. The more devices and is versions we can test on the better. :bignod
By the way, the current minimum for VoxWav is android 2.2
Only one person surveyed had 2.1 or lower. Hopefully that person either had multiple androids, or is able to upgrade.
I just tried it, following the instructions in the email you sent, James.
The app isn't able to see the shared folder, despite me being able to access the share from another machine, using the username and password I had setup. It gives a "Unable to copy wav to shared folder. Please check your settings. The network name cannot be found." error.
Edit: I deleted the share and recreated it with another name (changing the settings in voxwav) and tried with another user who has no password set (so, left the password field blank in voxwav), and it now gives a "logon failure" error. Is a password mandatory? What am I doing wrong?
One thing I observed is that it took a while, close to 8 seconds for the logon prompt to come up when I tried to access the share from the other machine. So could this be a timeout issue in voxwav?
Quote from: ajith on July 26, 2012, 11:40:57 AM
I just tried it, following the instructions in the email you sent, James.
The app isn't able to see the shared folder, despite me being able to access the share from another machine, using the username and password I had setup. It gives a "Unable to copy wav to shared folder. Please check your settings. The network name cannot be found." error.
- Make sure you have WRITE access with the user account that you are using. Test that you can write to the folder from another machine
- check all your settings again, and make sure you got the path right (no colons and use forward slashes, not backslashes)
example of good settings:
host: 192.168.1.105
path: c/wavwatch
user: james
password: 123mypassword!
it's not very robust yet so something like an extra slash at the end of the host or path could probably mess your up.
I've tested on two (very different) Android devices with no problems so I think it is likely that you've missed something in the setup.
- What are your Android specs (model / os version)?
- use an IP address for host (but also try using the computer name I guess)
- Are you using your new windows box or a virtual machine?
- try using the app called ES File explorer to access the shared folder and write to it.
- maybe try diabling your any firewalls temporarily?
Also, you can look in your /mnt/sdcard folder to make sure it is saving the wav locally first. It will be the default folder you see when you open ES File Explorer and the file is called vctemp.wav
If you want me to take a look so we can figure it out together, I'd be happy to using skype and/or teamviewer
Quote from: ajith on July 26, 2012, 11:40:57 AM
Edit: I deleted the share and recreated it with another name (changing the settings in voxwav) and tried with another user who has no password set (so, left the password field blank in voxwav), and it now gives a "logon failure" error. Is a password mandatory? What am I doing wrong?
One thing I observed is that it took a while, close to 8 seconds for the logon prompt to come up when I tried to access the share from the other machine. So could this be a timeout issue in voxwav?
Yes it is possible that a username and password is mandatory. I have not tested without.
Yes it is possible that it is timing out if it takes too long. But I'm not sure why that is happening. For me it's practically instant.
Quote from: jitterjames on July 26, 2012, 12:02:40 PM
- Make sure you have WRITE access with the user account that you are using. Test that you can write to the folder from another machine
- check all your settings again, and make sure you got the path right (no colons and forward slashes)
Did that! The "Save path on host" value is "d/wav" without the quotes. My shared folder is D:\wav
Quote from: jitterjames
- What are your Android specs (model / os version)?
Samsung Galaxy S2, Android 4.0.4, from here: http://forum.xda-developers.com/showthread.php?t=1436854
Quote from: jitterjames
- use an IP address for host (but also try using the computer name I guess)
Tried both ways!
Quote from: jitterjames
- Are you using your new windows box or a virtual machine?
You remembered! :) Yes, VoX and the shared folder are on the new Xtreamer HTPC box running Windows 7
Quote from: jitterjames
- try using the app called ES File explorer to access the shared folder and write to it.
ES could see the share and I could create a file in the shared folder!
Quote from: jitterjames
- maybe try diabling your any firewalls temporarily?
Turned off Windows Firewall
Quote from: jitterjames
Also, you can look in your /mnt/sdcard folder to make sure it is saving the wav locally first. It will be the default folder you see when you open ES File Explorer and the file is called vctemp.wav
It is there! It played my last command :)
Quote from: jitterjames
If you want me to take a look so we can figure it out together, I'd be happy to using skype and/or teamviewer
That would probably be the next step, after this!
I am currently on skype. you can use either voxcommando or jitterjames to find me.
Thanks to James, the trouble I was facing was identified.
My shared folder was D:\wav
In Voxwav's settings, I had specified the Save path on host as d/wav. This didn't work.
It works when this was changed to just wav
Explanation:
For d/wav to work, both D: as well as wav need to be shared. But I had only shared wav. So it didn't work.
Since I only shared wav, I need to have just wav as my Save path on host
Quote from: ajith on July 26, 2012, 01:53:45 PM
Thanks to James, the trouble I was facing was identified.
My shared folder was D:\wav
In Voxwav's settings, I had specified the Save path on host as d/wav. This didn't work.
It works when this was changed to just wav
Explanation:
For d/wav to work, both D: as well as wav need to be shared. But I had only shared wav. So it didn't work.
Since I only shared wav, I need to have just wav as my Save path on host
Thanks ajith, this works for me too :bonk
Hi James,
I've tested it with my Galaxy S2 with Android 4.03
1) for network path I used the same methode on Android as ajith (wavewatcher for the path: c:\wavwatcher) all other things do not work.
2) recording and send with the button work only fine with my headset. The onboard mic in the Galaxy is waste :'(
3) "tilt to speak" doesn't work on the Galaxy S2 - VoxWav send a big file to the wavwatch folder >2 MB without content and VC show this in the upper field "can't open wav file: c:\wavwatch\android8_19_55.wav" and create in the historie window a event "Wav.FailedRead".
Quote from: Kalle on July 26, 2012, 02:40:56 PM
Thanks ajith, this works for me too :bonk
and for me 8)
My first impressions and ideas after some tests on my Galaxy Nexus (Android 4.1.1) and Asus Transformer TF101 (Android 4.0.3):
- the tilt sensor is very responisive, it happend several times that the mic turned on and instantly off. I don't think that all devices behave equal, maybe you could add add a setting to "calibrate" the sensor
- i tested both of my devices with and without two different headsets and i don't get good results. (without headset: ~60-70%, with headset ~70-80%, kinect ~90%).* Would it be possible to have a different required confidence for .wav-files?
- i personally think it would be nice, if commands from the phone wouldn't need a prefix. i know i could use Mic.on and mic.off states to change from standby to on, but this would also enable my other mic.
- it would be great if there was an option in your app, that VoxMic is only enabled if a user chosen app has focus, or maybe a plugin for tasker to enable/disable the background service, or at least that the disables the service when i get an incoming call.
- sometimes i get a Wav.FailedRead error
* tested again with mey nexus without headset and got ~75-80% in two of three tries.
For the first version it leaves a very good mark! :clap
Quote from: Kalle on July 26, 2012, 03:24:38 PM
2) recording and send with the button work only fine with my headset. The onboard mic in the Galaxy is waste :'(
did you experiment with holding the phone at different distances from your mouth? Try 10cm, 30cm, 60cm.
Those of you getting the error in VC: " Wav.FailedRead"
I think this is probably happening mostly when the wav is too long, but possibly also when the wav is too short, or if two wav files s are sent very close together because of the tilt being too sensitive. Please test and confirm.
The very short files ~ 80 kB seem to be created due to sensor "noise" at the "crossover". Probably some kind of damping or hysteresis needs to be applied for the sensor readout.
I cannot explain the sometimes large ~ 10 MB in some cases - files that created and copied. Once this happens, it fails continuously thereafter for a while. These large files do not seem to contain any recording (only some hissing), but they do play in WMP, showing it is ~ 2 min long. And I don't recall recording that long either!
For the record, I find the recording quality with the onboard mic on my Galaxy S2 quite good, giving me > 90% accuracy in recognition. Far from "waste" :)
Quote from: jitterjames on July 26, 2012, 03:53:39 PM
Those of you getting the error in VC: " Wav.FailedRead"
I think this is probably happening mostly when the wav is too long, but possibly also when the wav is too short, or if two wav files s are sent very close together because of the tilt being too sensitive. Please test and confirm.
Yes James, I can confirm that - I used the phone to close to my mouth without a result, but if I hold it 30 cm in distance it work good.
Sometimes it creates very big files with no content 5-10 MB. I will test more what goes on ;)
Quote- the tilt sensor is very responisive, it happend several times that the mic turned on and instantly off. I don't think that all devices behave equal, maybe you could add add a setting to "calibrate" the sensor
Yeah, I'll be working on this.
Quote- i tested both of my devices with and without two different headsets and i don't get good results. (without headset: ~60-70%, with headset ~70-80%, kinect ~90%).* Would it be possible to have a different required confidence for .wav-files?
It is quite possible that if you are using a profile that was trained with Kinect it is muddying the waters because the sound on the kinect is actually kind of weird. you can't actually train with wav files (which really sucks - ** see note below) but depending on your accent you could try working from either a blank profile, or one that was trained with a good headset, or other very clean mic. Of course it is also possible that your headset or the mic on your phone are not that great. it is also possible that the software is not obtaining optimal quality, but I get excellent results with my tab using it as shown in the video. 90+ most of the time.
Quote- i personally think it would be nice, if commands from the phone wouldn't need a prefix. i know i could use Mic.on and mic.off states to change from standby to on, but this would also enable my other mic.
they don't (or should not) in fact you don't even need to have VC on. did you upgrade to 0.956? Yes I think we should be able to set the required confidence independently for wav files. I will do that at some point.
Quote- it would be great if there was an option in your app, that VoxMic is only enabled if a user chosen app has focus, or maybe a plugin for tasker to enable/disable the background service, or at least that the disables the service when i get an incoming call.
sure, but that's going to take a while. No idea how to do that yet, and many other things to do first.
** actually you CAN train with wav files. All you need to do is set the default mic to "what you hear" and then play the wav files with windows media player (or whatever). In order to get the wav files from your android, the easiest way is to use VoxWav when VoxCommando is not running, and then just go look in the wavwatch folder.
James, you could refactor the settings screen to accept the Samba share details in just one field, with the format "//<hostname>/<share-name>", as is always done in UNIX and other similar OSes. For accessing a Samba share, the drive and actual path information is not required. This will standardize the UI and also avoid the kind of issues we faced and fixed via Skype :)
You still need to get the share name right ;)
If you share the whole drive, like I do, then the share name is the drive name. Concatenating the two fields into one does not solve any problems afaict . OK, it might conform to some standard, but for me it is easier to switch between multiple VC machines that are set up the same way. I have three machines that I'm testing on and all have d/wavwatch shared. This way I can just change the name of the host. In the long run it will probably change quite a bit. I'd like to be ale to store settings for multiple hosts and be able to switch between them with a couple clicks, or even a voice command. ;D
That's cool then :)
Edit: ..but just a thought ;D : When you share a folder <Some_Drive>:\X\Y with share-name <Share-Y> , it is understood that any subdirectories under Y are also shared, i.e., one can go ahead and access the absolute path <Some_Drive>:\X\Y\Z with <Share-Y>\Z from a remote machine, but, it is neither apparent nor necessary that the parent folder(s) of Y, namely <Some_Drive>:\ or <Some_Drive>:\X\ are shared. This "standard" assumption was what caused the troubles for me, Kalle and Dave. And if you go by this logic, you can still accept the hostname separately in settings. The only modification you would then want to do is to mandate that the "root" of whatever forward-slashed path you're accepting in Save path on host be shared. i.e., for your example, ask people to share D:\ if they want to use D:\wavwatch as the wav-file folder. So no code change required actually.
Hope that made sense! Sorry to be persistent with this. I'll now get off your lawn :)
Quote from: Kalle on July 26, 2012, 03:24:38 PM
Hi James,
I've tested it with my Galaxy S2 with Android 4.03
1) for network path I used the same methode on Android as ajith (wavewatcher for the path: c:\wavwatcher) all other things do not work.
2) recording and send with the button work only fine with my headset. The onboard mic in the Galaxy is waste :'(
3) "tilt to speak" doesn't work on the Galaxy S2 - VoxWav send a big file to the wavwatch folder >2 MB without content and VC show this in the upper field "can't open wav file: c:\wavwatch\android8_19_55.wav" and create in the historie window a event "Wav.FailedRead".
point 2 and 3: after the new "VoxWav update" everthing works fine. The tilt crash is fixed :)
For my Galaxy S2 it was important to clearing the cache after the update (with a android task mananger)
The speech recognition rate is usually above 90 percent.
Very good job :clap
Quote from: ajith on July 27, 2012, 12:20:50 AM
That's cool then :)
Edit: ..but just a thought ;D : When you share a folder <Some_Drive>:\X\Y with share-name <Share-Y> , it is understood that any subdirectories under Y are also shared, i.e., one can go ahead and access the absolute path <Some_Drive>:\X\Y\Z with <Share-Y>\Z from a remote machine, but, it is neither apparent nor necessary that the parent folder(s) of Y, namely <Some_Drive>:\ or <Some_Drive>:\X\ are shared. This "standard" assumption was what caused the troubles for me, Kalle and Dave. And if you go by this logic, you can still accept the hostname separately in settings. The only modification you would then want to do is to mandate that the "root" of whatever forward-slashed path you're accepting in Save path on host be shared. i.e., for your example, ask people to share D:\ if they want to use D:\wavwatch as the wav-file folder. So no code change required actually.
Hope that made sense! Sorry to be persistent with this. I'll now get off your lawn :)
Yeah, my original instructions were not stellar, that's for sure. Eventually, if we don't switch to some kind or audio stream broadcast, then I hope to at least offer some kind of network explorer to make choosing the path easier.
And btw I appreciate your suggestions. If I was a bit quick to dismiss your suggestion, it's probably because I was a bit burned out from banging my head against the wall of code all day yesterday. ::club
get the code, report bugs, request features etc. here: http://code.google.com/p/voxwav/
Quote from: jitterjames on July 27, 2012, 09:18:36 AM
get the code, report bugs, request features etc. here: http://code.google.com/p/voxwav/
Awesome! Great to see that you are going to release the source code of the app, thanks James!
What about using Git (GitHub) instead of SVN?
Looking forward to test the app in my VC installation, although my setup using PocketAudio is working nicely ;-) I'm sure that VoxWav (I prefer VoxCommandroid :D) will be my default app once I try it, everything that comes from you is wonderful!
Thanks for all you work and best regards,
Aitor
Quote(I prefer VoxCommandroid ;) )
I prefer VoxMic ;)
Hey James,
Version 1.1.b works really well, no problems with the gyro anymore! :clap
And i don't know why, but my results are getting better :)
only sometimes it takes a lot of time to submit the command, but i guess it has nothing to do with your app, but with my wireless lan.
p.s. i had to uninstall the old version of VoxMic, to install the newer one, otherwise i got an error.
Quote from: Dave on July 28, 2012, 05:26:17 AM
Hey James,
Version 1.1.b works really well, no problems with the gyro anymore! :clap
And i don't know why, but my results are getting better :)
only sometimes it takes a lot of time to submit the command, but i guess it has nothing to do with your app, but with my wireless lan.
p.s. i had to uninstall the old version of VoxMic, to install the newer one, otherwise i got an error.
Are you using VC version 0.957? What kind of delays are we talking about?
It's important to use 0.957 to avoid a conflict between the android trying to copy and VC trying to open the file at the same time.
Quote from: Dave on July 28, 2012, 05:26:17 AM
p.s. i had to uninstall the old version of VoxMic, to install the newer one, otherwise i got an error.
I will try to figure out how to use the play store at some point which will hopefully make updates smoother.
QuoteAre you using VC version 0.957? What kind of delays are we talking about?
It's important to use 0.957 to avoid a conflict between the android trying to copy and VC trying to open the file at the same time.
Version 0.957 ???
Where can I found it?
On the Google code site. I put it there for convenience since the only change was to work with voxwav.
Quote from: jitterjames on July 28, 2012, 08:38:44 AM
Are you using VC version 0.957? What kind of delays are we talking about?
It's important to use 0.957 to avoid a conflict between the android trying to copy and VC trying to open the file at the same time.
Ah, sorry, didn't notice that there is a new version. Still using 0.956.
Sometimes the apps says "saving" on the recording button for a long time (maybe 15 or 20 seconds), ven with short commands...
I will update VC tomorrow when i'm back home and try again...
Hey James, I just wanted to update VC but I cannot find 0.957 anymore...
VoxWav 1.2 works like a charm, even the delays seem to be gone. :yay
When I started VoxWav 1.2 for the first time, I had a bug: Everytime VoxCommando recognized a command, VoxWav displayed "not recognized".
After restarting VoxWav it displays nothing, wether the command was recognized or not.
Quote from: Dave on July 29, 2012, 03:23:10 PM
Hey James, I just wanted to update VC but I cannot find 0.957 anymore...
VoxWav 1.2 works like a charm, even the delays seem to be gone. :yay
When I started VoxWav 1.2 for the first time, I had a bug: Everytime VoxCommando recognized a command, VoxWav displayed "not recognized".
After restarting VoxWav it displays nothing, wether the command was recognized or not.
Hi Dave, waiting for 0.958 James work on it ;)
https://www.dropbox.com/s/50xoinwb038cwql/VC%20Setup%200958.exe
I just uploaded VW 1.2b
I think it works pretty well with 0958. I've only tested quickly though.
I hope it works well for the rest of you.
Quote from: jitterjames on July 29, 2012, 03:41:52 PM
I just uploaded VW 1.2b
I think it works pretty well with 0958. I've only tested quickly though.
I hope it works well for the rest of you.
I can't make it work here, using VC 0.9.5.8 and the VoxWav.apk file that you have submitted to google code some minutes ago :-(
I think that my problem is that VoxCommando is trying to read the wav file so fast that the file transfer is not even completed. What does the android app do, wait until the recording is completed and then transmit the file or start sending the file as soon as you tilt the phone? Here is the event history that I can see in VC:
Event&&AV.Resuming
AV.Resuming
Event&AV.Mic.On
AV.Mic.On
Event&&AV.SmbTransfer.Begin
AV.SmbTransfer.Begin
Event&&AV.Mic.Off
AV.Mic.Off
Wav.FailedRead
Event&&AV.SmbTransfer.End
AV.SmbTransfer.End
Is VoxCommando trying to open the wav file before it is transferred? In my shared "wavwatch" folder I can see many android*_*_**.wav files, and if I move them to any other folder and move them again to the shared folder then VC loads them properly and the command is triggered properly.
Am I the only one with this problem?
Many thanks for your work James, the app looks great for a beta version!
QuoteAm I the only one with this problem?
No you not the only one, I had the problem with "VoxWav 1.2".
If you have a program on your android device that can clearing the cache you can do this or restart your android device and test VoxWav again.
Kalle
Quote from: bodiroga on July 29, 2012, 04:05:49 PM
I can't make it work here, using VC 0.9.5.8 and the VoxWav.apk file that you have submitted to google code some minutes ago :-(
I think that my problem is that VoxCommando is trying to read the wav file so fast that the file transfer is not even completed. What does the android app do, wait until the recording is completed and then transmit the file or start sending the file as soon as you tilt the phone? Here is the event history that I can see in VC:
that was happening in previous versions, and in some cases there was a conflict where VC would start trying to read the file before the transfer was complete.
Here is what actually happens if you have VW 1.2b and VC 0.958
VW saves the WAV locally (into your sdcard folder) while you are recording. It then initiates a transfer to the samba share, saving the file as filname.tmp. VC should be ignoring this file because it is not interested in any files that do not end with .wav. When the transfer is complete, VW will rename the file to have a .wav extension. Now VC should see the file, open it, and perform it's analysis. I have it working perfectly here with two different androids (2.2 and 4.0.3) and several computers.
If you guys really have the latest version, I don't know what could be going wrong, but maybe you could try to uninstall VW completely and then install the latest version that is currently on the google code site.
Then you need to reenter the samba info to be able to connect. If you watch the folder when VC is off, and you send a nice long file you should be able to see the file appearing as a .tmp file and then changing to .wav
I'll upload a new version of VC later tonight, or first thing in the morning. Just to be safe, I'll get VC to wait a short period before trying to open the wav incase it is interfering with the file rename.
We have guests right now so I can't do it at the moment.
I've been struggling to get VC to recognize my commands through my $70 Android phone using 2.2 Froyo, while James was having much better luck, though still not perfect; it was pretty frustrating! This was made worse because my phone will pick up any ambient sound in the room -- it seemed way too sensitive. Music playing in the background was not an issue, since the auto-mute function was working, but if anyone started talking in the background while I was trying to issue a command, VC would not understand my command.
Today we decided to create a new untrained profile to compare. *Instantly* I started getting 98% - 99% recognition from Vox, and James is getting the same accuracy results using his tablet on the same untrained profile. I've had no problems since then, unless someone is talking in the background. That seems to be the nature of my phone. I don't know that this solution will work for non-native English speakers giving commands in English, but for me, changing to a new profile not "tuned" to James made all the difference, and the profile still works for James as well.
OK, I'm really not sure why it's not working for some of you, but hopefully this version will help some of you. Or at least the history window might provide a better clue of what it going on.
https://www.dropbox.com/s/1i6zcteg1z4yjzh/VC%20Setup%200959.exe
Quote from: bodiroga on July 29, 2012, 04:05:49 PM
I can't make it work here, using VC 0.9.5.8 and the VoxWav.apk file that you have submitted to google code some minutes ago :-(
I think that my problem is that VoxCommando is trying to read the wav file so fast that the file transfer is not even completed. What does the android app do, wait until the recording is completed and then transmit the file or start sending the file as soon as you tilt the phone? Here is the event history that I can see in VC:
Event&&AV.Resuming
AV.Resuming
Event&AV.Mic.On
AV.Mic.On
Event&&AV.SmbTransfer.Begin
AV.SmbTransfer.Begin
Event&&AV.Mic.Off
AV.Mic.Off
Wav.FailedRead
Event&&AV.SmbTransfer.End
AV.SmbTransfer.End
Is VoxCommando trying to open the wav file before it is transferred? In my shared "wavwatch" folder I can see many android*_*_**.wav files, and if I move them to any other folder and move them again to the shared folder then VC loads them properly and the command is triggered properly.
Am I the only one with this problem?
Many thanks for your work James, the app looks great for a beta version!
I had the same problem, but this is not because voxcommando is traying to read the file very fast.
The problem is if you have a mic enable in the pc VoxCommando cant read the .wav files, so...disable the mic in pc and everything will work well (version 0.958).
I test for now with two devices :
1 - Motorolla Atrix with android 3.2, works but its very sensitive, sometimes i have to put the cellphone 50cm way, its not a good device for this app.
2- Galaxy Tab S G70, works perfectly. I get 94% or more of accuracy. Very well job James.
Sometimes the app crash, happens when the VoxCommando dont recognize the command and the app try to get the feedback udp, if voxcommando dont send the feedback the app crash (at least i think so.)
Congratulaion James and thanks.
Clayton
Quote from: claymic on July 30, 2012, 08:08:23 AM
Sometimes the app crash, happens when the VoxCommando dont recognize the command and the app try to get the feedback udp, if voxcommando dont send the feedback the app crash (at least i think so.)
My bad, i was using the version 1.2, now i install the version 1.2b and the crashing stop. Very thanks for your support (email) James.
I am very happy with this app. Works very nice.
Clayton
Ok, with the last version of VC 0.9.5.9 VoxWav is working fine in my Samsung Galaxy S2 (CM9), but now my setup using VAC with a USB mic and PocketAudio (I repeat both sounds into Line 1 and I have configured Line 1 as my standard device in windows recording settings -similar setup to the one that Kalle described here: http://voxcommando.com/forum/index.php?topic=825.msg6877#msg6877-) is no working :( :( WTF! If I use the default record program that comes with windows everything works fine, Line 1 is recognized as the record device and I can here my voice in the wma file, but VC doesn't hear anything in Line 1 (the bottom green line doesn't move). Yesterday with 0.9.5.8 the setup was working fine, but after upgrading to 0.9.5.9 I can't make it work (and installing a different version of VC, 0.9.5.6 for example, doesn't help).
James (or Kalle), do you know what's happening here???
Many thanks for your help!
Sorry, if going back to a previous version doesn't work, then I have no idea, but it must only be a coincidence that it happened when you installed 0.959
James
Only for undertand how this work.
After the .wav files recorded and salved in the shared folder, how VoxCommando read it ? There is a way to manage the volume or booster of this recorded audios ?
I test in two pcs. In the first pc my Galaxy G70 works perfectly, i have to put the galaxy very close to my mouth and there is no interference of the surround sound.
But in the second pc was diferent, i have to put the G70 30cm from my mouth and now i get a lot of surround sound interference. I dont understand why.
I try to cover the mic but dont work very well.
Thanks James
Clayton
Quote from: jitterjames on July 30, 2012, 10:19:02 AM
Sorry, if going back to a previous version doesn't work, then I have no idea, but it must only be a coincidence that it happened when you installed 0.959
James, I don't know if this says something to you, but if I enable the option "Check the mic state every x seconds" I can see lots of "time since mic event...", "trying to reset" and "mic reset" events in the history tab, so it's clear that VC can't see my "Line 1" mic (remember that the mic is working fine in Windows, tested with the build in recording program").
Ohhh, fuck, I have opened the log file and this is the error I'm getting:
30/07/2012 16:03:50 237 focused: VoxCommando
30/07/2012 16:03:51 742 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
30/07/2012 16:04:01 755 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
30/07/2012 16:04:10 837 focused: Idle
30/07/2012 16:04:26 802 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
30/07/2012 16:04:51 835 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
30/07/2012 16:04:52 985 focused: VoxCommando
30/07/2012 16:04:54 55 error opening options.xml
30/07/2012 16:05:08 40 focused: Idle
30/07/2012 16:05:16 875 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
30/07/2012 16:05:26 882 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
30/07/2012 16:05:35 140 focused: VoxCommando
30/07/2012 16:05:36 903 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
30/07/2012 16:05:56 210 focused: explorer
30/07/2012 16:05:56 934 unable to reset speech engine: System.InvalidOperationException: No se puede realizar esta operación mientras el reconocedor está realizando el reconocimiento.
en System.Speech.Recognition.RecognizerBase.SetInputToDefaultAudioDevice()
en System.Speech.Recognition.SpeechRecognitionEngine.SetInputToDefaultAudioDevice()
en VoxCommando.MainWindow.setModeReset()
Does this say anything to you?
Many thanks for your help as always James, you know how much I admire your work ;)
Quote from: bodiroga on July 30, 2012, 11:11:47 AM
James, I don't know if this says something to you, but if I enable the option "Check the mic state every x seconds" I can see lots of "time since mic event...", "trying to reset" and "mic reset" events in the history tab, so it's clear that VC can't see my "Line 1" mic (remember that the mic is working fine in Windows, tested with the build in recording program").
don't enable "Check the mic state every x seconds". Set it to 0
Also, reboot.
Quote from: claymic on July 30, 2012, 10:29:07 AM
James
Only for undertand how this work.
After the .wav files recorded and salved in the shared folder, how VoxCommando read it ? There is a way to manage the volume or booster of this recorded audios ?
I test in two pcs. In the first pc my Galaxy G70 works perfectly, i have to put the galaxy very close to my mouth and there is no interference of the surround sound.
But in the second pc was diferent, i have to put the G70 30cm from my mouth and now i get a lot of surround sound interference. I dont understand why.
I try to cover the mic but dont work very well.
Thanks James
Clayton
I don't understand the question: "how VoxCommando read it"
It opens it like any other file. What can I say about it?
If you are using the same wav file from the same android phone, and getting better results with one PC than the other, then the difference must be in the way you have trained your speech profile. Perhaps one machine learned using a louder microphone, so now it is better able to understand when the wav is loud.
Thanks James
Hello James,
With VoxMic 1.2 and VC 959, I get the following in the VC history window:
AV.Mic.On
Event&&AV.SmbTransfer.Begin
AV.SmbTransfer.Begin
Event&&AV.Mic.Off
AV.Mic.Off
Wav.Renamed
Wav.FailedRead.OnRename
Event&&AV.SmbTransfer.End
AV.SmbTransfer.End
Event&&AV.Destroyed (closed)
The recorded audio gets transferred just fine, with the voice properly recorded. However, VC doesn't seem to try to recognize it. Again, a case of trying to read prematurely?
Edit: Same thing as above with VC 958 (and VoxMic 1.2)
Edit2: The download page did say it was VoxWav "1.2b" however, I see that the version is "1.2" in the "Manage Apps" screen on my mobile phone. Is it supposed to show "1.2b" ?
Nope. VC does not read prematurely anymore. This was solved by using a .tmp file and then renaming it to .wav However, as some users have noticed, if you have an active "normal" microphone listening at the time that the wav is renamed, there is a conflict when VC tries to open the wav even though it is processing sound already (from the microphone).
This will be resolved in the next release, but to test it you can simply put VC into off mode to see if it fixes the problem. Or you could even use the events generated by VoxWav to turn VC off when a .wav is on its way, and then back on when the wav processing is complete.
I've already fixed this, but I have a couple other things I want to do before I upload 0.960.
Also, note there is currently a bit of confusion because I put out a second version of VoxWav 1.2 that I called "b" but there is no way to know for sure if you are using 1.2 or 1.2b. If you are not sure you could reinstall from the apk currently on google code, since it is the latest. I will release VoxWav 1.3 soon so there will be less confusion, but again I want to make a few changes first.
Yes, it works after disabling VC. Waiting for VoxMic 1.3 :)
Oh, just noticed its now called VoxWav instead of VoxMic. So that's final then, the name that is? :)
nothing is final :)
and for the record, the solution to this particular problem you are having will be in VC 0.96, not in VW 1.3
Quote from: ajith on July 30, 2012, 01:28:40 PM
Yes, it works after disabling VC. Waiting for VoxMic 1.3 :)
Oh, just noticed its now called VoxWav instead of VoxMic. So that's final then, the name that is? :)
Yes the same for me - everything works perfect if I used in VC-OFF mode.
Very good results with 98-100 percent can be achieved if you use a new blank Windows Speech Profile. (Thanks to Nime5ter to figure this out :hugs by special'K ;))
For me, VoxMic its a good name ;D
Quote from: Kalle on July 30, 2012, 02:13:03 PM
Yes for me the same - everything works perfect if I used in VC-OFF mode.
Very good results with 98-100 percent can be achieved if you use a new blank Windows Speech Profile. (Thanks to Nime5ter to figure this out :hugs by special'K ;))
Kalle
I was thinking to create a new profile, the problem is...How i can make the tutorial for speech reconigtion using my android? Almost all my commands are in portuguese.
Quote from: claymic on July 30, 2012, 02:19:34 PM
Kalle
I was thinking to create a new profile, the problem is...How i can make the tutorial for speech reconigtion using my android? Almost all my commands are in portuguese.
I do not understand the question - I've open windows speech settings -> create a new profile -> cancel the window to ask you what mic you use -> apply
Thats all what I do - I use it with english and german language and it works.
Quote from: claymic on July 30, 2012, 02:19:34 PM
I was thinking to create a new profile, the problem is...How i can make the tutorial for speech reconigtion using my android? Almost all my commands are in portuguese.
Quote from: jitterjames on July 26, 2012, 04:40:08 PM
** actually you CAN train with wav files. All you need to do is set the default mic to "what you hear" and then play the wav files with windows media player (or whatever). In order to get the wav files from your android, the easiest way is to use VoxWav when VoxCommando is not running, and then just go look in the wavwatch folder.
Kalle, my english its bad, i am sorry for that. But Dave understood.
Very thanks to both. I will create a new profile an make a new train.
Quote from: claymic on July 30, 2012, 02:44:09 PM
Kalle, my english its bad, i am sorry for that. But Dave understood.
Very thanks to both. I will create a new profile an make a new train.
Haha, not your english is bad, my is bad - sorry :bonk
By the way: you can use the attach group to set VC in off-mode after recognized a command from "VoxMic". So you can use VoxMic without problems in VC959.
[attachment=1]
Very thanks Kalle, i will get this right now.
I have uploaded 0.960. It should allow you to use android at the same time that you have a microphone with VC in "on" or "standby" modes now. You may still choose to use the events to turn VC to "off" mode though, if that works for you.
http://voxcommando.com/forum/index.php?topic=868
there is also a little something new in this version that might turn out to be quite handy for some. or maybe not... :biglaugh
I test right now and works with all my android devices. Very thanks James
First, work with my external microphone with any problem. i dont have to put vox off anymore.
Tested with :
Motorla atrix : fade .4 reverse fade .6 reverse vol 0.8
Galaxy S G270 vol 0.15
Galaxy Mini fade 0.5 vol 0.15
Tablet Genesis GT 7200 fade .4 reverse fade .6 reverse vol 0.5 (the worst perfomance here, but works)
And James, i was worry about two things : The window of SOX (you already solved) and the delay of the command using this method, but its very fast, only 15ms, at least in my htpc usin SSD.
Congratulations again, works very nice with all my devices and pcs, i can choose use the Android or the Microphone.
Clayton
Yes, 960b has fixed the simultaneous mic/VC ON issue.
A couple of things I have observed -
- The VoxWav "Tilt On/Off" button does not seem to have any effect on the AV.Mic.ON/OFF events being fired off by VoxWav. That is, AV.Mic.ON/OFF events are always fired when the phone is tilted 90 deg and back, irrespective of whether the VoxWav Tilt button is ON or OFF.
- Whenever AV.Mic.ON/OFF events are fired, they come in fours. i.e., each time it is a burst of 4 AV.Mic.ON events or 4 AV.Mic.OFF events.
Quote from: ajith on July 31, 2012, 11:35:05 AM
Yes, 960b has fixed the simultaneous mic/VC ON issue.
A couple of things I have observed -
- The VoxWav "Tilt On/Off" button does not seem to have any effect on the AV.Mic.ON/OFF events being fired off by VoxWav. That is, AV.Mic.ON/OFF events are always fired when the phone is tilted 90 deg and back, irrespective of whether the VoxWav Tilt button is ON or OFF.
- Whenever AV.Mic.ON/OFF events are fired, they come in fours. i.e., each time it is a burst of 4 AV.Mic.ON events or 4 AV.Mic.OFF events.
Wow. Interesting. :o That's the beauty of beta testing. All the androids seem to have their own way of doing things. I'm curious to see if I can fix this in the next release. You'll have to tell me since I can't recreate it here. Thanks for the heads up. :-)
New version (1.3)...
new thread: http://voxcommando.com/forum/index.php?topic=872
;D