Command recognition on a Sony VAIO UX280
Submitted by jojo on Sun, 04/27/2008 - 00:55.
Hi,
I would appreciate any advice/suggestions regarding a speech controlled wheelchair I am constructing. Go here http://abm.wso.net/redirect/wheelchair.html for a video and overview (Windows Internet Explorer only- my apologies.)
In the current implementation, my control program is running on a desktop PC (3.5MHz, 2GB RAM) with the wheelchair (as you can see in the video) operating as a wirelessly connected robot. This is so I can debug the program without risking life and limb. Here's how it works:
my control program is a Java program that accepts input in the form of text into a swing text component and generates output control signals to the wireless USB-connected signal generator electronics. I am running Vocola and Dragon NaturallySpeaking preferred version 7.3 under Windows XP professional. I run DNS in command mode and have a Vocola command vocabulary of about 30 different variable argument instructions. Each of the Vocola commands causes a text string to be entered into the Java program. The Java program parses the strings into instructions for the hardware. This works fine. There is a significant delay (of about one second) between my uttering a command and text appearing in the Java program input window. This is acceptable, but it would be nice if I could make it shorter.
Now, in the next version of the wheelchair I will be sitting in the wheelchair and the control program will be running on a Sony VAIO UX280 with a 1.2GB Intel solo processor and 1GB of RAM. I will be experimenting with different methods of speech-recognition in this considerably weaker environment, and I would appreciate any advice about alternatives to try. I am aware of the various programs discussed in this forum and it would be very helpful if I could get some expert advice on what is most likely to succeed. The goal is rapid and accurate recognition of a small set of variable argument commands in a computationally limited, probably noisy, environment.
Best regards,
Joe
I'm no expert but since none
I'm no expert but since none has stepped forward I suggest this may not be the best forum for this inquiry. I can't direct you to a better one, but I hope someone who can notices this post and helps out. FWIW, it sounds fascinating, but I wonder if there are any commercial solutions out there? Maybe a programming or machine-control forum would be more useful. It seems to me that there should be better programming environments than the one you've hacked.
Good luck and post back please if you make any progress.
Bruce
Hi Bruce, Thanks for the
Hi Bruce,
Thanks for the response.
It seems to me that there should be better programming environments than the one you've hacked.
Undoubtedly. The current system is cobbled together from pieces that I already had on hand for other projects (Dragon NaturallySpeaking, Vocola, LabJack, Java programs and libraries). In some final version (if there is such a thing) things should be much more lightweight.
I fear that perhaps, by giving so much detail in my original message, I confuse the issue about what I'm looking for. Essentially, all I need is a program that will convert, quickly and with high accuracy in an acoustically noisy environment, a small set of variable argument commands into text. I'll keep looking.
Regards,
Joe