09-20-2009 12:31 AM
I am building the same app, under Vista and a 64bit processor. On the code for the callback, after using a unbundle to get SpeechRecognizedEventArgs e, I cant access its properties because it throws a Property Node: Invalid Refnum class
Any idea why this might be?
(using init2.vi provided here)
09-21-2009 05:20 PM
05-22-2010 08:30 AM
I need help in my final project. I am trying to make a security system for a house based on voice recognition and also by entering a code or password. I need help in doing both. I don`t know from where to start. I just know the basic functions of labview.
I will appreciate your ideas and comments.
05-22-2010 03:56 PM
Well, what I want to exactly do is to simulate an alarm system for a house, in which you type a password and you say a word code... My idea is to first save a specific voice in a data base in labview, which i don't know how to do it, so that when I want to come into the house I type the correct password and the system recognizes my voice so it desactivates the alarm, and whenever I want to activate the alarm I do the same thing. The thing is that i will have alarms in different parts of the house, for example the main door, the kitchen, the living room..etc And what i think would be nice is that if i could say "activate living room" it recognizes the voice that is permitted plus it activates that alarm.
05-24-2010 10:28 AM
Ok, I understand what you're looking for, but unfortunately the microsoft functionality (from what I know) can not distinguish unique voice patterns which limits it's ability to do what you are looking for.
I have heard of people doing voice processing using LV but have not seen an actual implementation.
Good luck in your searching.
PS, if the need to detect a specific person goes away, stop back here w/ questions regarding getting going w/ what MS can do.
05-24-2010 10:50 PM
Ok, Thank You! As I was researching I found this document, it's in spanish, i don't know if you might understand it but it kind of like shows how to identify a specific voice. They show the block diagram and everything maybe you can help me try to understand what were they exactly trying to do.
05-25-2010 05:42 PM
It's difficult to see the block diagrams in much detail but it general it looks like they train the VI by taking a signal, normalizing it, and performing a correlation on it. This value is then stored as the 'password'. When they record the user again, they perform the same sequence on it and compare the results. If the correlation values are similar, they pass it.
It looks like they might be doing a few iterations of the processing as it looks more complicated than the simple steps listed above but I think that's the general idea of what's going on.