Thanks DJ, we integrated AIMEC:3 with the online 20Q.net in realtime and it was a brilliant demo for our robot, as It nearly always gets what onlookers had thought of. The 20Q Ai was invented by the brilliant Canadian Robin Burgener who I have had good fortune to meet and talk to a number of times as he is a fellow toy inventor and we both licensed major toy products to Radica/Mattel.
Try 20Q its a lot of fun and the Ai mostly guesses correctly even when you give some wrong answers!
Wow, that's amazing. I only time to play 3 games and I only beat it once but it did guess in 22 questions. It guessed pengunes, a desk and the Borg (you know, the Swedish guys in Star Trek).
How could I put that into my robot? It would be great to play this with the voice responce like in your video.
Troy, the secret for high accuracy speech recognition is the microphone quality and in the early days training. With this early robot I used a professional wireless micro wearable mic which usually clipped to my shirt pocket or collar. The SNR (signal to noise ratio) of this mic is amazing and as it is closer to the speech source it tends to filter out external sounds even load music. These mics are made for this as they designed for public speaking and stage presentation.
This AIMEC in 2006 used the SAPI:5.1 speech recognition engine that was part of XP, this older engine only worked with grammar files and had no dictation mode like the current engine 6 has. With the robot we had grammar files that were for hundreds of phrases that it worked with and it was extremely accurate and very rarely got a false recognition, the worst thing that could happen was it missed the phrase and you have to repeat it. In Win 7 the engine 6 is more advanced with a dictation mode were you do not need to have grammar files so you can say anything and it will try to determine what the phrase or word is. The dictation mode is extremely difficult to do and you do get lots of errors but training really helps. Win 7 etc can still use the grammar mode and much more accurate recognition is possible. We now use Dragon (DNS11) with our current robots as we find its dictation mode is really good and its works right out of the box with limited training required.
Ha! Tony that's awesome. Reminds me of ANIMALS that was found on the Apple ][ MS DOS Master Disk!
Thanks DJ, we integrated AIMEC:3 with the online 20Q.net in realtime and it was a brilliant demo for our robot, as It nearly always gets what onlookers had thought of. The 20Q Ai was invented by the brilliant Canadian Robin Burgener who I have had good fortune to meet and talk to a number of times as he is a fellow toy inventor and we both licensed major toy products to Radica/Mattel.
Try 20Q its a lot of fun and the Ai mostly guesses correctly even when you give some wrong answers!
Here it is http://www.20q.net/
I've just realised, I need to add a lot more functions to Jarvis...
That's pretty darn awesome
Wow, that's amazing. I only time to play 3 games and I only beat it once but it did guess in 22 questions. It guessed pengunes, a desk and the Borg (you know, the Swedish guys in Star Trek)
.
How could I put that into my robot? It would be great to play this with the voice responce like in your video.
I wonder. if you could put this guessing game in ARC, could you put a game like hangman or something like that too?
I gotta remember to utilize this , very cool , it can keep kids and adults alike busy for a few minutes trying to outwit him.
This is incredible! Flawless recognition and immediate responses. What am I missing?
Troy, the secret for high accuracy speech recognition is the microphone quality and in the early days training. With this early robot I used a professional wireless micro wearable mic which usually clipped to my shirt pocket or collar. The SNR (signal to noise ratio) of this mic is amazing and as it is closer to the speech source it tends to filter out external sounds even load music. These mics are made for this as they designed for public speaking and stage presentation.
This AIMEC in 2006 used the SAPI:5.1 speech recognition engine that was part of XP, this older engine only worked with grammar files and had no dictation mode like the current engine 6 has. With the robot we had grammar files that were for hundreds of phrases that it worked with and it was extremely accurate and very rarely got a false recognition, the worst thing that could happen was it missed the phrase and you have to repeat it. In Win 7 the engine 6 is more advanced with a dictation mode were you do not need to have grammar files so you can say anything and it will try to determine what the phrase or word is. The dictation mode is extremely difficult to do and you do get lots of errors but training really helps. Win 7 etc can still use the grammar mode and much more accurate recognition is possible. We now use Dragon (DNS11) with our current robots as we find its dictation mode is really good and its works right out of the box with limited training required.