Robot playing Guitar Hero

[shmokes]

http://www.youtube.com/watch?v=icsDptr7Umw&eurl=http://gizmodo.com/index.php?refId=383209

[ahofle]

LOL even the robot can't get 100%.

[Cakemeister]

Yeah, I saw that on [H]. The bot has problems with sustains and it doesn't do any hammer-on/pull-offs or alt-strums. So any note that has to be repeated faster than the recovery time of the strumming solenoid can't be played.

I am more curious about the input than about the output. How does the bot "see" what notes to play? Do they cheat and enter a note chart?

[shmokes]

I don't think they cheat.  I think they're using a camera, cos you can hear a guy saying not to start the software running until the graphical fret-board appeared on-screen, suggesting that something needed to see it, or the software would get confused.

[AtomSmasher]

According to the description "The system plays Guitar Hero by analyzing the video signal."  which my guess means the video goes through the robot before going to the tv, which should make a lot more accurate then a video camera in front of the screen.

[massive88]

When someone goes to that much work on something, you really think they can't make decent video footage of it?  Or it the footage wasn't great, refilm it...?  15 minutes of camera work versus 15 hours of assembly?

If its their senior EE project, and anything like my senior ME project at UT, it took a hell of a lot more than 15 hours.

Also since its April, I would assume they aren't done yet, or at least will have a more final version wrapped up in a few weeks for a final presentation around finals time.

Though I thought the same thing when they were saying that the background fret board is one that it has problems with, if you are making a video, reset and get it on a good board.  Anyway I imagine this vid was just a casual work in progress video.

[AtomSmasher]

According to the description "The system plays Guitar Hero by analyzing the video signal."  which my guess means the video goes through the robot before going to the tv, which should make a lot more accurate then a video camera in front of the screen.

You only see it briefly, but there's a laptop sitting next to the PC and it looks like it may be running some kind of sequencer program.

If the camera was truly analyzing a video signal, it wouldn't really matter if the software was turned on early, now would it?  If the robot was clattery because of no signal or some kind of video level issue, then all they'd have to do is throw a switch on the power until it was ready to go.  Synching software means they're up to something.

The guy says to wait for the fret board to come up, so my guess was that when the software analyzes the signal, it looks for the fret board since the two sides of the board are the one constant it would use.  They also say "it doesn't like this board" which I'm guessing the board has too much clutter on it, which makes it harder for the software to read.  If the notes were preprogrammed in, why would a different fret board make a difference?  It sounds to me like the notes are not preprogrammed in.  It wouldn't be too difficult to make a program to read the notes from the video, so I don't know why they wouldn't (assuming they had access to programmers).

[massive88]

It wouldn't be too difficult to make a program to read the notes from the video, so I don't know why they wouldn't (assuming they had access to programmers).

Half of all double E's are pretty much pure programmers in school.  Most of my EE friends went into the computer science industry after school.   I would assume aTm is the same way.

[AtomSmasher]

It wouldn't be too difficult to make a program to read the notes from the video, so I don't know why they wouldn't (assuming they had access to programmers).

Half of all double E's are pretty much pure programmers in school.  Most of my EE friends went into the computer science industry after school.   I would assume aTm is the same way.

While the EE's generally do know some programming, from my experience most wouldn't know how to program something as advanced as they needed for this.  I had a number of EE friends in college and we even had a few classes where the CS and EE people worked together on various projects and the programming we did was generally well over the head of the EE majors (just as we generally didn't understand how they made the electronics for the projects).  Since there is a lot of clutter on the screen, my guess the best way to program it to eliminate errors would be with a neural network (I made a facial recognition program using a neural net back in college, and it would of had the same basic principles needed for this program), which isn't too difficult if you know what your doing, but I don't think I knew any EE majors who knew how to program one.

[massive88]

Probably not, but they have a whole semester to design, research, inquire, and learn; which is a big part of the senior design project.  Its not like this project is a few weeks in the making.  Well, again at least mine wasnt, different school and different major, so take that for what its worth, both engineering though   

Anyway I just mean to indicate that they probably did program something to recognize the notes, not on the skills of an EE compared to a CS major.

[AtomSmasher]

Probably not, but they have a whole semester to design, research, inquire, and learn; which is a big part of the senior design project.  Its not like this project is a few weeks in the making.  Well, again at least mine wasnt, different school and different major, so take that for what its worth, both engineering though   

Anyway I just mean to indicate that they probably did program something to recognize the notes, not on the skills of an EE compared to a CS major.

My initial comment was that it wouldn't be difficult to program assuming they had access to programmers, it is possible the programmer they had access to was actually an EE major on their team (although I do believe that to be unlikely).  However, if they had to take the time to learn how to program it first, then it definately would of been a lot easier for them to just preprogram the notes, which is why I added the programmer part to my comment.

[ahofle]

God, he's an aggie too?!
Hook 'em!
 

[massive88]

I'm still a little confused as to why they used relays to drive each solenoid rather than some type of transistor set up.  I understand they want to isolate the voltages, but relays add a lot of delay... obviously not too much.

Drop 'em a comment and ask!  http://slashbot.wordpress.com/

Hook 'em!
 

[Samstag]

When my wife went to aTm 10 years ago first-year EEs were affectionately referred to as "pre-Business" because so many dropped the major to become pointy-haired bosses.

[shmokes]

If the camera was truly analyzing a video signal, it wouldn't really matter if the software was turned on early, now would it? 

Whether the signal is being processed before it hits the TV or by a camera after it hits the TV, the fact that the software can't be started until the fretboard appears suggests that the notes aren't being fed to it.  That's all I was saying.

[FrizzleFried]

One thing I know for sure...

...I'll never get 96% on GHIII expert mode for any of the songs...

I can consistently get 98-99% on easy mode...85-95% in medium mode...after that I suck something fierce.


I still can't get better than 3 starts for Reigning Blood even on Easy...song kicks ---my bottom---.

[Thurman]

Reminded me of the band "Captured by Robots".  Except this guy made all of his robots play real instruments......

[youtube]http://www.youtube.com/watch?v=7B0t_v8NYjI[/youtube]