Help Wanted - MSc Project

[Gerk]

A (quite geeky) idea from me would be something that analyses and corrects phase in recorded audio, i.e. the ability to determine the direction a mic was as compared to the sound source when it was captured, and possibly a way to correct this

[CosmicDolphin]

I'd like a program that analyses great melody & chord structures and comes up with new ones using AI.

 

CD

then we'd be into defining a criteria for great ... maybe based on No.1 hits, record sales, actually there could be something in that... analysis of all the past no.1s for specific genres using fingerprint type algorithms with a goal to artificially creating new songs that people instantly like.. hmmmm 

Sounds good , then you could sell it to us here 

CD

[McLovin]

I'd like a program that analyses great melody & chord structures and comes up with new ones using AI.

 

CD

then we'd be into defining a criteria for great ... maybe based on No.1 hits, record sales, actually there could be something in that... analysis of all the past no.1s for specific genres using fingerprint type algorithms with a goal to artificially creating new songs that people instantly like.. hmmmm 

Sounds good , then you could sell it to us here 

CD

Woooe down,... we get the membership discount!  Free!!

[napoleonboot]

Ideas:
How about something about social interactions / behaviour in an online community for music creation, compared to the tradution prcess. Could branch out into team roles (e.g. Belbin), motivation (Maslow) etc. Maybe a bit too social science and not scientific enough.

Alternatively how about emerging business models for monetising music, such as peple giving albums away free, and getting most income from gigs. Maybe a bit too business and not scientific enough.

I did an MSc a bout a hundred years ago (actually 1988) at Ford Motor Company, on boring stuff to do with computer systems for car design. And then an MBA about 10 years ago, again on boring stuff to do with decision making. I envy you doing something interesting. I'll give it more thought.

George

[snowdoguk]

Ideas:
How about something about social interactions / behaviour in an online community for music creation, compared to the tradution prcess. Could branch out into team roles (e.g. Belbin), mtivaiotn (Maslow) etc. Maybe a bit too social science and not scientific enough.

Alternatively how about emerging business models for monetising music, such as peple givign albums away free, and getting most income from gigs. Maybe a bit too business and not scientific enough.

I did an MSc a bout a hundred years ago (actually 1988) at Ford Motor Company, on boring stuff to do with computer systems for car design. And then an MBA about 10 years ago, again on boring stuff to do with decision making. I envy you doing something interesting. I'll give it more thought.

George

Some interesting ideas in there George. I had a long chat with the head of dept. at Glamorgan Uni in Cardiff this afternoon. It seems they focus more on the creative side of things rather than the technical. So something along the lines of the social interaction of an online collab community interspersed with some actual collabs with own mixing and mastering and video might do the trick.

Appreciate the input ... thanks very much
Cheers
Mike

[detune]

I'll mention something I've looked into doing myself and that there doesn't seem to be a good implentation of: convincingly changing the gender of a vocal track.  A passable gender change can be accomplished by pitch-shifting up a few semitones, and adjusting the formant with a plugin.  But it's not 100% convincing, and you end up with something that doesn't sound quite human. 

Imagine the possibilities if this could be done right.

Oops, just read your latest note snowdog--maybe this is not the best suggestion.

Mark

[snowdoguk]

I'll mention something I've looked into doing myself and that there doesn't seem to be a good implentation of: convincingly changing the gender of a vocal track.  A passable gender change can be accomplished by pitch-shifting up a few semitones, and adjusting the formant with a plugin.  But it's not 100% convincing, and you end up with something that doesn't sound quite human. 

Imagine the possibilities if this could be done right.

Oops, just read your latest note snowdog--maybe this is not the best suggestion.

Mark

Hey Mark

Great suggestion, any ideas are welcome since they are all things I could discuss as possibilities for a project with the Uni...  cool idea, many thanks

[snowdoguk]

A (quite geeky) idea from me would be something that analyses and corrects phase in recorded audio, i.e. the ability to determine the direction a mic was as compared to the sound source when it was captured, and possibly a way to correct this

I haven't answered this post yet Gerk (until now) because I was struggling to understand the actual problem, which probably highlights my lack of knowledge in anything mic related 

[napoleonboot]

another idea

synthesised instuments are taking over from real instruments in some ways, and can be very convincing when done well. Billy is the best I've ever heard at this, he once did a sax solo in a song and nobody would ever be able to tell it wasnt real.

But.. interestingly, nobody ever does synthesised guitar solos. Why is this - what charavterises guitar that makes it much more difficult to synthesise than (say) drums. Vocals is even harder. Something in there perhaps?


George

[Gerk]

A (quite geeky) idea from me would be something that analyses and corrects phase in recorded audio, i.e. the ability to determine the direction a mic was as compared to the sound source when it was captured, and possibly a way to correct this

I haven't answered this post yet Gerk (until now) because I was struggling to understand the actual problem, which probably highlights my lack of knowledge in anything mic related 

Hehehe understood.  In that case my suggestion is probably not something you would want to jump into 2 feet first LOL

[CosmicDolphin]

But.. interestingly, nobody ever does synthesised guitar solos. Why is this - what charavterises guitar that makes it much more difficult to synthesise than (say) drums. Vocals is even harder. Something in there perhaps?


George

It's beacuse they sound rubbish....to play like a guitar you need a guitar like controller otherwise it just sounds fake. There are guitars that generate modelled sounds of different guitars or other string instruments that sound totally realistic , but you still need to be a guitarist to make it work.

CD

[snowdoguk]

I'll mention something I've looked into doing myself and that there doesn't seem to be a good implentation of: convincingly changing the gender of a vocal track.  A passable gender change can be accomplished by pitch-shifting up a few semitones, and adjusting the formant with a plugin.  But it's not 100% convincing, and you end up with something that doesn't sound quite human. 

Imagine the possibilities if this could be done right.

Oops, just read your latest note snowdog--maybe this is not the best suggestion.

Mark

I changed tack after all Mark and am now nearing the end of the first semester of a very technical based MSc in Digital Music Processing, so your idea is actually quite appealing at the moment 

[snowdoguk]

A (quite geeky) idea from me would be something that analyses and corrects phase in recorded audio, i.e. the ability to determine the direction a mic was as compared to the sound source when it was captured, and possibly a way to correct this

After 6 months of hard study I'm a bit better placed to consider this now but still don't fully understand what you mean. I can appreciate that due to the room response there would be a component of the direct sound, and then some early reflections followed by the late ones.... but I don't see how it could possibly be determined in which direction the mic is pointing unless you have a very detailed model of the room shape/size and contents and know the exact location of the source, which would render it for anything practical pretty useless...     phase difference and correction between two mics monitoring the same source is probably already covered and relatively trivial... any more thoughts on this Gerk?
cheers
Mike

[Gerk]

I see what you're saying that without a comparison you can't easily tell.

How about another suggestion, which has probably already been done, but might be fun to try ... How about something to emulate the way mid/side works using only a single sound source?  For mid side explanation it's better to look it up than have me explain it if you're not familiar with that technique.

[snowdoguk]

I see what you're saying that without a comparison you can't easily tell.

How about another suggestion, which has probably already been done, but might be fun to try ... How about something to emulate the way mid/side works using only a single sound source?  For mid side explanation it's better to look it up than have me explain it if you're not familiar with that technique.

will do 

all suggestions are welcome... to do with any aspect of digital music processing, including feature extraction, which I'm actually very interested in, it's a hot topic at the moment due to all the DRM and copyright stuff going on.

thanks
Mike