My sister Lubna has this weird habit of doing a bad karaoke act by turning on voice cancellation in the speaker settings and playing any song she (thinks she) can pull off. She's been doing this for at least a year now. Now she's started this new thing where she cancels the singer's voice, but doesn't sing herself(Thank God...) - instead she tries to pick out the different instruments being used.
She was doing this a few minutes ago, and asked me for suggestions. Now I have rather weird musical tastes, and it so happens that 3 of my favourite songs are in Japanese. And no, I don't understand them, even though I remember them word for word...
Anyway, there's a particular favourite of mine that goes by the name Zutto Kimi No Soba De. I've always admired the music of this song - it's pretty complex, in the sense that if you listen closely, you can hear 3 different tunes. They are similar, but not exactly alike. What I really love is the way they seem to blend in when you listen to them. Combine that with the superbly trained voice of the main singer, and you get one hell of a song.
To cut a long story short, I told her to try this song out. The effect was unbelievable.
Though voice cancellation removes the voice of the main singer, you can hear the chorus very clearly - something that's almost impossible in the uncancelled version. The lead singer just drowns the entire chorus out otherwise, so you only get a sort of vague hint of them crooning in the background.
So we played the song, and whoa! The chorus was amazing! They sang the same stuff as the lead singer, but with a slightly different tune(key? pitch? Really need to figure out what these terms mean...).
I really don't remember seeing this kind of symmetry in anything else - except mathematics - no wonder they say music and math are related...Just think about it - 2 or 3 different instruments playing closely related(but not exactly similar) tunes, and all the voices in the chorus singing away in another slightly different tune - and all that melding together seamlessly with the voice of the singer to create one fantastic piece of music...Terrific.
I'm pretty sure that composers have known about this melding effect for a long time - the words 'harmony' and 'counterpoint' come to mind. Sadly, as far as music is concerned, they're just words to me, and I'm too lazy to look them up...
And here's a question for anyone who understands filters or digital signal processing. I figure voice cancellation is accomplished by filtering out most of the frequency components of human speech. What I notice is that some instruments in some songs become noticeably louder when I cancel voice - why is this? I'm pretty sure it's not my ears playing tricks on me...
TopCoder update: I managed all 3 problems in SRM 256 - unfortunately, my 550 failed. I still haven't tested it to figure out why. Still, my score was good enough that I managed to place 5th in the room, and my rating is up 44 points...
Was a bit depressed at first, but I realized later that this was actually a good thing. College attendance worries mean that I'll be missing the next two matches. This means the next event I do will be the TopCoder Open Qualification Round - and I'll be doing it at the top of division 2, rather than at the bottom of division 1! Better chances of advancement, and the chance to win some dough :-D...
PS: This does NOT mean I'm happy about having lectures at 8 in the morning. Who the hell wants to wake up at 6 AM four days a week, and then attend college until 4 or 5 PM? No wonder most people look terribly sleepy by Friday...
Subscribe to:
Post Comments (Atom)
6 comments:
My hunch on the chorus is that its the lead singers voice, but its distorted due to voice cancellation and sounds like a chorus.
Nope. It's definitely the chorus. You can hear them in the background accompanying the lead in the uncancelled version of the song.
TopCoder: At first sight you don't respect this part of the spec: "If there is no swap that would lower the magic score, return the magic score without any swaps."
Voice cancellation: I too think there is no way a filter can possibly separate correctly voice from instruments. It's a best effort thing...
I really should have been clearer in this post - the filter doesn't completely remove the singer's voice - most of the time I can hear a softer version in the background. In cases where the singer is singing in a rather weird way, it fails miserably. Example: my sister tried it on Joss Stone's You had me, and the opening section came through exactly like the uncancelled version. Once she went back to a normal voice, it got mostly cancelled...
As for the 550 - I really should test it in the practice rooms to see why it failed...it handled all systests, and my algo was simple brute force - I don't see what could have gone wrong...
Btw, why don't you respect that spec at first sight? Any particular reason?
I was talking about your solution: it fails to conform with that part of the spec :) (i.e. I used "respect" in sense (2) not (1): http://dictionary.reference.com/search?q=respect) A simple fix is to change "for(int j = i+1..." to "for(int j = i...".
Oh - I thought you were talking about the problem statement, not my solution...I'll go over to the practice rooms and fix it up today.
From my vague memory of my solution, I think you're right - what I should have done is calculated the minimum of all possible swaps and the original score. A one line change should do the trick...
Post a Comment