Somehow I couldn't shake the thought that there may be a simple way to do it.
A while ago i made a decoder for some DominoEX radio communication signals. It was a small microcontroller that took an audio signal from a radio receiver and decomposed it into its frequency components, analyzed these and turned them into data. The frequency range of this signal coincidentally corresponds to that of the human voice.
I thought that there may be a way to use part of this system to look for tones of singing. This was the idea.
I made this quick'n dirty prototype using a AVR microcontroller...