Singing Voice to Amplitude and Fundamental Frequency
in [DSP]
|
Prev: ESQ Centurion Men's_Watch Watch 07300907 - Replica Watch Fake
Next: Question on Frame Number in GSM synchronization channel
From: Magnus.Moraberg on 19 Apr 2008 08:48 Hi, When I did my ungraduate studies I designed a simple software algorithm to extract note information from a guitar track. I extracted notes from guitar chords using an sliding fft. Taking the magnitufe spectrum, I multiplied the value in one bin by that of its prime multiples. In other words, I multipled the value in bin 10 by that found in bin 20, 30, 50, 70, 110, and 130. This worked great when extracting notes from major and minor type chords where notes share a large number of harmonics. Since then I've designed a number of electronaic instruments which rather than taking midi input, take a note value and an amplitude value pertainent to a point in time. Therefore the signal sent to the instrument looks something like this - 00ms A4+40 0dB 10ms A4+30 5dB 20ms A4+20 10dB 30ms A4+10 15dB 40ms A4+00 20dB 50ms A4-10 25dB 60ms A4-20 30dB What I'm interested in now is process a human voice in order to extract the same information listed above. So I wish to convert a singing voice to the following - A(t)*sin(2*pi*f(t) + ph(t)) I think my old approach to pitch tracking is not required here and was wondering if you guys cold suggest some schemes, papers etc which might be of interest to me. Perhaps there is software available also which I might be interested in. Thanks for your help, Barry. |