Thursday, 30 April 2015

Wavelet has strength over spectrogram

Convert audio to midi.

I have managed to decode a music into its notes accurately on my 41st birthday. My work is motivated by playing the demo music in my baby $7 toy piano. The world work is in wavelet direction.


Published 030515. It is a rough version. 


1. Save signal 050515

2. Export midi

3. Get fundamental note. Use fft. Remove harmonic.

4. Keyboard widget

5. Sound module does not run.

6. Play sound with a vertical ruler. 070515

7. Real time period finder

8. Midi in moniter

9. midi out control

10. Spectrograph

11. Unlimited time by scrolling

12. Start asyn morlet with signal


270415 finish wavelet

280415 test in starbuck. Could not use as the 1sec is too short. Amd itmay capture harminic instead. The wavelet is too wide. Try make it smaller. Try sampling freq in scale. Try wavelet constant. Try 10sec. Then 30 sec. Find fft cannot work. Divide. Try demo music. Surprisedby tge result.

has a well define freq note with time. Spectrogram is real time. Wavelet takes 7 sec to generate 1 sec image with only 14 freq band. It seem to be able to decode.

No comments:

Post a Comment