drum beat analysis using audio processing

hello everyone, i have a problem regarding on audio signal.

i have a sample audio signal, it is a drum beat with a different instrument on it. here a the envelope signal that i already convert it from time domain graph.

i used three drum instruments which is snare, kick pedal and cymbal. so from the diagram as you can see the first 15 signal are snare, next the second 16 signal are kick pedal, next the 14 signal are snare and the last 15 signal are cymbal. as you can see the shape of the signal for those instrument are different. so, how am i going to detect the difference between each drum instruments based on the signal shape ?

I've been looking for this solution for quite long, so i hope there gonna be a solution from you guys, thank you for helping me.

[Image Analyst]

Can you draw lines on the plot to indicate the four sections?

[faiz hazizi]

Here i provide the picture and the file also sir:

[Image Analyst]

Actually you forgot to upload the data file. Please edit the above post and attach 'Mixinstruments.wav' so we can run your code.

[faiz hazizi]

Sorry sir, here are the files.

I also forgot to mention, from the picture that I attached, the S letter is for snare, K is for Kick pedal and C is for cymbal.

Answer 1

[Von Duesenberg]

It depends on how much annotated data you have. I would start with raw labeled waveforms and try to adapt the example from the documentation here or the one there . They both use Long Short Term Memory networks.