import wave, struct
from struct import *
from time import sleep
import numpy as np
import matplotlib.pyplot as plt
from scipy.optimize import curve_fit

def func(x,a,b,c):
    return c * np.exp( -np.power( (x-a)/b, 2) )

 
# Ouverture du fichier wav a decrypter
wav_original = wave.open("pianoDm.wav",'rb')
 
nchannels = wav_original.getnchannels()     # Returns number of audio channels (1 for mono, 2 for stereo)

framerate = wav_original.getframerate()     # Returns sampling frequency

nframes   = wav_original.getnframes()       # Returns number of audio frames


# Decouper le fichier en plusieurs parties (une note par partie)
frequences, freq_gauss = [], []
larg_frame = 44100
for posi in range(0,nframes,larg_frame):
 
    # Sequence contenant une note
    wav_original.setpos(posi)
    donnee = wav_original.readframes(larg_frame)
    data = struct.unpack('%sh' % (larg_frame*nchannels ), donnee)
    # Transformee de Fourier
    w     = np.fft.fft(data)
    sig   = np.real(w * w.conjugate())
    freqs = np.fft.fftfreq(len(w)) * framerate
 
    # Estimation de la frequence
    idx = np.argmax(sig)
    f0, maxi = np.abs(freqs[idx]), sig[idx]
    frequences.append( f0 )
    
    # Ajustement par une gaussienne
    ind = np.where( np.abs(freqs - f0) < 20 )
    popt, pcov = curve_fit( func, freqs[ind], sig[ind]/maxi, p0=[f0,1,1] )
    a, b, c = popt
    freq_gauss.append(a)

wav_original.close()

for res in (frequences,freq_gauss):
    print(np.round(res, 0))