process.py

from scipy.io.wavfile import read, write
import numpy as np
import matplotlib.pyplot as plt
import scipy.signal as signal
import os

def fitness(f1,f2):
    f1_norm = f1/sum(f1)
    f2_norm = f2/sum(f2)

    MSE = 0
    for i in range(len(f1)):
        d = f1_norm[i] - f2_norm[i]     # diff in each time slice

        MSE_t = 0
        for j in range(len(d)):
            MSE_t += d[j]**2        # SE in time slice

        MSE += MSE_t/len(f1)            # total up each slice and take the mean
    
    return MSE

def crop(audio,t):
    fs = 44100
    length = len(audio)
    audio_crop = []
    for i in range(length):
        x, y = audio[i]
        if y>0.9:
            for j in range(i, i+fs*t):
                p = audio[j]
                audio_crop.append(p)
            break

    return np.array(audio_crop)
    

def freq(data):
    fs = 44100
    f, t, Sxx = signal.spectrogram(np.array(data[:,1]), fs, nperseg = 3000, nfft = 10000 )
    print(len(f),len(t))
    plt.figure()
    plt.pcolormesh(t, f, Sxx)
    # plt.title('n1,n2 = {},{}'.format(n1,n2))
    plt.ylabel('Frequency [Hz]')
    plt.xlabel('Time [sec]')
    # plt.ylim(0,3000)
    return Sxx
    # plt.ylim(0, 2000)

def convert_data_x(input, output):
    files = os.listdir(input)
    for i in range(200):
        print('Converting File {}'.format(i))
        audio = read(input + '/' + files[i])[1]
        audio_crop = crop(audio,2)
        params = np.load(direc+'paramsdata/params{}.npy'.format(i))[param_list]
        
        np.save(output + '/audio{}'.format(i),np.array([audio_crop,params]),allow_pickle=True)

def convert_data_f(input, output):
    files = os.listdir(input)
    print(len(files))
    for i in range(len(files)):
        print('Converting File {}'.format(i))
        audio = read(input + '/' + files[i])[1]
        audio_crop = crop(audio,2)
        image = img(audio_crop)
        size = np.shape(image)[1]
        if size != 67:
            print('WRONG SIZE', size)
        params = np.load(direc+'paramsdata/params{}.npy'.format(i))[param_list]
        
        np.save(output + '/img{}'.format(i),np.array([image,params]),allow_pickle=True)
        # np.save(output + '/image{}'.format(i),np.array(image),allow_pickle=True)

def img(audio):
    fs = 44100
    f, t, Sxx = signal.spectrogram(np.array(audio[:,1]), fs, nperseg = 1500 , nfft = 10000 )

    # plt.imshow(Sxx, aspect ='auto')
    # plt.show()
    return Sxx

if __name__ == "__main__":
    # param_list = [0,1,2,3,4,5,6,
    # 
    # ,8,9,10,14,15,16,17,18,19,20,21,22,23,24,25] # all oscillators
    param_list = [0,1,14,15,16,17,18,19,20,21,22,23,24,25] # single oscillator

    direc = '../data/1osc/testdata'
    audio = read(direc + '/' + 'audiodata/audio1.wav')[1]
    img(audio)
    # convert_data_f(direc+'audiodata',direc+'TRAIN_f')
    # convert_data_x('../data/1osc/audiodata','../data/1osc/TRAIN_x')