def p_function5(p):
'function : PLOT FOURIERTRANSFORM exprname COMMA expression'
pol = FourierTransform(names[p[3]], int(p[5])).compute()
p[0] = pol
def p_function3(p):
'function : PLOT FOURIERTRANSFORM LPARENT expression RPARENT'
p[0] = str(p[1]) + '(' + str(p[3]) + ')'
pol = FourierTransform(p[3]).compute()
#print pol
p[0] = pol
def p_function2(p):
'function : PLOT FOURIERTRANSFORM exprname'
equation, number_pointsamples = names[p[3]].split(',')
pol = FourierTransform(equation, int(number_pointsamples)).compute()
# print pol
p[0] = pol
def p_function1(p):
'function : FOURIERTRANSFORM LPARENT expression COMMA expression RPARENT'
pol = FourierTransform(p[3], int(p[5])).computefft()
p[0] = pol
def p_function(p):
'function : PLOT FOURIERTRANSFORM expression COMMA expression'
pol = FourierTransform(p[3], int(p[5])).compute()
#print pol
p[0] = pol
softmask.load_state_dict(torch.load(softmaskModelPathName))
softmask.eval()
#---------------------------------------------------------------------------------------------------------
freqBandCnt = 240 # number of frequency bands to produce from fast Fourier transform
(frame_size, frame_stride) = (0.030, 0.010
) # audio frame size and frame stride in seconds
# load the background white noise file
noiseFilePathName = 'C:\\Data\\WAV\\Environment\\' + 'Whitenoise2_16K' + '.wav'
#
channel, spectrogram, phasegram, frqcyBnd, sample_rate, sample_cnt, fft = FourierTransform.FFT(
noiseFilePathName,
0,
frame_size,
frame_stride,
duration=4,
emphasize=False)
spect_noise = torch.as_tensor(spectrogram, dtype=torch.float)
spect_noise /= 25 # The noise was scaled down in amplitude when it was mixed with the pure voice. It needs to be scaled down by the same scale.
(_, _, frameCnt_noise) = np.shape(spect_noise)
# load a sample audio file
pathName = "C:\\Data\\TIMIT\\Data\\TRAIN\\DR1\\MTRR0\\"
audioFile = "SI918.WAV"
#
#pathName = "C:\\Data\\TIMIT\\Data\\TEST\\DR1\\FJEM0\\"
#audioFile = "SX364.WAV"
#
# load a sample audio file