"""
Multiple Implementations of Power Spectral Entropy
"""
from utils import readMonoWav
import matplotlib.pyplot as plt
import numpy as np
np.set_printoptions(edgeitems=50)
# sampRate, sig = readMonoWav('./METU Recordings/hh2_48kHz_Mono_32bitFloat.wav')
# sampRate, sig = readMonoWav('./METU Recordings/hh2_breath/hh2_04_00.34.164_270_en.wav')
# sampRate, sig = readMonoWav('./METU Recordings/hh2_breath/hh2_09_01.20.741_134_en2.wav')
# sampRate, sig = readMonoWav('./METU Recordings/hh2_breath/hh2_09_01.20.741_134_en3_16bit.wav')
# sampRate, sig = readMonoWav('./METU Recordings/hh2_breath/hh2_23_03.02.050_149_tr.wav')
sampRate, sig = readMonoWav(
'./METU Recordings/hh2_withEdges/hh2_random001.wav')
# print(sig)
print(len(sig))
print(sampRate)
print('######')
readWholeFile = True # set to False for trials only
if readWholeFile is False:
# take a sample for better visualising, huge chunks crashes in hilbert()
sig = sig[sampRate * 0:sampRate * 1] # read only first 1 second
# sig = sig / max(sig) # normalization
partLen = sampRate // 100 # divide into 10 ms parts
plot_spectrogram(noise_stft_db, title='Noise')
if visual:
plot_statistics_and_filter(mean_freq_noise, std_freq_noise, noise_thresh, smoothing_filter)
if visual:
plot_spectrogram(sig_stft_db, title='Signal')
if visual:
plot_spectrogram(sig_mask, title='Mask applied')
if visual:
plot_spectrogram(sig_stft_db_masked, title='Masked signal')
if visual:
plot_spectrogram(recovered_spec, title='Recovered spectrogram')
return recovered_signal
filepath = './METU Recordings/hh2_bg/hh2_00.09.569-2266_bg.wav'
_, bgData = utils.readMonoWav(filepath)
filepath = './METU Recordings/hh2_speechAndBreath/hh2_00.17.614-1629_spAndBr.wav'
filename = str(os.path.basename(filepath).rsplit('.', 1)[0])
sampRate, speechData = utils.readMonoWav(filepath)
sd.play(speechData, sampRate)
time.sleep(len(speechData)/sampRate)
sd.stop()
output = removeNoise(audio_clip=speechData, noise_clip=bgData, verbose=False, visual=False)
# output = removeNoise(audio_clip=output, noise_clip=bgData, verbose=False, visual=False)
sd.play(output, sampRate)
time.sleep(len(output)/sampRate)
sd.stop()
scipy.io.wavfile.write('outputs/before.wav', sampRate, speechData)
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_00.21.576-326_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_00.24.897-274_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_00.31.490-324_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_00.42.814-516_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_01.18.383-252_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_01.32.227-273_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_01.35.925-301_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_02.59.372-200_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_03.18.690-234_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_03.35.168-364_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_03.47.911-201_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_04.25.179-287_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_04.35.304-254_nv.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_05.04.223-208_nv.wav'
sampRate, sig = readMonoWav(filepath)
filename = str(os.path.basename(filepath).rsplit('.', 1)[0])
# todo - remove when mid signal tests completed
mid = len(sig) // 2
centerFrameLenBySamples = min(int(sampRate * 0.2),
len(sig)) # center frame is 200 ms
sig = sig[mid - (centerFrameLenBySamples // 2):mid +
(centerFrameLenBySamples // 2)]
print('FileName:', filename)
print('Length:', len(sig), 'samples')
print('SamplingRate:', sampRate)
print('######')
readWholeFile = True # set to False for trials only
import numpy as np
import os
import utils
import sounddevice as sd
import time
# filepath = './TED Recordings/BillGates_2009/BillGates_2009.wav'
# filepath = './TED Recordings/BillGates_2009/bg/BillGates_2009_00.55.000-571_bg.wav'
# filepath = './TED Recordings/BillGates_2009/breath/BillGates_2009_00.32.850-525_breath.wav'
# filepath = './TED Recordings/BillGates_2009/non-voiced/BillGates_2009_00.21.576-326_nv.wav'
# filepath = './TED Recordings/BillGates_2009/speech/BillGates_2009_01.32.610-614_spch.wav'
# filepath = './METU Recordings/hh2_bg/hh2_00.09.569-2266_bg.wav'
# filepath = './METU Recordings/hh2_breath/hh2_08_01.05.315_246_en.wav'
filepath = './METU Recordings/hh2_speech/hh2_01.28.155-896_sp.wav'
sampRate, sig = utils.readMonoWav(filepath)
filename = str(os.path.basename(filepath).rsplit('.', 1)[0])
sp = np.fft.fft(sig)
freq = np.fft.fftfreq(len(sig))
plotlen = int(len(freq) / 2)
plt.figure(figsize=(10, 10))
plt.subplot(311)
plt.title('Signal: ' + filename)
plt.plot(sig)
plt.subplot(323)
plt.title('FFT Real')
plt.plot(freq[:plotlen], sp.real[:plotlen])
plt.subplot(325)
plt.title('FFT Imaginary')
plt.plot(freq[:plotlen], sp.imag[:plotlen])
"""
Code taken from:
https://pytftb.readthedocs.io/en/master/nonstationary_signals.html#instantaneous-frequency
[2018.08.30]: Tried, works as example
"""
import numpy as np
from tftb.generators import fmlin
from tftb.processing import inst_freq, plotifl
from utils import readMonoWav
import numpy
numpy.set_printoptions(edgeitems=50)
signal = readMonoWav('./METU Recordings/hh2_48kHz_Mono_32bitFloat.wav')[1]
print(signal, len(signal))
print('##########################################################################')
# signal, _ = fmlin(256)
# time_samples = np.arange(3, 257)
ifr = inst_freq(signal)[0]
print(ifr, len(ifr))
# todo - atili: try more with whole_speech and hh2 wave files
# plotifl(time_samples, ifr)