def analysisSynthesis(params, signal):
outsignal = array(0)
# framecutter > windowing > FFT > IFFT > OverlapAdd
frames = cutFrames(params, signal)
w = std.Windowing(type="hann")
fft = std.FFT(size=params['frameSize'])
ifft = std.IFFT(size=params['frameSize'])
overl = std.OverlapAdd(frameSize=params['frameSize'],
hopSize=params['hopSize'])
counter = 0
for f in frames:
#outframe = OverlapAdd(frameSize = params['frameSize'], hopSize = params['hopSize'])(IFFT(size = params['frameSize'])(FFT(size = params['frameSize'])(Windowing()(f))))
# STFT analysis
infft = fft(w(f))
# here we could apply spectral transformations
outfft = infft
# STFT synthesis
ifftframe = ifft(outfft)
of = ifftframe
outframe = overl(of)
if counter >= (params['frameSize'] / (2 * params['hopSize'])):
outsignal = numpy.append(outsignal, outframe)
counter += 1
return outsignal
def analysisSynthesis(params, signal):
outsignal = array(0)
signal = numpy.append(signal, zeros(params['frameSize']/2))
frames = cutFrames(params, signal)
w = std.Windowing(type = "hann");
fft = std.FFT(size = params['frameSize']);
ifft = std.IFFT(size = params['frameSize']);
overl = std.OverlapAdd (frameSize = params['frameSize'], hopSize = params['hopSize'], gain = 1./params['frameSize']);
counter = 0
for f in frames:
# STFT analysis
infft = fft(w(f))
# here we could apply spectral transformations
outfft = infft
# STFT synthesis
ifftframe = ifft(outfft)
of = ifftframe
outframe = overl(of)
if counter >= (params['frameSize']/(2*params['hopSize'])):
outsignal = numpy.append(outsignal,outframe)
counter += 1
return outsignal
def analysisSynthesisStandard(params, signal):
w = std.Windowing(type = "hann");
fft = std.FFT(size = params['frameSize']);
ifft = std.IFFT(size = params['frameSize']);
overl = std.OverlapAdd (frameSize = params['frameSize'], hopSize = params['hopSize'], gain = 1./params['frameSize']);
# add half window of zeros to input signal to reach same ooutput length
signal = numpy.append(signal, zeros(params['frameSize']/2))
frames = cutFrames(params, signal)
outsignal = []
counter = 0
outframe = array(0)
for f in frames:
outframe = overl(ifft(fft(w(f))))
outsignal = numpy.append(outsignal,outframe)
outsignal = outsignal [2*params['hopSize']:]
return outsignal
import matplotlib.pylab as plt
if mode == 'standard':
# create an audio loader and import audio file
loader = std.MonoLoader(filename = inputFilename, sampleRate = 44100)
audio = loader()
print("Duration of the audio sample [sec]:")
print(len(audio)/44100.0)
w = std.Windowing(type = "hann");
fft = std.FFT(size = framesize);
ifft = std.IFFT(size = framesize);
overl = std.OverlapAdd (frameSize = framesize, hopSize = hopsize);
awrite = std.MonoWriter (filename = outputFilename, sampleRate = 44100);
for frame in std.FrameGenerator(audio, frameSize = framesize, hopSize = hopsize):
# STFT analysis
infft = fft(w(frame))
# here we could apply spectral transformations
outfft = infft
# STFT synthesis
ifftframe = ifft(outfft)
out = overl(ifftframe)
def analysis_synthesis_spr_model_standard(self, params, signal):
pool = essentia.Pool()
# Streaming Algos for Sine Model Analysis
w = es.Windowing(type="hann")
fft = es.FFT(size=params['fftSize'])
smanal = es.SineModelAnal(
sampleRate=params['sampleRate'],
maxnSines=params['maxnSines'],
magnitudeThreshold=params['magnitudeThreshold'],
freqDevOffset=params['freqDevOffset'],
freqDevSlope=params['freqDevSlope'])
# Standard Algos for Sine Model Analysis
smsyn = es.SineModelSynth(sampleRate=params['sampleRate'],
fftSize=params['frameSize'],
hopSize=params['hopSize'])
ifft = es.IFFT(size=params['frameSize'])
overlSine = es.OverlapAdd(frameSize=params['frameSize'],
hopSize=params['hopSize'],
gain=1. / params['frameSize'])
overlres = es.OverlapAdd(frameSize=params['frameSize'],
hopSize=params['hopSize'],
gain=1. / params['frameSize'])
fft_original = []
# analysis
for frame in es.FrameGenerator(signal,
frameSize=params["frameSize"],
hopSize=params["hopSize"]):
frame_fft = fft(w(frame))
fft_original.append(frame_fft)
freqs, mags, phases = smanal(frame_fft)
pool.add("frequencies", freqs)
pool.add("magnitudes", mags)
pool.add("phases", phases)
# remove short tracks
minFrames = int(params['minSineDur'] * params['sampleRate'] /
params['hopSize'])
pool = self.cleaningSineTracks(pool, minFrames)
# synthesis
sineTracksAudio = np.array([])
resTracksAudio = np.array([])
for frame_ix, _ in enumerate(pool["frequencies"]):
sine_frame_fft = smsyn(pool["magnitudes"][frame_ix],
pool["frequencies"][frame_ix],
pool["phases"][frame_ix])
res_frame_fft = fft_original[frame_ix] - sine_frame_fft
sine_outframe = overlSine(ifft(sine_frame_fft))
sineTracksAudio = np.append(sineTracksAudio, sine_outframe)
res_outframe = overlres(ifft(res_frame_fft))
resTracksAudio = np.append(resTracksAudio, res_outframe)
sineTracksAudio = sineTracksAudio.flatten()[-len(signal):]
resTracksAudio = resTracksAudio.flatten()[-len(signal):]
#print("len signal", len(signal), "len res", len(resTracksAudio))
return essentia.array(signal), essentia.array(
sineTracksAudio), essentia.array(resTracksAudio)
