def transformation_synthesis(inputFile1,
fs,
hfreq1,
hmag1,
stocEnv1,
inputFile2,
hfreq2,
hmag2,
stocEnv2,
hfreqIntp=np.array([0, 0, .1, 0, .9, 1, 1, 1]),
hmagIntp=np.array([0, 0, .1, 0, .9, 1, 1, 1]),
stocIntp=np.array([0, 0, .1, 0, .9, 1, 1, 1])):
"""
Transform the analysis values returned by the analysis function and synthesize the sound
inputFile1: name of input file 1
fs: sampling rate of input file 1
hfreq1, hmag1, stocEnv1: hps representation of sound 1
inputFile2: name of input file 2
hfreq2, hmag2, stocEnv2: hps representation of sound 2
hfreqIntp: interpolation factor between the harmonic frequencies of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
hmagIntp: interpolation factor between the harmonic magnitudes of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
stocIntp: interpolation factor between the stochastic representation of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# morph the two sounds
yhfreq, yhmag, ystocEnv = HPST.hpsMorph(hfreq1, hmag1, stocEnv1, hfreq2,
hmag2, stocEnv2, hfreqIntp,
hmagIntp, stocIntp)
# synthesis
y, yh, yst = HPS.hpsModelSynth(yhfreq, yhmag, np.array([]), ystocEnv, Ns,
H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(
inputFile1)[:-4] + '_hpsMorph.wav'
UF.wavwrite(y, fs, outputFile)
def main(inputFile='../../sounds/sax-phrase-short.wav', window='blackman', M=601, N=1024, t=-100, minSineDur=0.1, nH=100, minf0=350, maxf0=700, f0et=5, harmDevSlope=0.01, stocf=0.1): """ inputFile: input sound file (monophonic with sampling rate of 44100) window: analysis window type (rectangular, hanning, hamming, blackman, blackmanharris) M: analysis window size; N: fft size (power of two, bigger or equal than M) t: magnitude threshold of spectral peaks; minSineDur: minimum duration of sinusoidal tracks nH: maximum number of harmonics; minf0: minimum fundamental frequency in sound maxf0: maximum fundamental frequency in sound; f0et: maximum error accepted in f0 detection algorithm harmDevSlope: allowed deviation of harmonic tracks, higher harmonics have higher allowed deviation stocf: decimation factor used for the stochastic approximation """ # size of fft used in synthesis Ns = 512 # hop size (has to be 1/4 of Ns) H = 128 # read input sound (fs, x) = UF.wavread(inputFile) # compute analysis window w = get_window(window, M) # compute the harmonic plus stochastic model of the whole sound hfreq, hmag, hphase, stocEnv = HPS.hpsModelAnal(x, fs, w, N, H, t, nH, minf0, maxf0, f0et, harmDevSlope, minSineDur, Ns, stocf) # synthesize a sound from the harmonic plus stochastic representation y, yh, yst = HPS.hpsModelSynth(hfreq, hmag, hphase, stocEnv, Ns, H, fs) # output sound file (monophonic with sampling rate of 44100) outputFileSines = 'output_sounds/' + os.path.basename(inputFile)[:-4] + '_hpsModel_sines.wav' outputFileStochastic = 'output_sounds/' + os.path.basename(inputFile)[:-4] + '_hpsModel_stochastic.wav' outputFile = 'output_sounds/' + os.path.basename(inputFile)[:-4] + '_hpsModel.wav' # write sounds files for harmonics, stochastic, and the sum UF.wavwrite(yh, fs, outputFileSines) UF.wavwrite(yst, fs, outputFileStochastic) UF.wavwrite(y, fs, outputFile) return x, fs, hfreq, stocEnv, y
def analysis(inputFile='../../sounds/sax-phrase-short.wav',
window='blackman',
M=601,
N=1024,
t=-100,
minSineDur=0.1,
nH=100,
minf0=350,
maxf0=700,
f0et=5,
harmDevSlope=0.01,
stocf=0.1):
"""
Analyze a sound with the harmonic plus stochastic model
inputFile: input sound file (monophonic with sampling rate of 44100)
window: analysis window type (rectangular, hanning, hamming, blackman, blackmanharris)
M: analysis window size
N: fft size (power of two, bigger or equal than M)
t: magnitude threshold of spectral peaks
minSineDur: minimum duration of sinusoidal tracks
nH: maximum number of harmonics
minf0: minimum fundamental frequency in sound
maxf0: maximum fundamental frequency in sound
f0et: maximum error accepted in f0 detection algorithm
harmDevSlope: allowed deviation of harmonic tracks, higher harmonics have higher allowed deviation
stocf: decimation factor used for the stochastic approximation
returns inputFile: input file name; fs: sampling rate of input file,
hfreq, hmag: harmonic frequencies, magnitude; mYst: stochastic residual
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# read input sound
(fs, x) = UF.wavread(inputFile)
# compute analysis window
w = get_window(window, M)
# compute the harmonic plus stochastic model of the whole sound
hfreq, hmag, hphase, mYst = HPS.hpsModelAnal(x, fs, w, N, H, t, nH, minf0,
maxf0, f0et, harmDevSlope,
minSineDur, Ns, stocf)
# synthesize the harmonic plus stochastic model without original phases
y, yh, yst = HPS.hpsModelSynth(hfreq, hmag, np.array([]), mYst, Ns, H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(
inputFile)[:-4] + '_hpsModel.wav'
UF.wavwrite(y, fs, outputFile)
# create figure to plot
plt.figure(figsize=(12, 9))
# frequency range to plot
maxplotfreq = 15000.0
# plot the input sound
plt.subplot(3, 1, 1)
plt.plot(np.arange(x.size) / float(fs), x)
plt.axis([0, x.size / float(fs), min(x), max(x)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('input sound: x')
# plot spectrogram stochastic compoment
plt.subplot(3, 1, 2)
numFrames = int(mYst[:, 0].size)
sizeEnv = int(mYst[0, :].size)
frmTime = H * np.arange(numFrames) / float(fs)
binFreq = (.5 * fs) * np.arange(sizeEnv * maxplotfreq /
(.5 * fs)) / sizeEnv
plt.pcolormesh(
frmTime, binFreq,
np.transpose(mYst[:, :int(sizeEnv * maxplotfreq / (.5 * fs)) + 1]))
plt.autoscale(tight=True)
# plot harmonic on top of stochastic spectrogram
if (hfreq.shape[1] > 0):
harms = hfreq * np.less(hfreq, maxplotfreq)
harms[harms == 0] = np.nan
numFrames = int(harms[:, 0].size)
frmTime = H * np.arange(numFrames) / float(fs)
plt.plot(frmTime, harms, color='k', ms=3, alpha=1)
plt.xlabel('time (sec)')
plt.ylabel('frequency (Hz)')
plt.autoscale(tight=True)
plt.title('harmonics + stochastic spectrogram')
# plot the output sound
plt.subplot(3, 1, 3)
plt.plot(np.arange(y.size) / float(fs), y)
plt.axis([0, y.size / float(fs), min(y), max(y)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('output sound: y')
plt.tight_layout()
plt.show(block=False)
return inputFile, fs, hfreq, hmag, mYst
def transformation_synthesis(
inputFile,
fs,
hfreq,
hmag,
mYst,
freqScaling=np.array([0, 1.2, 2.01, 1.2, 2.679, .7, 3.146, .7]),
freqStretching=np.array([0, 1, 2.01, 1, 2.679, 1.5, 3.146, 1.5]),
timbrePreservation=1,
timeScaling=np.array([0, 0, 2.138, 2.138 - 1.0, 3.146, 3.146])):
"""
transform the analysis values returned by the analysis function and synthesize the sound
inputFile: name of input file
fs: sampling rate of input file
hfreq, hmag: harmonic frequencies and magnitudes
mYst: stochastic residual
freqScaling: frequency scaling factors, in time-value pairs (value of 1 no scaling)
freqStretching: frequency stretching factors, in time-value pairs (value of 1 no stretching)
timbrePreservation: 1 preserves original timbre, 0 it does not
timeScaling: time scaling factors, in time-value pairs
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# frequency scaling of the harmonics
hfreqt, hmagt = HT.harmonicFreqScaling(hfreq, hmag, freqScaling,
freqStretching, timbrePreservation,
fs)
# time scaling the sound
yhfreq, yhmag, ystocEnv = HPST.hpsTimeScale(hfreqt, hmagt, mYst,
timeScaling)
# synthesis from the trasformed hps representation
y, yh, yst = HPS.hpsModelSynth(yhfreq, yhmag, np.array([]), ystocEnv, Ns,
H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(
inputFile)[:-4] + '_hpsModelTransformation.wav'
UF.wavwrite(y, fs, outputFile)
# create figure to plot
plt.figure(figsize=(12, 6))
# frequency range to plot
maxplotfreq = 15000.0
# plot spectrogram of transformed stochastic compoment
plt.subplot(2, 1, 1)
numFrames = int(ystocEnv[:, 0].size)
sizeEnv = int(ystocEnv[0, :].size)
frmTime = H * np.arange(numFrames) / float(fs)
binFreq = (.5 * fs) * np.arange(sizeEnv * maxplotfreq /
(.5 * fs)) / sizeEnv
plt.pcolormesh(
frmTime, binFreq,
np.transpose(ystocEnv[:, :int(sizeEnv * maxplotfreq / (.5 * fs)) + 1]))
plt.autoscale(tight=True)
# plot transformed harmonic on top of stochastic spectrogram
if (yhfreq.shape[1] > 0):
harms = yhfreq * np.less(yhfreq, maxplotfreq)
harms[harms == 0] = np.nan
numFrames = int(harms[:, 0].size)
frmTime = H * np.arange(numFrames) / float(fs)
plt.plot(frmTime, harms, color='k', ms=3, alpha=1)
plt.xlabel('time (sec)')
plt.ylabel('frequency (Hz)')
plt.autoscale(tight=True)
plt.title('harmonics + stochastic spectrogram')
# plot the output sound
plt.subplot(2, 1, 2)
plt.plot(np.arange(y.size) / float(fs), y)
plt.axis([0, y.size / float(fs), min(y), max(y)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('output sound: y')
plt.tight_layout()
plt.show()
def analysis(inputFile='../../sounds/sax-phrase-short.wav', window='blackman', M=601, N=1024, t=-100,
minSineDur=0.1, nH=100, minf0=350, maxf0=700, f0et=5, harmDevSlope=0.01, stocf=0.1):
"""
Analyze a sound with the harmonic plus stochastic model
inputFile: input sound file (monophonic with sampling rate of 44100)
window: analysis window type (rectangular, hanning, hamming, blackman, blackmanharris)
M: analysis window size
N: fft size (power of two, bigger or equal than M)
t: magnitude threshold of spectral peaks
minSineDur: minimum duration of sinusoidal tracks
nH: maximum number of harmonics
minf0: minimum fundamental frequency in sound
maxf0: maximum fundamental frequency in sound
f0et: maximum error accepted in f0 detection algorithm
harmDevSlope: allowed deviation of harmonic tracks, higher harmonics have higher allowed deviation
stocf: decimation factor used for the stochastic approximation
returns inputFile: input file name; fs: sampling rate of input file,
hfreq, hmag: harmonic frequencies, magnitude; mYst: stochastic residual
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# read input sound
(fs, x) = UF.wavread(inputFile)
# compute analysis window
w = get_window(window, M)
# compute the harmonic plus stochastic model of the whole sound
hfreq, hmag, hphase, mYst = HPS.hpsModelAnal(x, fs, w, N, H, t, nH, minf0, maxf0, f0et, harmDevSlope, minSineDur, Ns, stocf)
# synthesize the harmonic plus stochastic model without original phases
y, yh, yst = HPS.hpsModelSynth(hfreq, hmag, np.array([]), mYst, Ns, H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(inputFile)[:-4] + '_hpsModel.wav'
UF.wavwrite(y,fs, outputFile)
# create figure to plot
plt.figure(figsize=(12, 9))
# frequency range to plot
maxplotfreq = 15000.0
# plot the input sound
plt.subplot(3,1,1)
plt.plot(np.arange(x.size)/float(fs), x)
plt.axis([0, x.size/float(fs), min(x), max(x)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('input sound: x')
# plot spectrogram stochastic compoment
plt.subplot(3,1,2)
numFrames = int(mYst[:,0].size)
sizeEnv = int(mYst[0,:].size)
frmTime = H*np.arange(numFrames)/float(fs)
binFreq = (.5*fs)*np.arange(sizeEnv*maxplotfreq/(.5*fs))/sizeEnv
plt.pcolormesh(frmTime, binFreq, np.transpose(mYst[:,:sizeEnv*maxplotfreq/(.5*fs)+1]))
plt.autoscale(tight=True)
# plot harmonic on top of stochastic spectrogram
if (hfreq.shape[1] > 0):
harms = hfreq*np.less(hfreq,maxplotfreq)
harms[harms==0] = np.nan
numFrames = int(harms[:,0].size)
frmTime = H*np.arange(numFrames)/float(fs)
plt.plot(frmTime, harms, color='k', ms=3, alpha=1)
plt.xlabel('time (sec)')
plt.ylabel('frequency (Hz)')
plt.autoscale(tight=True)
plt.title('harmonics + stochastic spectrogram')
# plot the output sound
plt.subplot(3,1,3)
plt.plot(np.arange(y.size)/float(fs), y)
plt.axis([0, y.size/float(fs), min(y), max(y)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('output sound: y')
plt.tight_layout()
plt.show(block=False)
return inputFile, fs, hfreq, hmag, mYst
def transformation_synthesis(inputFile, fs, hfreq, hmag, mYst, freqScaling = np.array([0, 1.2, 2.01, 1.2, 2.679, .7, 3.146, .7]),
freqStretching = np.array([0, 1, 2.01, 1, 2.679, 1.5, 3.146, 1.5]), timbrePreservation = 1,
timeScaling = np.array([0, 0, 2.138, 2.138-1.0, 3.146, 3.146])):
"""
transform the analysis values returned by the analysis function and synthesize the sound
inputFile: name of input file
fs: sampling rate of input file
hfreq, hmag: harmonic frequencies and magnitudes
mYst: stochastic residual
freqScaling: frequency scaling factors, in time-value pairs (value of 1 no scaling)
freqStretching: frequency stretching factors, in time-value pairs (value of 1 no stretching)
timbrePreservation: 1 preserves original timbre, 0 it does not
timeScaling: time scaling factors, in time-value pairs
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# frequency scaling of the harmonics
hfreqt, hmagt = HT.harmonicFreqScaling(hfreq, hmag, freqScaling, freqStretching, timbrePreservation, fs)
# time scaling the sound
yhfreq, yhmag, ystocEnv = HPST.hpsTimeScale(hfreqt, hmagt, mYst, timeScaling)
# synthesis from the trasformed hps representation
y, yh, yst = HPS.hpsModelSynth(yhfreq, yhmag, np.array([]), ystocEnv, Ns, H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(inputFile)[:-4] + '_hpsModelTransformation.wav'
UF.wavwrite(y,fs, outputFile)
# create figure to plot
plt.figure(figsize=(12, 6))
# frequency range to plot
maxplotfreq = 15000.0
# plot spectrogram of transformed stochastic compoment
plt.subplot(2,1,1)
numFrames = int(ystocEnv[:,0].size)
sizeEnv = int(ystocEnv[0,:].size)
frmTime = H*np.arange(numFrames)/float(fs)
binFreq = (.5*fs)*np.arange(sizeEnv*maxplotfreq/(.5*fs))/sizeEnv
plt.pcolormesh(frmTime, binFreq, np.transpose(ystocEnv[:,:sizeEnv*maxplotfreq/(.5*fs)+1]))
plt.autoscale(tight=True)
# plot transformed harmonic on top of stochastic spectrogram
if (yhfreq.shape[1] > 0):
harms = yhfreq*np.less(yhfreq,maxplotfreq)
harms[harms==0] = np.nan
numFrames = int(harms[:,0].size)
frmTime = H*np.arange(numFrames)/float(fs)
plt.plot(frmTime, harms, color='k', ms=3, alpha=1)
plt.xlabel('time (sec)')
plt.ylabel('frequency (Hz)')
plt.autoscale(tight=True)
plt.title('harmonics + stochastic spectrogram')
# plot the output sound
plt.subplot(2,1,2)
plt.plot(np.arange(y.size)/float(fs), y)
plt.axis([0, y.size/float(fs), min(y), max(y)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('output sound: y')
plt.tight_layout()
plt.show()
frmTime = H * np.arange(numFrames) / float(fs)
binFreq = fs * np.arange(N * maxplotfreq / fs) / N
plt.pcolormesh(frmTime, binFreq, np.transpose(mXr[:, : N * maxplotfreq / fs + 1]))
plt.xlabel("time (sec)")
plt.ylabel("frequency (Hz)")
plt.title("spectrogram of residual")
plt.autoscale(tight=True)
plt.tight_layout()
plt.savefig("cello-phrase-residual.png")
# compute the harmonic plus stochastic model
stocf = 0.5
hfreq, hmag, hphase, mYst = HPS.hpsModelAnal(
x, fs, w, N, H, t, nH, minf0, maxf0, f0et, harmDevSlope, minSineDur, Ns, stocf
)
y, yh, yst = HPS.hpsModelSynth(hfreq, hmag, np.array([]), mYst, Ns, H, fs)
UF.wavwrite(yst, fs, "cello-phrase-stochastic.wav")
UF.wavwrite(yh, fs, "cello-phrase-harmonic.wav")
UF.wavwrite(y, fs, "cello-phrase-synthesis.wav")
# plot the spectrogram of the stochastic component
plt.figure(6, figsize=(16, 4.5))
maxplotfreq = 5000.0
numFrames = int(mYst[:, 0].size)
Nst = 2 * int(mYst[0, :].size)
lastbin = int(Nst * maxplotfreq / fs)
maxplotfreq = fs * (lastbin - 1) / Nst
frmTime = H * np.arange(numFrames) / float(fs)
binFreq = fs * np.arange(lastbin) / Nst
plt.pcolormesh(frmTime, binFreq, np.transpose(mYst[:, :lastbin]))
plt.xlabel("time (sec)")
Ns = 512 H = 128 (fs1, x1) = UF.wavread(inputFile1) (fs2, x2) = UF.wavread(inputFile2) w1 = get_window(window1, M1) w2 = get_window(window2, M2) hfreq1, hmag1, hphase1, stocEnv1 = HPS.hpsModelAnal(x1, fs1, w1, N1, H, t1, nH, minf01, maxf01, f0et1, harmDevSlope1, minSineDur1, Ns, stocf) hfreq2, hmag2, hphase2, stocEnv2 = HPS.hpsModelAnal(x2, fs2, w2, N2, H, t2, nH, minf02, maxf02, f0et2, harmDevSlope2, minSineDur2, Ns, stocf) hfreqIntp = np.array([0, 0, .1, 0, .9, 1, 1, 1]) hmagIntp = np.array([0, 0, .1, 0, .9, 1, 1, 1]) stocIntp = np.array([0, 0, .1, 0, .9, 1, 1, 1]) yhfreq, yhmag, ystocEnv = HPST.hpsMorph(hfreq1, hmag1, stocEnv1, hfreq2, hmag2, stocEnv2, hfreqIntp, hmagIntp, stocIntp) y, yh, yst = HPS.hpsModelSynth(yhfreq, yhmag, np.array([]), ystocEnv, Ns, H, fs1) UF.wavwrite(y,fs1, 'hps-morph-total.wav') plt.figure(figsize=(12, 9)) # frequency range to plot maxplotfreq = 15000.0 # plot spectrogram stochastic component of sound 1 plt.subplot(3,1,1) numFrames = int(stocEnv1[:,0].size) sizeEnv = int(stocEnv1[0,:].size) frmTime = H*np.arange(numFrames)/float(fs1) binFreq = (.5*fs1)*np.arange(sizeEnv*maxplotfreq/(.5*fs1))/sizeEnv plt.pcolormesh(frmTime, binFreq, np.transpose(stocEnv1[:,:int(sizeEnv*maxplotfreq/(.5*fs1)+1)]))
def transformation_synthesis(inputFile1,
fs,
hfreq1,
hmag1,
stocEnv1,
inputFile2,
hfreq2,
hmag2,
stocEnv2,
hfreqIntp=np.array([0, 0, .1, 0, .9, 1, 1, 1]),
hmagIntp=np.array([0, 0, .1, 0, .9, 1, 1, 1]),
stocIntp=np.array([0, 0, .1, 0, .9, 1, 1, 1])):
"""
Transform the analysis values returned by the analysis function and synthesize the sound
inputFile1: name of input file 1
fs: sampling rate of input file 1
hfreq1, hmag1, stocEnv1: hps representation of sound 1
inputFile2: name of input file 2
hfreq2, hmag2, stocEnv2: hps representation of sound 2
hfreqIntp: interpolation factor between the harmonic frequencies of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
hmagIntp: interpolation factor between the harmonic magnitudes of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
stocIntp: interpolation factor between the stochastic representation of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# morph the two sounds
yhfreq, yhmag, ystocEnv = HPST.hpsMorph(hfreq1, hmag1, stocEnv1, hfreq2,
hmag2, stocEnv2, hfreqIntp,
hmagIntp, stocIntp)
# synthesis
y, yh, yst = HPS.hpsModelSynth(yhfreq, yhmag, np.array([]), ystocEnv, Ns,
H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(
inputFile1)[:-4] + '_hpsMorph.wav'
UF.wavwrite(y, fs, outputFile)
# create figure to plot
plt.figure(figsize=(12, 9))
# frequency range to plot
maxplotfreq = 15000.0
# plot spectrogram of transformed stochastic compoment
plt.subplot(2, 1, 1)
numFrames = int(ystocEnv[:, 0].size)
sizeEnv = int(ystocEnv[0, :].size)
frmTime = H * np.arange(numFrames) / float(fs)
binFreq = (.5 * fs) * np.arange(sizeEnv * maxplotfreq /
(.5 * fs)) / sizeEnv
plt.pcolormesh(
frmTime, binFreq,
np.transpose(ystocEnv[:, :int(sizeEnv * maxplotfreq / (.5 * fs)) + 1]))
plt.autoscale(tight=True)
# plot transformed harmonic on top of stochastic spectrogram
if (yhfreq.shape[1] > 0):
harms = np.copy(yhfreq)
harms = harms * np.less(harms, maxplotfreq)
harms[harms == 0] = np.nan
numFrames = int(harms[:, 0].size)
frmTime = H * np.arange(numFrames) / float(fs)
plt.plot(frmTime, harms, color='k', ms=3, alpha=1)
plt.xlabel('time (sec)')
plt.ylabel('frequency (Hz)')
plt.autoscale(tight=True)
plt.title('harmonics + stochastic spectrogram')
# plot the output sound
plt.subplot(2, 1, 2)
plt.plot(np.arange(y.size) / float(fs), y)
plt.axis([0, y.size / float(fs), min(y), max(y)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('output sound: y')
plt.tight_layout()
plt.show()
def transformation_synthesis(inputFile1, fs, hfreq1, hmag1, stocEnv1, inputFile2, hfreq2, hmag2, stocEnv2,
hfreqIntp = np.array([0, 0, .1, 0, .9, 1, 1, 1]), hmagIntp = np.array([0, 0, .1, 0, .9, 1, 1, 1]), stocIntp = np.array([0, 0, .1, 0, .9, 1, 1, 1])):
"""
Transform the analysis values returned by the analysis function and synthesize the sound
inputFile1: name of input file 1
fs: sampling rate of input file 1
hfreq1, hmag1, stocEnv1: hps representation of sound 1
inputFile2: name of input file 2
hfreq2, hmag2, stocEnv2: hps representation of sound 2
hfreqIntp: interpolation factor between the harmonic frequencies of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
hmagIntp: interpolation factor between the harmonic magnitudes of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
stocIntp: interpolation factor between the stochastic representation of the two sounds, 0 is sound 1 and 1 is sound 2 (time,value pairs)
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# morph the two sounds
yhfreq, yhmag, ystocEnv = HPST.hpsMorph(hfreq1, hmag1, stocEnv1, hfreq2, hmag2, stocEnv2, hfreqIntp, hmagIntp, stocIntp)
# synthesis
y, yh, yst = HPS.hpsModelSynth(yhfreq, yhmag, np.array([]), ystocEnv, Ns, H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(inputFile1)[:-4] + '_hpsMorph.wav'
UF.wavwrite(y, fs, outputFile)
# create figure to plot
plt.figure(figsize=(12, 9))
# frequency range to plot
maxplotfreq = 15000.0
# plot spectrogram of transformed stochastic compoment
plt.subplot(2,1,1)
numFrames = int(ystocEnv[:,0].size)
sizeEnv = int(ystocEnv[0,:].size)
frmTime = H*np.arange(numFrames)/float(fs)
binFreq = (.5*fs)*np.arange(sizeEnv*maxplotfreq/(.5*fs))/sizeEnv
plt.pcolormesh(frmTime, binFreq, np.transpose(ystocEnv[:,:sizeEnv*maxplotfreq/(.5*fs)+1]))
plt.autoscale(tight=True)
# plot transformed harmonic on top of stochastic spectrogram
if (yhfreq.shape[1] > 0):
harms = np.copy(yhfreq)
harms = harms*np.less(harms,maxplotfreq)
harms[harms==0] = np.nan
numFrames = int(harms[:,0].size)
frmTime = H*np.arange(numFrames)/float(fs)
plt.plot(frmTime, harms, color='k', ms=3, alpha=1)
plt.xlabel('time (sec)')
plt.ylabel('frequency (Hz)')
plt.autoscale(tight=True)
plt.title('harmonics + stochastic spectrogram')
# plot the output sound
plt.subplot(2,1,2)
plt.plot(np.arange(y.size)/float(fs), y)
plt.axis([0, y.size/float(fs), min(y), max(y)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('output sound: y')
plt.tight_layout()
plt.show(block=False)
w = np.blackman(601)
N = 1024
t = -100
nH = 100
minf0 = 350
maxf0 = 700
f0et = 5
minSineDur = .1
harmDevSlope = 0.01
Ns = 512
H = Ns / 4
stocf = .2
hfreq, hmag, hphase, mYst = HPS.hpsModelAnal(x, fs, w, N, H, t, nH, minf0,
maxf0, f0et, harmDevSlope,
minSineDur, Ns, stocf)
y, yh, yst = HPS.hpsModelSynth(hfreq, hmag, hphase, mYst, Ns, H, fs)
maxplotfreq = 10000.0
plt.figure(1, figsize=(9, 7))
plt.subplot(311)
plt.plot(np.arange(x.size) / float(fs), x, 'b')
plt.autoscale(tight=True)
plt.title('x (sax-phrase-short.wav)')
plt.subplot(312)
numFrames = int(mYst[:, 0].size)
sizeEnv = int(mYst[0, :].size)
frmTime = H * np.arange(numFrames) / float(fs)
binFreq = (.5 * fs) * np.arange(sizeEnv * maxplotfreq / (.5 * fs)) / sizeEnv
plt.pcolormesh(frmTime, binFreq,
def transformation_synthesis_stereo(inputFile, fs, hfreq, hmag, mYst, freqScaling = np.array([0, 1.2, 2.01, 1.2, 2.679, .7, 3.146, .7]),
freqStretching = np.array([0, 1, 2.01, 1, 2.679, 1.5, 3.146, 1.5]), timbrePreservation = 1,
timeScaling = np.array([0, 0, 2.138, 2.138-1.0, 3.146, 3.146]),
inputSound=[]):
"""
transform the analysis values returned by the analysis function and synthesize the sound
inputFile: name of input file
fs: sampling rate of input file
hfreq, hmag: harmonic frequencies and magnitudes
mYst: stochastic residual
freqScaling: tuple (for L and R channels) of arrays of frequency scaling factors, in time-value pairs (value of 1 no scaling)
freqStretching: tuple (for L and R channels) of arrays of frequency stretching factors, in time-value pairs (value of 1 no stretching)
timbrePreservation: tuple (for L and R channels) of arrays of 1 preserves original timbre, 0 it does not
timeScaling: tuple (for L and R channels) of arrays of time scaling factors, in time-value pairs
inputSound: original sound, used for auto-attenuation of the output sound
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# frequency scaling of the harmonics
hfreqtLeft , hmagtLeft = HT.harmonicFreqScaling(hfreq, hmag, freqScaling[0], freqStretching[0], timbrePreservation[0], fs)
hfreqtRight, hmagtRight = HT.harmonicFreqScaling(hfreq, hmag, freqScaling[1], freqStretching[1], timbrePreservation[1], fs)
# time scaling the sound
yhfreqLeft , yhmagLeft , ystocEnvLeft = HPST.hpsTimeScale(hfreqtLeft , hmagtLeft , mYst, timeScaling[0])
yhfreqRight, yhmagRight, ystocEnvRight = HPST.hpsTimeScale(hfreqtRight, hmagtRight, mYst, timeScaling[1])
# synthesis from the trasformed hps representation
yLeft, yhLeft, ystLeft = HPS.hpsModelSynth(yhfreqLeft , yhmagLeft , np.array([]), ystocEnvLeft , Ns, H, fs)
yRight, yhRight, ystRight = HPS.hpsModelSynth(yhfreqRight, yhmagRight, np.array([]), ystocEnvRight, Ns, H, fs)
# write output sound
outputFile = 'output_sounds/' + os.path.basename(inputFile)[:-4] + '_hpsModelTransformation.wav'
UF.wavwriteStereo(yLeft, yRight, fs, outputFile, inputSound)
# create figure to plot
plt.figure(figsize=(12, 12))
# frequency range to plot
maxplotfreq = 15000.0
def plotTransformedSignals(channelName,subplot1,subplot2,yhfreq,ystocEnv,y):
# plot spectrogram of transformed stochastic compoment
plt.subplot(subplot1)
numFrames = int(ystocEnv[:,0].size)
sizeEnv = int(ystocEnv[0,:].size)
frmTime = H*np.arange(numFrames)/float(fs)
binFreq = (.5*fs)*np.arange(sizeEnv*maxplotfreq/(.5*fs))/sizeEnv
plt.pcolormesh(frmTime, binFreq, np.transpose(ystocEnv[:,:sizeEnv*maxplotfreq/(.5*fs)+1]))
plt.autoscale(tight=True)
# plot transformed harmonic on top of stochastic spectrogram
if (yhfreq.shape[1] > 0):
harms = yhfreq*np.less(yhfreq,maxplotfreq)
harms[harms==0] = np.nan
numFrames = int(harms[:,0].size)
frmTime = H*np.arange(numFrames)/float(fs)
plt.plot(frmTime, harms, color='k', ms=3, alpha=1)
plt.xlabel('time (sec)')
plt.ylabel('frequency (Hz)')
plt.autoscale(tight=True)
plt.title('harmonics + stochastic spectrogram (' + channelName + ')')
# plot the output sound
plt.subplot(subplot2)
plt.plot(np.arange(y.size)/float(fs), y)
plt.axis([0, y.size/float(fs), min(y), max(y)])
plt.ylabel('amplitude')
plt.xlabel('time (sec)')
plt.title('output sound: y (' + channelName + ')')
plotTransformedSignals('Left' , 411, 413, yhfreqLeft, ystocEnvLeft, yLeft)
plotTransformedSignals('Right', 412, 414, yhfreqRight, ystocEnvRight, yRight)
plt.tight_layout()
plt.show()
sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../software/models/'))
import hpsModel as HPS
import utilFunctions as UF
if __name__ == '__main__':
(fs, x) = UF.wavread(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../sounds/sax-phrase.wav'))
w = np.blackman(601)
N = 1024
t = -100
nH = 100
minf0 = 350
maxf0 = 700
f0et = 5
maxnpeaksTwm = 5
minSineDur = .1
harmDevSlope = 0.01
Ns = 512
H = Ns/4
stocf = .2
hfreq, hmag, hphase, mYst = HPS.hpsModelAnal(x, fs, w, N, H, t, nH, minf0, maxf0, f0et, harmDevSlope, minSineDur, Ns, stocf)
y, yh, yst = HPS.hpsModelSynth(hfreq, hmag, hphase, mYst, Ns, H, fs)
# UF.play(y, fs)
# UF.play(yh, fs)
# UF.play(yst, fs)
UF.wavwrite(y,fs,'sax-phrase-total-synthesis.wav')
UF.wavwrite(yh,fs,'sax-phrase-harmonic-component.wav')
UF.wavwrite(yst,fs,'sax-phrase-stochastic-component.wav')
def main(
inputFile="../../sounds/sax-phrase-short.wav",
window="blackman",
M=601,
N=1024,
t=-100,
minSineDur=0.1,
nH=100,
minf0=350,
maxf0=700,
f0et=5,
harmDevSlope=0.01,
stocf=0.1,
):
"""
inputFile: input sound file (monophonic with sampling rate of 44100)
window: analysis window type (rectangular, hanning, hamming, blackman, blackmanharris)
M: analysis window size; N: fft size (power of two, bigger or equal than M)
t: magnitude threshold of spectral peaks; minSineDur: minimum duration of sinusoidal tracks
nH: maximum number of harmonics; minf0: minimum fundamental frequency in sound
maxf0: maximum fundamental frequency in sound; f0et: maximum error accepted in f0 detection algorithm
harmDevSlope: allowed deviation of harmonic tracks, higher harmonics have higher allowed deviation
stocf: decimation factor used for the stochastic approximation
"""
# size of fft used in synthesis
Ns = 512
# hop size (has to be 1/4 of Ns)
H = 128
# read input sound
(fs, x) = UF.wavread(inputFile)
# compute analysis window
w = get_window(window, M)
# compute the harmonic plus stochastic model of the whole sound
hfreq, hmag, hphase, stocEnv = HPS.hpsModelAnal(
x, fs, w, N, H, t, nH, minf0, maxf0, f0et, harmDevSlope, minSineDur, Ns, stocf
)
# synthesize a sound from the harmonic plus stochastic representation
y, yh, yst = HPS.hpsModelSynth(hfreq, hmag, hphase, stocEnv, Ns, H, fs)
# output sound file (monophonic with sampling rate of 44100)
outputFileSines = "output_sounds/" + os.path.basename(inputFile)[:-4] + "_hpsModel_sines.wav"
outputFileStochastic = "output_sounds/" + os.path.basename(inputFile)[:-4] + "_hpsModel_stochastic.wav"
outputFile = "output_sounds/" + os.path.basename(inputFile)[:-4] + "_hpsModel.wav"
# write sounds files for harmonics, stochastic, and the sum
UF.wavwrite(yh, fs, outputFileSines)
UF.wavwrite(yst, fs, outputFileStochastic)
UF.wavwrite(y, fs, outputFile)
# create figure to plot
plt.figure(figsize=(12, 9))
# frequency range to plot
maxplotfreq = 15000.0
# plot the input sound
plt.subplot(3, 1, 1)
plt.plot(np.arange(x.size) / float(fs), x)
plt.axis([0, x.size / float(fs), min(x), max(x)])
plt.ylabel("amplitude")
plt.xlabel("time (sec)")
plt.title("input sound: x")
# plot spectrogram stochastic component
plt.subplot(3, 1, 2)
numFrames = int(stocEnv[:, 0].size)
sizeEnv = int(stocEnv[0, :].size)
frmTime = H * np.arange(numFrames) / float(fs)
binFreq = (0.5 * fs) * np.arange(sizeEnv * maxplotfreq / (0.5 * fs)) / sizeEnv
plt.pcolormesh(frmTime, binFreq, np.transpose(stocEnv[:, : sizeEnv * maxplotfreq / (0.5 * fs) + 1]))
plt.autoscale(tight=True)
# plot harmonic on top of stochastic spectrogram
if hfreq.shape[1] > 0:
harms = hfreq * np.less(hfreq, maxplotfreq)
harms[harms == 0] = np.nan
numFrames = harms.shape[0]
frmTime = H * np.arange(numFrames) / float(fs)
plt.plot(frmTime, harms, color="k", ms=3, alpha=1)
plt.xlabel("time (sec)")
plt.ylabel("frequency (Hz)")
plt.autoscale(tight=True)
plt.title("harmonics + stochastic spectrogram")
# plot the output sound
plt.subplot(3, 1, 3)
plt.plot(np.arange(y.size) / float(fs), y)
plt.axis([0, y.size / float(fs), min(y), max(y)])
plt.ylabel("amplitude")
plt.xlabel("time (sec)")
plt.title("output sound: y")
plt.tight_layout()
plt.show(block=False)
