changes to lectures

AUTHORS

@@ -1,3 +1,3 @@
-Fabià Serra Arrizabalaga
 Xavier Serra
+Fabià Serra Arrizabalaga
 Sankalp Gulati

README.md

@@ -1,7 +1,7 @@
 sms-tools
 =========
 
-<p>Spectral modeling analysis and synthesis tools written in python and C for sound and music applications, plus some complementary material.</p>
+<p>Spectral modeling analysis and synthesis tools written in python and C for sound and music applications, plus complementary teaching material.</p>
 
 <p> In order to use all the software you have to install version 2.7 of python and the following modules: iPython, Matplotlib, Numpy, Scipy, PyAudio, PySide, Cython. Some of the code also requires to install the <a href="http://essentia.upf.edu/"> Essentia library</a>.  </p>
 
@@ -11,9 +11,9 @@ sms-tools
 
 <p>There are examples of analysis/transformation/synthesis in the examples directory. All the sounds used in the examples are in the sounds directory.</p>
 
-<p>In order to use the C functions, which will run most the code faster, you need to compile basicFunctions_C. Once Cython is installed, in the Terminal go to the directory software/utilFunctions_C and write <code> python compileModule.py build_ext --inplace </code> (don't bother if it appears a warning) </p>
+<p>In order to use the C functions, which will run most the code faster, you need to compile basicFunctions_C. Once Cython is installed, in the Terminal go to the directory software/utilFunctions_C and write <code> python compileModule.py build_ext --inplace </code> </p>
 
-<p>All this code is used in several classes that I teach. The slides and demo code I use in class is available in the lectures directory.</p>
+<p>All this code is used in several classes that I teach. The slides and demo code used in class is available in the lectures directory.</p>
 
 
 

lectures/7-Sinusoidal-plus-residual-model/plots-code/LPC.py

@@ -0,0 +1,41 @@
+import numpy as np
+import matplotlib.pyplot as plt
+from scipy.signal import hamming, hanning, triang, blackmanharris, resample
+import math
+import sys, os, time
+from scipy.fftpack import fft, ifft
+import essentia.standard as ess
+
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/utilFunctions/'))
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/models/'))
+import waveIO as WIO
+
+
+lpc = ess.LPC(order=14)
+N= 512
+(fs, x) = WIO.wavread('../../../sounds/soprano-E4.wav')
+first = 20000
+last = first+N
+x1 = x[first:last]
+X = fft(hamming(N)*x1) 
+mX = 20 * np.log10(abs(X[:N/2]))
+
+coeff = lpc(x1)
+Y = fft(coeff[0], N) 
+mY = 20 * np.log10(abs(Y[:N/2]))
+
+
+plt.figure(1)
+plt.subplot(2,1,1)
+plt.plot(np.arange(first, last)/float(fs), x[first:last], 'b')
+plt.axis([first/float(fs), last/float(fs), min(x[first:last]), max(x[first:last])])
+plt.title('x (soprano-E4.wav)')
+
+plt.subplot(2,1,2)
+plt.plot(np.arange(0, fs/2.0, fs/float(N)), mX-max(mX), 'r', label="mX")
+plt.plot(np.arange(0, fs/2.0, fs/float(N)), -mY-max(-mY)-3, 'k', label="mY")
+plt.legend()
+plt.axis([0, fs/2, -60, 3])
+plt.title('magnitude spectrum and LPC approximation')
+
+plt.show()

lectures/7-Sinusoidal-plus-residual-model/plots-code/envelope-approx.py

@@ -0,0 +1,60 @@
+import numpy as np
+import matplotlib.pyplot as plt
+from scipy.signal import hamming, hanning, triang, blackmanharris, resample
+import math
+import sys, os, time
+from scipy.fftpack import fft, ifft
+
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/utilFunctions/'))
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/models/'))
+import waveIO as WIO
+import dftAnal
+
+def stochasticModelFrame(x, w, N, stocf) :
+  # x: input array sound, w: analysis window, N: FFT size,  
+  # stocf: decimation factor of mag spectrum for stochastic analysis
+  hN = N/2                                                 # size of positive spectrum
+  hM = (w.size)/2                                          # half analysis window size
+  pin = hM                                                 # initialize sound pointer in middle of analysis window       
+  fftbuffer = np.zeros(N)                                  # initialize buffer for FFT
+  yw = np.zeros(w.size)                                    # initialize output sound frame
+  w = w / sum(w)                                           # normalize analysis window
+  #-----analysis-----             
+  xw = x[pin-hM:pin+hM] * w                              # window the input sound
+  X = fft(xw)                                            # compute FFT
+  mX = 20 * np.log10( abs(X[:hN]) )                      # magnitude spectrum of positive frequencies
+  mXenv = resample(np.maximum(-200, mX), mX.size*stocf)  # decimate the mag spectrum     
+  pX = np.angle(X[:hN])
+  #-----synthesis-----
+  mY = resample(mXenv, hN)                               # interpolate to original size
+  pY = 2*np.pi*np.random.rand(hN)                        # generate phase random values
+  Y = np.zeros(N, dtype = complex)
+  Y[:hN] = 10**(mY/20) * np.exp(1j*pY)                   # generate positive freq.
+  Y[hN+1:] = 10**(mY[:0:-1]/20) * np.exp(-1j*pY[:0:-1])  # generate negative freq.
+  fftbuffer = np.real( ifft(Y) )                         # inverse FFT
+  y = fftbuffer*N/2                                  
+  return mX, pX, mY, pY, y
+
+
+# example call of stochasticModel function
+if __name__ == '__main__':
+  (fs, x) = WIO.wavread('../../../sounds/ocean.wav')
+  w = np.hanning(512)
+  N = 512
+  stocf = .1
+  envSize = (N * stocf) // 2
+  maxFreq = 10000.0
+  lastbin = N*maxFreq/fs
+  first = 4000
+  last = first+w.size
+  mX, pX = dftAnal.dftAnal(x[first:last], w, N)
+  mXenv = resample(np.maximum(-200, mX), envSize)
+  mY = resample(mXenv, N/2)
+  plt.figure(1)
+  plt.plot(float(fs)*np.arange(0, N/2)/N, mX, 'r', label=r'$a$')
+  plt.plot(float(fs/2.0)*np.arange(0, envSize)/envSize, mXenv, color='k', label=r'$\tilde a$')
+  plt.plot(float(fs)*np.arange(0, N/2)/N, mY, 'g', label=r'$b$')
+  plt.legend()
+  plt.axis([0, maxFreq, -80, max(mX)+3])
+  plt.title('envelope approximation')
+  plt.show()

lectures/7-Sinusoidal-plus-residual-model/plots-code/stochasticModelAnalSynth.py

@@ -0,0 +1,50 @@
+import numpy as np
+import matplotlib.pyplot as plt
+from scipy.signal import hamming, hanning, resample
+from scipy.fftpack import fft, ifft
+import time
+import sys, os
+
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/utilFunctions/'))
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/models/'))
+
+import waveIO as WIO
+import stochasticModelAnal, stochasticModel,stftAnal
+
+(fs, x) = WIO.wavread('../../../sounds/ocean.wav')
+w = np.hamming(512)
+N = 512
+H = 256
+stocf = .1
+mXenv = stochasticModelAnal.stochasticModelAnal(x, w, N, H, stocf)
+y = stochasticModel.stochasticModel(x, w, N, H, stocf)
+mX, pX = stftAnal.stftAnal(x, fs, w, N, H)
+
+
+plt.figure(1)
+plt.subplot(411)
+plt.plot(np.arange(x.size)/float(fs), x,'b')
+plt.title('input sound x=wavread(ocean.wav)')
+plt.axis([0,x.size/float(fs),min(x),max(x)])
+
+plt.subplot(412)
+numFrames = int(mX[:,0].size)
+frmTime = H*np.arange(numFrames)/float(fs)                             
+binFreq = np.arange(N/2)*float(fs)/N                         
+plt.pcolormesh(frmTime, binFreq, np.transpose(mX))
+plt.title('magnitude spectrogram; M=512, N=512, H=256')
+plt.autoscale(tight=True)
+
+plt.subplot(413)
+numFrames = int(mXenv[:,0].size)
+frmTime = H*np.arange(numFrames)/float(fs)                             
+binFreq = np.arange(stocf*N/2)*float(fs)/(stocf*N)                       
+plt.pcolormesh(frmTime, binFreq, np.transpose(mXenv))
+plt.title('stochastic approximation; stocf=.1')
+plt.autoscale(tight=True)
+
+plt.subplot(414)
+plt.plot(np.arange(x.size)/float(fs), y,'b')
+plt.title('y')
+plt.axis([0,y.size/float(fs),min(y),max(y)])
+plt.show()

lectures/7-Sinusoidal-plus-residual-model/plots-code/stochasticModelFrame.py

@@ -1,38 +1,23 @@
 import numpy as np
 import matplotlib.pyplot as plt
-from scipy.signal import hamming, hanning, resample
+from scipy.signal import hamming, hanning, triang, blackmanharris, resample
+import math
+import sys, os, time
 from scipy.fftpack import fft, ifft
-import time
 
-import sys, os
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/utilFunctions/'))
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/models/'))
+import waveIO as WIO
 
-sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../../code/basicFunctions/'))
-sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../../code/basicFunctions_C/'))
-
-import smsWavplayer as wp
-
-try:
-  import basicFunctions_C as GS
-except ImportError:
-  import smsGenSpecSines as GS
-  print "%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%"
-  print "NOTE: Cython modules for some functions were not imported, the processing will be slow"
-  print "%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%"
-
-
-def stochasticModel(x, w, N, stocf) :
+def stochasticModelFrame(x, w, N, stocf) :
   # x: input array sound, w: analysis window, N: FFT size,  
   # stocf: decimation factor of mag spectrum for stochastic analysis
-  # y: output sound
-
   hN = N/2                                                 # size of positive spectrum
   hM = (w.size)/2                                          # half analysis window size
   pin = hM                                                 # initialize sound pointer in middle of analysis window       
   fftbuffer = np.zeros(N)                                  # initialize buffer for FFT
   yw = np.zeros(w.size)                                    # initialize output sound frame
   w = w / sum(w)                                           # normalize analysis window
-  ws = hanning(w.size)*2                                   # synthesis window
-
   #-----analysis-----             
   xw = x[pin-hM:pin+hM] * w                              # window the input sound
   X = fft(xw)                                            # compute FFT
@@ -41,46 +26,46 @@ def stochasticModel(x, w, N, stocf) :
   pX = np.angle(X[:hN])
   #-----synthesis-----
   mY = resample(mXenv, hN)                               # interpolate to original size
-  pY = 2*np.pi*np.random.rand(hN)                      # generate phase random values
+  pY = 2*np.pi*np.random.rand(hN)                        # generate phase random values
   Y = np.zeros(N, dtype = complex)
   Y[:hN] = 10**(mY/20) * np.exp(1j*pY)                   # generate positive freq.
   Y[hN+1:] = 10**(mY[:0:-1]/20) * np.exp(-1j*pY[:0:-1])  # generate negative freq.
-
   fftbuffer = np.real( ifft(Y) )                         # inverse FFT
-  y = ws*fftbuffer*N/2                     # overlap-add
-
+  y = fftbuffer*N/2                                  
   return mX, pX, mY, pY, y
 
 
 # example call of stochasticModel function
 if __name__ == '__main__':
-  (fs, x) = wp.wavread('../../../../sounds/ocean.wav')
+  (fs, x) = WIO.wavread('../../../sounds/ocean.wav')
   w = np.hanning(1024)
   N = 1024
   stocf = 0.2
   maxFreq = 10000.0
   lastbin = N*maxFreq/fs
   first = 1000
   last = first+w.size
-  mX, pX, mY, pY, y = stochasticModel(x[first:last], w, N, stocf)
+  mX, pX, mY, pY, y = stochasticModelFrame(x[first:last], w, N, stocf)
 
   plt.figure(1)
   plt.subplot(4,1,1)
-  plt.plot(np.arange(first, last)/float(fs), x[first:last]*w)
-  plt.axis([first/float(fs), last/float(fs), min(x[first:last]*w), max(x[first:last]*w)])
-  plt.title('xw')
+  plt.plot(np.arange(first, last)/float(fs), x[first:last])
+  plt.axis([first/float(fs), last/float(fs), min(x[first:last]), max(x[first:last])])
+  plt.title('x (ocean.wav)')
   plt.subplot(4,1,2)
-  plt.plot(np.arange(0, fs/2.0, fs/float(N)), mX, 'b')
-  plt.plot(np.arange(0, fs/2.0, fs/float(N)), mY, 'r')
+  plt.plot(np.arange(0, fs/2.0, fs/float(N)), mX, 'r', label="mX")
+  plt.plot(np.arange(0, fs/2.0, fs/float(N)), mY, 'k', label="mY")
+  plt.legend()
   plt.axis([0, maxFreq, -80, max(mX)+3])
-  plt.title('mX (blue), mY (red)')
+  plt.title('magnitude spectrum and approximation')
   plt.subplot(4,1,3)
-  plt.plot(np.arange(0, fs/2.0, fs/float(N)), pX, 'b')
-  plt.plot(np.arange(0, fs/2.0, fs/float(N)), pY-np.pi, 'r')
+  plt.plot(np.arange(0, fs/2.0, fs/float(N)), pX, 'c', label="pX")
+  plt.plot(np.arange(0, fs/2.0, fs/float(N)), pY-np.pi, 'k', label="pY")
   plt.axis([0, maxFreq, -np.pi, np.pi]) 
-  plt.title('pX (blue), pY (red)')
+  plt.legend()
+  plt.title('phase spectrum and random')
   plt.subplot(4,1,4)
-  plt.plot(np.arange(first, last)/float(fs), y)
+  plt.plot(np.arange(first, last)/float(fs), y, 'b')
   plt.axis([first/float(fs), last/float(fs), min(y), max(y)])
-  plt.title('yw')
+  plt.title('y')
   plt.show()

lectures/7-Sinusoidal-plus-residual-model/plots-code/stochasticSynthesisFrame.py

@@ -1,38 +1,23 @@
 import numpy as np
 import matplotlib.pyplot as plt
-from scipy.signal import hamming, hanning, resample
+from scipy.signal import hamming, hanning, triang, blackmanharris, resample
+import math
+import sys, os, time
 from scipy.fftpack import fft, ifft
-import time
 
-import sys, os
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/utilFunctions/'))
+sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../software/models/'))
+import waveIO as WIO
 
-sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../../code/basicFunctions/'))
-sys.path.append(os.path.join(os.path.dirname(os.path.realpath(__file__)), '../../../../code/basicFunctions_C/'))
-
-import smsWavplayer as wp
-
-try:
-  import basicFunctions_C as GS
-except ImportError:
-  import smsGenSpecSines as GS
-  print "%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%"
-  print "NOTE: Cython modules for some functions were not imported, the processing will be slow"
-  print "%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%"
-
-
-def stochasticModel(x, w, N, stocf) :
+def stochasticModelFrame(x, w, N, stocf) :
   # x: input array sound, w: analysis window, N: FFT size,  
   # stocf: decimation factor of mag spectrum for stochastic analysis
-  # y: output sound
-
   hN = N/2                                                 # size of positive spectrum
   hM = (w.size)/2                                          # half analysis window size
   pin = hM                                                 # initialize sound pointer in middle of analysis window       
   fftbuffer = np.zeros(N)                                  # initialize buffer for FFT
   yw = np.zeros(w.size)                                    # initialize output sound frame
   w = w / sum(w)                                           # normalize analysis window
-  ws = hanning(w.size)*2                                   # synthesis window
-
   #-----analysis-----             
   xw = x[pin-hM:pin+hM] * w                              # window the input sound
   X = fft(xw)                                            # compute FFT
@@ -41,40 +26,38 @@ def stochasticModel(x, w, N, stocf) :
   pX = np.angle(X[:hN])
   #-----synthesis-----
   mY = resample(mXenv, hN)                               # interpolate to original size
-  pY = 2*np.pi*np.random.rand(hN)                      # generate phase random values
+  pY = 2*np.pi*np.random.rand(hN)                        # generate phase random values
   Y = np.zeros(N, dtype = complex)
   Y[:hN] = 10**(mY/20) * np.exp(1j*pY)                   # generate positive freq.
   Y[hN+1:] = 10**(mY[:0:-1]/20) * np.exp(-1j*pY[:0:-1])  # generate negative freq.
-
   fftbuffer = np.real( ifft(Y) )                         # inverse FFT
-  y = ws*fftbuffer*N/2                     # overlap-add
-
+  y = fftbuffer*N/2                                  
   return mX, pX, mY, pY, y
 
 
 # example call of stochasticModel function
 if __name__ == '__main__':
-  (fs, x) = wp.wavread('../../../../sounds/ocean.wav')
+  (fs, x) = WIO.wavread('../../../sounds/ocean.wav')
   w = np.hanning(1024)
   N = 1024
   stocf = 0.1
   maxFreq = 10000.0
   lastbin = N*maxFreq/fs
   first = 1000
   last = first+w.size
-  mX, pX, mY, pY, y = stochasticModel(x[first:last], w, N, stocf)
+  mX, pX, mY, pY, y = stochasticModelFrame(x[first:last], w, N, stocf)
 
   plt.figure(1)
   plt.subplot(3,1,1)
-  plt.plot(np.arange(0, fs/2.0, fs/float(N)), mY, 'b')
+  plt.plot(np.arange(0, fs/2.0, fs/float(N)), mY, 'r', label="mY")
   plt.axis([0, maxFreq, -80, max(mX)+3])
-  plt.title('mY')
+  plt.title('magnitude spectrum (approx of mX)')
   plt.subplot(3,1,2)
-  plt.plot(np.arange(0, fs/2.0, fs/float(N)), pY-np.pi, 'b')
+  plt.plot(np.arange(0, fs/2.0, fs/float(N)), pY-np.pi, 'c', label="pY")
   plt.axis([0, maxFreq, -np.pi, np.pi]) 
-  plt.title('pY')
+  plt.title('phase spectrum (random)')
   plt.subplot(3,1,3)
-  plt.plot(np.arange(first, last)/float(fs), y)
+  plt.plot(np.arange(first, last)/float(fs), y, 'b')
   plt.axis([first/float(fs), last/float(fs), min(y), max(y)])
-  plt.title('yw')
+  plt.title('y')
   plt.show()