Maria@4: # -*- coding: utf-8 -*-
Maria@4: """
Maria@4: Created on Tue Feb  2 22:26:10 2016
Maria@4: 
Maria@4: @author: mariapanteli
Maria@4: """
Maria@4: import smoothiecore as s
Maria@4: import numpy
Maria@4: import scipy.signal
Maria@4: # import librosa
Maria@4: 
Maria@4: 
Maria@4: class PitchBihist:
Maria@4:     def __init__(self, win2sec=8):
Maria@4:         self.y = None
Maria@4:         self.sr = None
Maria@4:         self.chroma = None
Maria@4:         self.chromasr = None
Maria@4:         self.bihist = None
Maria@4:         self.win2sec = win2sec
Maria@4: 
Maria@4:     def bihist_from_chroma(self, filename='test.wav', secondframedecomp=True):
Maria@4:         self.chroma, self.chromasr = s.get_smoothie_for_bihist(filename=filename, hopinsec=0.005)
Maria@4:         if secondframedecomp:
Maria@4:             win2 = int(round(8*self.chromasr))
Maria@4:             hop2 = int(round(0.5*self.chromasr))
Maria@4:             nbins, norigframes = self.chroma.shape
Maria@4:             if norigframes<win2:
Maria@4:                 nframes = 1
Maria@4:             else:
Maria@4:                 nframes = int(1+numpy.floor((norigframes-win2)/float(hop2)))
Maria@4:             bihistframes = numpy.empty((nbins*nbins, nframes))
Maria@4:             for i in range(nframes):  # loop over all 8-sec frames
Maria@4:                 frame = self.chroma[:, (i*hop2):min((i*hop2+win2),norigframes)]
Maria@4:                 bihist = self.bihistogram(frame)
Maria@4:                 bihist = numpy.reshape(bihist, -1)
Maria@4:                 bihistframes[:, i] = bihist
Maria@4:             self.bihist = bihistframes
Maria@4:         else:
Maria@4:             self.bihist = numpy.reshape(self.bihistogram(), -1)
Maria@4: 
Maria@4:     def bihist_from_melodia(self, filename='sample_melodia.csv', secondframedecomp=True, stop_sec=None):
Maria@4:         def hz_to_cents(freq_Hz, ref_Hz=32.703, n_cents=1200):
Maria@4:             """ convert frequency values from Hz to cents
Maria@4:                 reference frequency at C1 
Maria@4:             """        
Maria@4:             freq_cents = numpy.round(n_cents * numpy.log2(freq_Hz/ref_Hz))
Maria@4:             return freq_cents
Maria@4:         def wrap_to_octave(cents, octave_length=1200):
Maria@4:             """ wrap to a single octave 0-1200
Maria@4:             """
Maria@4:             octave_cents = cents % octave_length
Maria@4:             return octave_cents
Maria@4:         
Maria@4:         n_bins = 60
Maria@4:         data = numpy.loadtxt(filename, delimiter=',')
Maria@4:         times, freqs = (data[:, 0], data[:, 1])
Maria@4:         self.chromasr = 1. / (times[1] - times[0])
Maria@4:         #self.chromasr = 128.
Maria@4:         if stop_sec is not None:
Maria@4:             stop_idx = numpy.where(times < stop_sec)[0]
Maria@4:             times, freqs = times[stop_idx], freqs[stop_idx]
Maria@4:         freqs[freqs<=0] = numpy.nan
Maria@4:         #melody = freqs[freqs>0]
Maria@4:         melody = freqs
Maria@4:         n_frames = len(melody)
Maria@4:         melody_cents = hz_to_cents(melody, n_cents=n_bins)
Maria@4:         melody_octave = wrap_to_octave(melody_cents, octave_length=n_bins)
Maria@4:         melody_matrix = numpy.zeros((n_bins, n_frames))
Maria@4:         for time, pitch in enumerate(melody_octave):
Maria@4:             if not numpy.isnan(pitch):
Maria@4:                 melody_matrix[int(pitch), time] = 1
Maria@4:         if secondframedecomp:
Maria@4:             win2 = int(round(self.win2sec*self.chromasr))
Maria@4:             hop2 = int(round(0.5*self.chromasr))
Maria@4:             nbins, norigframes = melody_matrix.shape
Maria@4:             if norigframes<win2:
Maria@4:                 nframes = 1
Maria@4:                 win2 = norigframes
Maria@4:             else:
Maria@4:                 nframes = int(1+numpy.floor((norigframes-win2)/float(hop2)))
Maria@4:             bihistframes = numpy.empty((nbins*nbins, nframes))
Maria@4:             for i in range(nframes):  # loop over all 8-sec frames
Maria@4:                 frame = melody_matrix[:, (i*hop2):(i*hop2+win2)]
Maria@4:                 bihist = self.bihistogram(frame)
Maria@4:                 bihist = numpy.reshape(bihist, -1)
Maria@4:                 bihistframes[:, i] = bihist
Maria@4:             self.bihist = bihistframes
Maria@4:         else:
Maria@4:             self.bihist = self.bihistogram(melody_matrix)
Maria@4:         return self.bihist
Maria@4: 
Maria@4:     def bihistogram(self, spec, winsec=0.5, align=True):
Maria@4:         win = int(round(winsec*self.chromasr))
Maria@4:         ker = numpy.concatenate([numpy.zeros((win, 1)), numpy.ones((win+1, 1))], axis=0)
Maria@4:         spec = spec.T  # transpose to have franes as rows in convolution
Maria@4: 
Maria@4:         # energy threshold
Maria@4:         thr = 0.3*numpy.max(spec)
Maria@4:         spec[spec < max(thr, 0)] = 0
Maria@4: 
Maria@4:         # transitions via convolution
Maria@4:         tra = scipy.signal.convolve2d(spec, ker, mode='same')
Maria@4:         tra[spec > 0] = 0
Maria@4: 
Maria@4:         # multiply with original
Maria@4:         B = numpy.dot(tra.T, spec)
Maria@4: 
Maria@4:         # normalize
Maria@4:         mxB = numpy.max(B)
Maria@4:         mnB = numpy.min(B)
Maria@4:         if mxB != mnB:
Maria@4:             B = (B - mnB)/float(mxB-mnB)
Maria@4: 
Maria@4:         # circshift to highest?
Maria@4:         if align:
Maria@4:             ref = numpy.argmax(numpy.sum(spec, axis=0))
Maria@4:             B = numpy.roll(B, -ref, axis=0)
Maria@4:             B = numpy.roll(B, -ref, axis=1)
Maria@4:         return B
Maria@4: 
Maria@4:     def bihist_from_precomp_chroma(self, align=False):
Maria@4:         win2 = int(round(self.win2sec*self.chromasr))
Maria@4:         hop2 = int(round(0.5*self.chromasr))
Maria@4:         nbins, norigframes = self.chroma.shape
Maria@4:         if norigframes<win2:
Maria@4:             nframes = 1
Maria@4:         else:
Maria@4:             nframes = int(1+numpy.floor((norigframes-win2)/float(hop2)))
Maria@4:         bihistframes = numpy.empty((nbins*nbins, nframes))
Maria@4:         for i in range(nframes):  # loop over all 8-sec frames
Maria@4:             frame = self.chroma[:, (i*hop2):min((i*hop2+win2),norigframes)]
Maria@4:             bihist = self.bihistogram(frame, align=align)
Maria@4:             bihist = numpy.reshape(bihist, -1)
Maria@4:             bihistframes[:, i] = bihist
Maria@4:         self.bihist = bihistframes
Maria@4: 
Maria@4:     def get_pitchbihist(self, filename='test.wav'):
Maria@4:         self.bihist_from_chroma(filename=filename)
Maria@4:         return self.bihist
Maria@4:         
Maria@4:     def get_pitchbihist_from_chroma(self, chroma=[], chromasr=[]):
Maria@4:         self.chroma = chroma
Maria@4:         self.chromasr = chromasr
Maria@4:         self.bihist_from_precomp_chroma(align=False)
Maria@4:         return self.bihist
Maria@4: 
Maria@4: 
Maria@4: if __name__ == '__main__':
Maria@4:     pb = PitchBihist()
Maria@4:     pb.get_pitchbihist()