def beat_spectrum_unit_fromFile(self,
fname,
framesize=1024,
hopsize=512,
n_dim=128):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
return self.beat_spectrum_unit(x, fs, framesize, hopsize, n_dim)
def beat_spectrum_4beat_fromFile(self,
fname,
bpm,
framesize=1024,
hopsize=512,
n_lags=256):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
return self.beat_spectrum_4beat(x, fs, bpm, framesize, hopsize, n_lags)
def spectrogram(self,
fname,
framesize=1024,
hopsize=512,
window='hamming'):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
X, F, T = spectral.stft(x, framesize, hopsize, fs, window)
X = sp.absolute(X)
return X, F, T
def segmental_spectral_kurtosis(self,
fname,
bpm,
framesize=2048,
hopsize=512,
segsize=32):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
SK = spectral.spectral_kurtosis(x, framesize, hopsize, fs)
SK = segment.normalize_time_axis_by_bpm(SK, framesize, hopsize, fs,
bpm, segsize)
return SK
def segmental_mel_spectrogram_bpm(self,
fname,
bpm,
beatunit=16,
framesize=1024,
hopsize=512,
n_ceps=22):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
melspe = timbre.mel_spectrogram(x, framesize, hopsize, fs, 'barthann',
7000, n_ceps)
melspe = segment.normalize_time_axis_by_bpm(melspe, bpm, beatunit,
framesize, hopsize, fs)
trans_melspe = segment.transform_time_axis_to_one_bar(melspe, beatunit)
return trans_melspe
def segmental_cqt_bpm(self,
fname,
bpm,
beatunit=16,
framesize=1024,
hopsize=512,
n_cq=88,
n_per_semitone=1,
fmin=60.0):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
X, F, T = spectral.cqt(x, framesize, hopsize, fs,
n_cq * n_per_semitone, n_per_semitone, fmin)
Xseg = segment.normalize_time_axis_by_bpm(X.T, bpm, beatunit,
framesize, hopsize, fs).T
return Xseg
def segmental_mfcc_bpm(self,
fname,
bpm,
beatunit=16,
framesize=1024,
hopsize=512,
max_freq=22050,
n_ceps=22,
nopower=False):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
mfccdata = timbre.mfcc(x, framesize, hopsize, fs, max_freq, n_ceps)
if nopower:
mfccdata = mfccdata[:, 1:]
mfccdata = segment.normalize_time_axis_by_bpm(mfccdata, bpm, beatunit,
framesize, hopsize, fs)
trans_mfcc = segment.transform_time_axis_to_one_bar(mfccdata, beatunit)
return trans_mfcc
def segmental_beat_spectrogram(self,
fname,
bpm,
segsize=32,
framesize=1024,
hopsize=512,
n_lags=256):
x, fs, bit = audioio.wavread(fname)
x = audioio.stereo_to_mono(x)
# BS = rhythm.beat_spectrogram_kurth_4beat_force20sec_cy(x, framesize, hopsize, fs, 'hamming', bpm, n_lags)
# BS = signal.smoothByBeat(BS, framesize, hopsize, fs, bpm, segsize)
BS = sp.zeros((segsize * 4, n_lags))
BS[:] = rhythm.beat_spectrum_kurth_4beat_force20sec_cy(
x, framesize, hopsize, fs, 'barthann', bpm, n_lags)
# B = rhythm.beat_spectrum_kurth_4beat_force20sec_cy(x, framesize, hopsize, fs, 'hamming', bpm, n_lags)
# BS = sp.zeros( (segsize*4, 256) )
# for f in xrange(segsize*4):
# BS[f,:] = B
return BS