def motograms_from_file(file_name, wav_file_dir):
hf = h5py.File(file_name, 'r')
motograms = list()
for md5, stim_group in hf['motograms'].iteritems():
wav_file_name = os.path.join(wav_file_dir, '%s.wav' % md5)
if not os.path.exists(wav_file_name):
print 'No such file: %s' % wav_file_name
continue
intercept = stim_group.attrs['intercept']
slope = stim_group.attrs['slope']
wf = WavFile(wav_file_name, log_spectrogram=False)
wf.analyze(min_freq=300,
max_freq=8000,
spec_sample_rate=1000.0,
freq_spacing=125.0)
for split_name, split_group in stim_group.iteritems():
mgram = np.array(split_group['motogram'])
time = mgram[:, 0]
alpha = mgram[:, 1]
beta = -1.0 * (
slope * alpha + intercept
) # sign inversion from difference between Hedi's and Mike's models
mu = mgram[:, 2]
sigma1 = mgram[:, 3]
sigma2 = mgram[:, 4]
fitvals = mgram[:, 5]
motogram = Motogram()
motogram.alpha = alpha
motogram.beta = beta
motogram.mu = mu
motogram.sigma1 = sigma1
motogram.sigma2 = sigma2
motogram.fitvals = fitvals
motogram.hedi_spectrogram = np.array(split_group['spectrogram'])
motogram.hedi_spectrogram_freq = np.array(
split_group['frequencies'])
motogram.start_time = split_group.attrs['start_time'] / 1000.0
motogram.end_time = split_group.attrs['end_time'] / 1000.0
motogram.md5 = md5
motogram.wav_file = wf
motogram.wav_file_name = wav_file_name
motograms.append(motogram)
hf.close()
return motograms
def motograms_from_file(file_name, wav_file_dir):
hf = h5py.File(file_name, 'r')
motograms = list()
for md5,stim_group in hf['motograms'].iteritems():
wav_file_name = os.path.join(wav_file_dir, '%s.wav' % md5)
if not os.path.exists(wav_file_name):
print 'No such file: %s' % wav_file_name
continue
intercept = stim_group.attrs['intercept']
slope = stim_group.attrs['slope']
wf = WavFile(wav_file_name, log_spectrogram=False)
wf.analyze(min_freq=300, max_freq=8000, spec_sample_rate=1000.0, freq_spacing=125.0)
for split_name,split_group in stim_group.iteritems():
mgram = np.array(split_group['motogram'])
time = mgram[:, 0]
alpha = mgram[:, 1]
beta = -1.0 * (slope*alpha + intercept) # sign inversion from difference between Hedi's and Mike's models
mu = mgram[:, 2]
sigma1 = mgram[:, 3]
sigma2 = mgram[:, 4]
fitvals = mgram[:, 5]
motogram = Motogram()
motogram.alpha = alpha
motogram.beta = beta
motogram.mu = mu
motogram.sigma1 = sigma1
motogram.sigma2 = sigma2
motogram.fitvals = fitvals
motogram.hedi_spectrogram = np.array(split_group['spectrogram'])
motogram.hedi_spectrogram_freq = np.array(split_group['frequencies'])
motogram.start_time = split_group.attrs['start_time'] / 1000.0
motogram.end_time = split_group.attrs['end_time'] / 1000.0
motogram.md5 = md5
motogram.wav_file = wf
motogram.wav_file_name = wav_file_name
motograms.append(motogram)
hf.close()
return motograms
def add_wav_file(self, file_name, rms_thresh=40.0):
wf = WavFile(file_name)
t, f, spec, spec_rms = log_spectrogram(
wf.data,
wf.sample_rate,
spec_sample_rate=self.spec_sample_rate,
freq_spacing=self.spec_freq_spacing,
min_freq=self.min_freq,
max_freq=self.max_freq)
self.spec_t = t
self.spec_f = f
for spectrum in spec[:, spec_rms < rms_thresh].transpose():
self.samples.append(spectrum)
def __init__(self, output, duration, dt):
self.output = output
self.duration = duration
self.dt = dt
wf = WavFile()
wf.sample_rate = 1.0 / dt
wf.data = self.output[:, 0]
wf.analyze()
self.wav_file = wf
