def abstract_cartoon(wave,
window_size=64,
resample_size=32768,
window_func=signal.hann,
freq=False,
norm=True):
#whole wave analysis using a set transform window
#TODO: resample the "cartoon" to a certain length and find trend.
#resample wave to a set length
resampled_wave = signal.resample(wave.data, resample_size)
sr = wave.sr
#segmentation:
segments = fileio.segment(resampled_wave, window_size)
#TODO: test this
#try frequency:
if freq:
transformed_segments = [fft_to_freq(fft_extract(seg, window_func), sr, 1)[0] for seg in segments]
else:
transformed_segments = [fft_extract(seg, window_func).max() for seg in segments]
#ft = fftpack.dct(wave.data * window_type(len(wave.data)))
#window this?
transformed_segments = syllable.moving_window(transformed_segments, window_size, window_func)
if norm:
transformed_segments = normalize(transformed_segments)
#1st derivative for trend detection:
#return np.diff(transformed_segments)
return transformed_segments
def plotstuff(filename):
a=fileio.Wave("audio/samples/"+filename+".wav")
am, fr = syllable.segmentation(a)
title(filename)
plot(feature.normalize(syllable.moving_window(am, 10)), label="amp")
#plot(feature.normalize(syllable.moving_window(fr, 10)), label="freq")
#plot(feature.normalize(am), label="amp")
#plot(feature.normalize(fr), label="freq")
legend(framealpha=0.5);
show()
def align_peaks(sample, alignment_target, num_peaks, peak_r1=1, peak_r2=64):
#doing a very direct and coarse alignment:
#peak_s = signal.find_peaks_cwt(sample, np.arange(peak_r1,peak_r2))
peak_s = list(signal.argrelmax(sample, order=32)[0])
peak_s_vals = np.array([sample[x] for x in peak_s]).argsort()[-num_peaks:]
peak_s = np.sort([peak_s[x] for x in peak_s_vals])
#topnpeak_t = signal.find_peaks_cwt(alignment_target, np.arange(peak_r1,peak_r2))
topnpeak_t = list(signal.argrelmax(sample, order=32)[0])
peak_t_vals = np.array([alignment_target[x] for x in topnpeak_t]).argsort()[-num_peaks:]
topnpeak_t = np.sort([topnpeak_t[x] for x in peak_t_vals])
sample_partitions = np.split(sample, peak_s)
chunk_len = []
chunk_len.append(topnpeak_t[0])
for i in range(1, len(topnpeak_t)):
chunk_len.append(topnpeak_t[i]-topnpeak_t[i-1])
chunk_len.append(len(alignment_target)-topnpeak_t[-1])
new_samples = [interpolate_wave(seg[0], seg[1]) for seg in zip(sample_partitions, chunk_len)]
#return np.hstack(new_samples)
return normalize(syllable.moving_window(np.hstack(new_samples), 16))
