class Onsets(BaseModel):
onset_prep = ArrayWithUnitsFeature(SlidingWindow,
needs=fft_feature,
wscheme=HalfLapped() *
Stride(frequency=1, duration=3),
store=False)
complex_domain = ArrayWithUnitsFeature(ComplexDomain,
needs=onset_prep,
store=False)
sliding_detection = ArrayWithUnitsFeature(
SlidingWindow,
needs=complex_domain,
wscheme=HalfLapped() * Stride(frequency=1, duration=11),
padwith=5,
store=False)
slices = TimeSliceFeature(MovingAveragePeakPicker,
needs=sliding_detection,
aggregate=np.median,
store=True)
class WithOnsets(self.STFT):
transience = ArrayWithUnitsFeature(MeasureOfTransience,
needs=self.STFT.fft,
store=True)
sliding_detection = ArrayWithUnitsFeature(
SlidingWindow,
needs=transience,
wscheme=self.wscheme * Stride(frequency=1, duration=10),
padwith=5,
store=False)
slices = TimeSliceFeature(MovingAveragePeakPicker,
needs=sliding_detection,
aggregate=np.median,
store=True)
class Document(BaseModel):
raw = ByteStreamFeature(ByteStream,
chunksize=2 * 44100 * 30 * 2,
store=True)
ogg = OggVorbisFeature(OggVorbis, needs=raw, store=True)
pcm = AudioSamplesFeature(AudioStream, needs=raw, store=True)
resampled = AudioSamplesFeature(Resampler,
needs=pcm,
samplerate=samplerate,
store=True)
windowed = ArrayWithUnitsFeature(SlidingWindow,
needs=resampled,
wscheme=windowing_scheme,
wfunc=OggVorbisWindowingFunc(),
store=False)
fft = ArrayWithUnitsFeature(FFT, needs=windowed, store=True)
chroma = ArrayWithUnitsFeature(Chroma,
needs=fft,
frequency_band=band,
store=True)
bark = ArrayWithUnitsFeature(BarkBands,
needs=fft,
frequency_band=band,
store=True)
bfcc = ArrayWithUnitsFeature(BFCC, needs=bark, store=True)
bfcc_sliding_window = ArrayWithUnitsFeature(
SlidingWindow,
needs=bfcc,
wscheme=windowing_scheme * Stride(frequency=2, duration=4),
store=True)
bfcc_pooled = ArrayWithUnitsFeature(Max,
needs=bfcc_sliding_window,
axis=1,
store=True)