class Sound(BaseModel):
"""
An audio processing pipeline that computes a frequency domain representation
of the sound that follows a geometric scale
"""
bark = zounds.ArrayWithUnitsFeature(zounds.BarkBands,
samplerate=samplerate,
stop_freq_hz=samplerate.nyquist,
needs=BaseModel.fft,
store=True)
long_windowed = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
wscheme=zounds.SampleRate(frequency=zounds.Milliseconds(358),
duration=zounds.Milliseconds(716)),
wfunc=zounds.OggVorbisWindowingFunc(),
needs=BaseModel.resampled,
store=True)
long_fft = zounds.ArrayWithUnitsFeature(zounds.FFT,
needs=long_windowed,
store=True)
freq_adaptive = zounds.FrequencyAdaptiveFeature(
zounds.FrequencyAdaptiveTransform,
transform=np.fft.irfft,
scale=scale,
window_func=np.hanning,
needs=long_fft,
store=False)
rasterized = zounds.ArrayWithUnitsFeature(lambda fa: fa.rasterize(64),
needs=freq_adaptive,
store=False)
class SoundWithNoSettings(BaseModel):
short_windowed = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
wscheme=windowing_scheme,
wfunc=zounds.OggVorbisWindowingFunc(),
needs=BaseModel.resampled)
fft = zounds.ArrayWithUnitsFeature(
zounds.FFT,
needs=short_windowed)
geom = zounds.ArrayWithUnitsFeature(
spectrogram,
needs=fft,
store=True)
log_spectrogram = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
wscheme=zounds.SampleRate(
frequency=windowing_scheme.frequency * (spectrogram_duration // 2),
duration=windowing_scheme.frequency * spectrogram_duration * 3),
needs=geom)
ls = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
wscheme=spectrogram_sample_rate,
needs=geom)
class Document(BaseModel):
bark = zounds.ArrayWithUnitsFeature(zounds.BarkBands,
samplerate=samplerate,
stop_freq_hz=samplerate.nyquist,
needs=BaseModel.fft,
store=True)
long_windowed = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
wscheme=zounds.SampleRate(frequency=zounds.Milliseconds(500),
duration=zounds.Seconds(1)),
wfunc=windowing_func,
needs=BaseModel.resampled,
store=True)
dct = zounds.ArrayWithUnitsFeature(zounds.DCT,
scale_always_even=True,
needs=long_windowed,
store=True)
mdct = zounds.FrequencyAdaptiveFeature(zounds.FrequencyAdaptiveTransform,
transform=scipy.fftpack.idct,
scale=scale,
needs=dct,
store=True)
class Sound(BaseModel):
bark = zounds.ArrayWithUnitsFeature(zounds.BarkBands,
needs=BaseModel.fft,
store=True)
chroma = zounds.ArrayWithUnitsFeature(zounds.Chroma,
needs=BaseModel.fft,
store=True)
class Sound(BaseModel):
windowed = zounds.ArrayWithUnitsFeature(zounds.SlidingWindow,
wscheme=wscheme,
needs=BaseModel.resampled)
mu_law = zounds.ArrayWithUnitsFeature(zounds.mu_law, needs=windowed)
categorical = zounds.ArrayWithUnitsFeature(categorical, needs=windowed)
class WithTimbre(STFT, Settings):
bark = zounds.ArrayWithUnitsFeature(
zounds.BarkBands,
needs=STFT.fft,
store=True)
bfcc = zounds.ArrayWithUnitsFeature(
zounds.BFCC,
needs=bark,
store=True)
class Snd(Sound):
embedding = zounds.ArrayWithUnitsFeature(
zounds.Learned,
learned=learned,
pipeline_func=lambda x: x.pipeline[:-1],
needs=Sound.ls)
hashed = zounds.ArrayWithUnitsFeature(zounds.Learned,
learned=learned,
needs=Sound.ls,
store=True)
class Sound(BaseModel):
windowed = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
wscheme=zounds.SampleRate(frequency=samplerate.frequency *
(SAMPLE_SIZE // 2),
duration=samplerate.frequency * SAMPLE_SIZE),
needs=BaseModel.resampled,
store=False)
perceptual = zounds.ArrayWithUnitsFeature(perceptual, needs=windowed)
decomposed = zounds.ArrayWithUnitsFeature(
lambda x: FrequencyDecomposition(x, bands).as_frequency_adaptive(),
needs=windowed)
class Sound(Resampled):
"""
A simple pipeline that computes a perceptually weighted modified discrete
cosine transform, and "persists" feature data in an in-memory store.
"""
windowed = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
needs=Resampled.resampled,
wscheme=zounds.HalfLapped(),
wfunc=zounds.OggVorbisWindowingFunc(),
store=True)
mdct = zounds.ArrayWithUnitsFeature(zounds.MDCT, needs=windowed)
weighted = zounds.ArrayWithUnitsFeature(lambda x: x * zounds.AWeighting(),
needs=mdct)
class Document(STFT, Settings):
"""
Inherit from a basic processing graph, and add a Modified Discrete Cosine
Transform feature
"""
mdct = zounds.ArrayWithUnitsFeature(
zounds.MDCT,
needs=STFT.windowed,
store=True)
class WithCodes(WithTimbre):
bfcc_kmeans = zounds.ArrayWithUnitsFeature(
zounds.Learned,
learned=BfccKmeans(),
needs=WithTimbre.bfcc,
store=True)
sliding_bfcc_kmeans = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
needs=bfcc_kmeans,
wscheme=windowing * zounds.Stride(frequency=30, duration=30),
store=False)
bfcc_kmeans_pooled = zounds.ArrayWithUnitsFeature(
zounds.Max,
needs=sliding_bfcc_kmeans,
axis=1,
store=True)
class WithCodes(WithOnsets):
bark_kmeans = zounds.ArrayWithUnitsFeature(
zounds.Learned,
# this feature will be computed using the learned K-Means clusters
learned=BarkKmeans(),
needs=WithOnsets.bark,
store=True)
pooled = zounds.VariableRateTimeSeriesFeature(
zounds.Pooled,
needs=(bark_kmeans, WithOnsets.slices),
op=np.max,
axis=0,
store=True)
class WithOnsets(STFT, Settings):
bark = zounds.ArrayWithUnitsFeature(
zounds.BarkBands,
needs=STFT.fft,
store=True)
transience = zounds.ArrayWithUnitsFeature(
zounds.MeasureOfTransience,
needs=STFT.fft,
store=True)
sliding_detection = zounds.ArrayWithUnitsFeature(
zounds.SlidingWindow,
needs=transience,
wscheme=windowing * zounds.Stride(frequency=1, duration=11),
padwith=5,
store=False)
slices = zounds.TimeSliceFeature(
zounds.MovingAveragePeakPicker,
needs=sliding_detection,
aggregate=np.median,
store=True)
class Sound(SoundWithNoSettings, ModelSettings):
hashed = zounds.ArrayWithUnitsFeature(IdentityNode,
needs=SoundWithNoSettings.ls,
store=True)
class Sound(BaseModel):
fake_hash = zounds.ArrayWithUnitsFeature(produce_fake_hash,
needs=BaseModel.fft,
store=True)
class Sound(BaseModel):
chroma = zounds.ArrayWithUnitsFeature(
zounds.Chroma,
frequency_band=band,
window=window,
needs=BaseModel.fft)
