Package `julius`

Julius contains different Digital Signal Processing algorithms implemented with TensorFlow, so that they are differentiable and available on GPU. Note that all the modules implemented here can be used inside a tf.function.

For now, I have implemented:

julius.resample: fast sinc resampling.
julius.fftconv: FFT based convolutions.
julius.lowpass: FIR low pass filter banks.
julius.filters: FIR high pass and band pass filters.
julius.bands: Decomposition of a waveform signal over mel-scale frequency bands.

Along that, you might found useful utilities in:

julius.core: DSP related functions.
julius.utils: Generic utilities.

Please checkout the Github repository for other informations. For a verification of the speed and correctness of Julius, check the benchmark module bench.

This package is named in this honor of Julius O. Smith, whose books and website were a gold mine of information for me to learn about DSP. Go checkout his website if you want to learn more about DSP.

Sub-modules

julius.bands: Decomposition of a signal over frequency bands in the waveform domain.
julius.core: Signal processing or TensorFlow related utilities.
julius.fftconv: Implementation of a FFT based 1D convolution in TensorFlow. While FFT is used in CUDNN for small kernel sizes, it is not the case for long ones, e.g …
julius.filters: FIR windowed sinc highpass and bandpass filters. Those are convenience wrappers around the filters defined in julius.lowpass.
julius.lowpass: FIR windowed sinc lowpass filters.
julius.resample: Differentiable, TensorFlow based resampling. Implementation of Julius O. Smith algorithm for resampling. See https://ccrma.stanford.edu/~jos/resample/ …
julius.utils: Non signal processing related utilities.