Shennong provides a wide range of speech features extraction algorithms as well as post-processing pipelines. It relies on Kaldi for most of the algorithms while providing simple to use Python API and command line interface.
-
See the complete documentation at https://docs.cognitive-ml.fr/shennong.
-
Attached paper available at https://arxiv.org/pdf/2112.05555.pdf.
-
See the detailed installation procedure here. On Linux, simply have a:
conda install -c coml -c conda-forge shennong -
Implementented models include filterbanks, MFCC, PLP, bottleneck, pitch, delta, CMVN, VAD, VTLN. See the complete list of available features here.