Skip to content

Latest commit

 

History

History
44 lines (27 loc) · 1.64 KB

README.md

File metadata and controls

44 lines (27 loc) · 1.64 KB

Features Extraction

Speech Features

Speech features are extracted from the Signal with sample_rate, frame_ms, stride_ms and num_feature_bins.

Speech features has the shape (B, T, num_feature_bins, num_channels) and it contains from 1-4 channels:

  1. Spectrogram, Log Mel Spectrogram, Log Gammatone Spectrogram or MFCCs
  2. Delta features: librosa.feature.delta from the features extracted on channel 1.
  3. Delta deltas features: librosa.feature.delta with order=2 from the features extracted on channel 1.
  4. Pitch features: librosa.core.piptrack from the signal

There are 2 classes for Speech Features Extraction: NumpySpeechFeaturizer (uses librosa) and TFSpeechFeaturizer (uses tf.signal). The tf based class does not support delta, delta_delta, pitch features yet.

Note: TFSpeechFeaturizer class should be used if you want to deploy tflite.

Spectrogram

Log Mel Spectrogram

MFCCs

Log Gammatone Spectrogram

Text Features

Text features are read as index from the file like the default tensorflow_asr.featurizers.english.txt plus 1 for the blank index.

The blank index is either 0 or num_classes - 1 where num_classes is number of characters in your language (exclude blank).

Class TextFeaturizer is initialized with the decoder_config as follows:

decoder_config:
    vocabulary: path_to_vocab_txt
    blank_at_zero: bool
    beam_width: int,
    lm_config: ...

TODO

  • Implement TFSpeechFeaturizer to extract delta, delta_delta, pitch features