• PARTS is a pattern recognition toolkit written in python, to quickly evaluate pattern recognition accuracies.
• It is based primarily on the 1DCNN. Something I have immense faith on, when it comes to time-series pattern classification.
• To carry out the evaluation, a series of steps are required: data preparation, training, testing.
• In cases involving feature extraction (instead of raw data), it is carried out live during the training phase, for each batch.
• An example use case can be found in example.ipynb

If you use this code for your work, please consider citing the corresponding paper(s):

1. Quartered Spectral Envelope and 1D-CNN-Based Classification of Normally Phonated and Whispered Speech
2. Literary and Colloquial Tamil Dialect Identification

• Example python notebook.
• Include a basic LSTM architecture.
• Data preparation classes for:
  • Whisper-normal speech classification using W-TIMIT.
  • Literary-colloquial Tamil speech classification using the Microsoft Dataset.
  • Sleep detection.