/MelgramCNN

Convolutional Neural Network(CNN) feature extraction from Mel-Spectrograms to predict laboratory induced faults in a signal.

Primary LanguagePython

MelgramCNN

MelgrammCNN is a Convolutional Neural Network used for a regression task. Features are extracted from MelSpectrograms to predict the remaining_time_left given acoustic signals emmited by a laboratory fault. This process is as follows:

(1) Generate MelSpectrograms given acoustic data

(2) Train a CNN on Generated MelSpectrograms.

(3) Predict remaining_time_left given a 150,000 size sample of acoustic data.

(data not included) This was an effort towards the LANL Kaggle Contest

Authors

  • William Mendoza Gopar
  • Brian Whiteaker
  • Premanand Kumar