Automatic Speech Recognition for Chhattisgarhi

• 5 words recognizer for Chhattisgarhi using MFCC and DTW


• 20 words recognizer for Chhattisgarhi using MFCC and DTW


• 58 words recognizer for Chhattisgarhi using MFCC and DTW


• 58 words recognizer for Chhattisgarhi using the following features:
  
  
  
  
  
  
  
  • MFCC
  
  
  • Mel Spectrogram
  
  
  • Chroma - stft
  
  
  • Tonnetz
  
  
  • Spectral - Contrast
  
  

  Which gives 193 features for each sample. We used 2088 samples for Training and 232 samples for testing out of 2320 total samples collected from 20 speakers, where each of the 58 words was recorded twice for each subject.
  
  (58 words 2 subjects 2 iterations= 2320 samples).
  

For more details, feel free to go through the report or the slides.

To read the report, click on the file:

“Automatic_Speech_Recognition_for_Chhattisgarhi.pdf”
present in the “Report_and_Slides” directory.

To view the presentation slides, click on the following link:

Slides