Speech Recognition System for Keyword Identification

– Implemented a Classification Model using 2D CNN (3 layers) on Audio Dataset for Keyword Identification.
– Extracted 13 MFCCs for each audio after truncating each audio to 22050 samples (1 sec).
– Managed to get training and testing accuracy of 95.66% and 95.14% after running for 40 epochs.
– Deployed the model as a Flask API and created a demo web-page for testing audios in front-end.