Natural Language DNN built on top of Keras (Tensorflow backend) to detect trigger words in real time audio inputs. The model is designed to take input of 10 second audio clips and predict whether trigger words are present.
- trigger_word_engine.ipynb: jupyter notebook for overall model training, testing and predictions
- twd.py: python code for designing and training the model
- twd_prediction.py: python code for predicting with the trained model
Adapt the model for use in real-time applications.