How rajeshraman91/deeplearning-model-for-audio-based-gesture-recognition-in-smartphones works
Project type and purpose
• This is a machine-learning training and inference pipeline for audio-based gesture recognition on smartphones.
• It uses two deep-learning model variants (CNN-GRU and LSTM) implemented in Keras/TensorFlow, with Jupyter notebooks and Python scripts for data parsing, preprocessing, model training, and saving the resulting HDF5 model files.
File structure
• Top level:
• README.md – project summary
• 2nd_phase/ – contains all code, notebooks, and model artifacts for phase two
• Inside 2nd_phase/:
• Input_Parsing.py – reads and organizes raw audio gesture data
• Input_Preprocessing.py – transforms parsed audio into model-ready features
• main_CNN_GRU.py & main_LSTM.py – training scripts for each model variant
• CNN_GRU_Keras.ipynb & keras_final.ipynb – interactive notebooks for exploratory modeling and final runs
• *.h5 files – serialized trained models (CNN_GRU.h5, LSTM.h5)
README
• Very concise: “Audio based Gesture recognition in smart phone by deep learning models – CNN-GRU and LSTM.”
Detail and essential architectural elements
• Script/notebook as its own node so reviewers see how responsibilities are divided.