gesture_classification

Classify 6 different hand gestures from EMG signals (8 channels) using Bidirectional-LSTM, achiving an accuracy of 91%.

https://docs.google.com/presentation/d/1S-tZIeXz5ZrqtYjcxkbFHGBd_ZCQBcK-lrb3u8WdbEw/edit?usp=sharing

https://colab.research.google.com/drive/1ilmsd0zbhoplAQs-3-lcVRK7BD2w266t?usp=sharing

https://docs.google.com/presentation/d/1odYnBfOFzMnyk0dq3HuuZ9Qx2JZ46el6a-IdfatW9ig/edit?usp=sharing

https://docs.google.com/document/d/1fqHOCs9yaafqJTCGch0EJN-a-lbHf-e6C8YGUgfd6uU/edit?usp=sharing

https://docs.google.com/document/d/1GbC57zn7GvMN1oU75FAP7olxzZ-ef7IPO0xspEOb8HI/edit?usp=sharing

https://docs.google.com/document/d/1VmB_VKGPw4U-Pls0y5d9pWIKKpnJDRvY4_jNekYAebk/edit?usp=sharing

https://docs.google.com/document/d/1ZZsI_LtQugMpogWA_1rXggQWapcsi9YxMs8zDaq7FXk/edit?usp=sharing

https://docs.google.com/document/d/1VmB_VKGPw4U-Pls0y5d9pWIKKpnJDRvY4_jNekYAebk/edit

Download Data

To download data required for training/testing, follow the instruction in README (the one in raw folder)

Before Starting

Delete README in raw folder and place holder in data folder, otherwise it will lead to error

To Start

Download requirements Follow documentation in main.py

Note

The model is already trained, and main.py is set up using trained model. (the version that trains is commented)

Evaluation Matrix

          precision    recall  f1-score   support

       1       0.99      0.98      0.99       367
       2       0.91      0.96      0.93       502
       3       0.90      0.86      0.88       523
       4       0.84      0.92      0.88       517
       5       0.90      0.84      0.87       527
       6       0.93      0.90      0.91       396

accuracy                           0.91      2832
macro avg      0.91      0.91      0.91      2832
weighted avg   0.91      0.91      0.91      2832