Machine learning ( RNN (Recurrent Neural Network) and SVM (Support Vector Machine) ) recognition/classification of 7 hand gestures using 6 channels of BioRadio 150 & BioCapture.
Note: This is for educational purposes, as part of the coursework for the Biorobotics & Cybernetics Course at RIT.
- Download the repo
- Open MatLab and change path to directory of repo
- Open DEMO.m and edit
DataDirandValidationDataDirto match the locations of the data on your computer - Run Options
UseValidationData: Set to 1 to train and test on the validation dataset. Set to 0 to use the collected EMG data.NeedDataForPlots: Set to 1 if training the SVM. Optionally set to 1 for training the RNN, if you want to view plots of the data.DemoCNN: Set to 1 to train a new RNN. Models with testing accuracies greater than 90% are automatically saved to the Models folder.DemoSVM: Set to 1 to train a new SVM. Models with testing accuracies greater than 90% are automatically saved to the Models folder.RunCount: How many times the model(s) should be trained before the overall confusion matrices are generated.TestRatio: What ratio of the data to reserve for testing.ShowModelDemo: Set to 1 to load an SVM trained model / RNN trained model. You will need to modify the paths in the DEMO.m file.BuildModels: Set to 1 to build new models. Set to 0 to skip building of new models.