FlexEMG Project: A wearable biosensing system with in-sensor adaptive machine learning for hand gesture recognition

Code and data repository for FlexEMG 2020 Nature Electronics publication. It is publicly available under GNU General Public License v3.

System requirements

Data collection:
- Graphical user interface: Python 3.4
- Device firmware: Libero SoC v11.7
Data analysis:
- Matlab R2019b
- Python 3.7 (package dependencies can be found in data_analysis/flexemg.yml

Repo structure

Offline dataset for model validation

Offline data collected from 5 subjects can be found in the submodule 'offline_data'.

Data collection code

Data collection code can be found in the subdirectory 'data_collection'. This directory includes the graphical user iterface (developed in PyQt) under the 'gui' subdirectory, as well as the firmware (FPGA design and ARM Cortex M3 code) located in 'FlexEMG_21Class_HDC'.

Data analysis

Data collected from real-time experiments located in 'data_analysis/realtime_data' directory as .mat files. Offline data encoded using simulation of hardware HD proection algorithm located in 'data_analysis/hw_encode'.

Main analysis files include:

'compare_snr.m' and 'compare_spectrum.m' for analysis of sEMG signal quality (Fig. 2)
'get_pca_projection.ipynb', 'plot_pca_dim.m', 'plot_pca.m' for analysis of hypervector dimensionality (Fig. 3)
'realtime_confusion.m', 'realtime_examples.m', 'realtime_transitions.m' for analysis of real-time experiment data and classification accuracy (Figs. 5, 6)