Hand Gesture Prediction using EMG Signals

Overview

This project implements a real-time hand gesture recognition model using 3-channel EMG signals. The model predicts four distinct gestures and maps them to keyboard controls for interactive applications such as gaming.

Recognized Gestures & Mapped Controls

• Bend Hand Left → Left Arrow (←)
• Bend Hand Right → Right Arrow (→)
• Flex Fingers → Up Arrow (↑)
• Pinch → Down Arrow (↓)

Device & Communication

These scripts are designed to work with a custom NPG (NeuroPlayground) device that streams real-time EMG data via WebSocket.

Installation

1. Clone the Repository

git clone https://github.com/CIumsy/Hand_Gesture_Prediction.git

2. Install Dependencies

Make sure you have Python installed. Then, navigate to the project directory in the command prompt and install the required libraries:

pip install -r requirements.txt

Running the Model

1. Start Gesture Prediction

Run the predict.py script to process real-time EMG signals and predict gestures:

python predict.py

2. Enable Keyboard Controls

Once predictions are running, execute game_controller.py to send keypresses based on detected gestures:

python game_controller.py

Usage

• Ensure that your NPG (NeuroPlayground) device is connected and streaming real-time data via WebSocket.
• Run predict.py first to generate predictions.
• Launch game_controller.py to map gestures to keyboard inputs.
• Use your hand movements to control applications that accept keyboard input.

Future Enhancements

• Expand Gesture Set: Add more complex gestures for enhanced control.
• Improve Model Accuracy: Fine-tune the model with additional data.
• Deploy on Embedded Devices: Optimize for low-power microcontrollers.

License

This project is open-source and available under the MIT License.