The Open Muscle project aims to revolutionize prosthetic technologies by harnessing the power of artificial intelligence (AI) to predict muscle contractions. Open Muscle is a machine learning prosthetic sensor suite that uses 12 pressure sensors on a bracelet for feature data and a LASK system comprised of piston springs for finger movement label data. The project’s goal is to provide free hardware and software for biometric machine learning datasets with feature and label data. ChatGPT, specifically the GPT-4 model, played a significant role in identifying the right library for the project, ultimately choosing Scikit-Learn based on its feedback.
The Open Muscle project aims to develop a low-cost sensor system for both forearm muscles and finger movements, evaluate machine learning models’ efficacy, and emphasize the importance of collecting and using biometric data in prosthetic technologies. The project faced challenges in acquiring clean and relevant data, and a more quantized approach was adopted. Machine learning, specifically artificial neural networks, were used to predict muscle contractions and improve prosthetic device functionality.
The Role of ChatGPT
ChatGPT, or GPT-4, assisted in identifying the right library and model for the project, providing feedback for advanced machine learning questions, and offering sample code based on project parameters. GPT-4 also played a vital role in troubleshooting and debugging code throughout the project, offering valuable insights and assistance as if an expert was present.
Data Collection & Model Training
The dataset used for Open Muscle had 12 sensors for the bracelet with timestamps and four labels from the LASK system with timestamps. These datasets were converted from JSON format to CSV format for model training. GPT-4 suggested three different training models, with the Random Forest Regressor ultimately being chosen. Challenges faced during the training process included noise in the data and limited training time.
Model Performance & Real-World Applications
The trained model had a Mean Average Error (MAE) value of approximately 25, equating to a 12% error rate. Despite the challenges, the model was able to make real-time predictions, demonstrating its potential for a significant impact on the prosthetic industry.
GitHub Link for Open Muscle: https://github.com/turfptax/openmuscle
Future Improvements & Call to Action
Next steps for the Open Muscle project include enhancing data quality, training the model with a larger dataset, improving hardware components, and creating a Python library for training and predicting sensor data. The project needs engineers, students, and enthusiasts to contribute to its development and spread the word.
The Open Muscle project, supported by AI and ChatGPT, aims to revolutionize prosthetic technologies by reducing sensor costs and enabling individuals to tackle projects that would typically require an entire team. The project’s success thus far is owed to the support and interest of various individuals and organizations, and its potential for creating a positive impact on the world is significant.