ROCK, PAPER, SCISSORS, spOcK!
Classifying Gestures by Reading Muscle Activity
Highlights and Results
The objective of this study is to apply the learnings of Machine Learning class to real world datasets like electrical muscle activity and explore the current trend of Artificial Intelligence in the field of Prosthetics.
Prosthetic implant is an artificial device that replaces a missing body part, which may be lost through trauma, disease, or congenital disorder. Prostheses intend to help people with disabilities to restore the normal functions of what they've lost. A way to improve these devices is through the help of an electromyography (EMG), an electrodiagnostic medicine technique for evaluating and recording the electrical activity produced by skeletal muscles, which then can be analyzed by a computer to detect the gesture intended to be made by the muscles.
In this study, Machine Learning algorithms have been used to 11,678 rows of muscle activity Kaggle data to create a model that can predict the classification of a gesture among 4 classes: rock, paper, scissors, or OK, with a test accuracy of 96.54%, which is higher than the methods used by others who have explored the dataset.
In addition, companies like Neuralink by Elon Musk, Kernel by Bryan Johnson, and CTRL-Labs by Thomas Reardon have done breakthroughs as they aim to read and write from the brain by tapping directly into the nervous system to decode individual neurons.
Methodology
The following steps were taken for this study:
- Data Description and Processing
- Exploratory Data Analysis
- Machine Learning Models
- Results and Conclusion
- Applications and Advancements