Conclusion.
This project aimed to use gesture recognition for controlling IoT networks to perform certain actions. In order to achieve this the GCAS system takes advantage of the simplicity and ease of the MQTT protocol for connectivity among different IoT networks, AWS S3 for data storage, and AWS SageMaker for their machine learning infrastructure. The project was able to successfully read raw sensor data from a wearable during a gesture, train and predict off these gestures, and push the prediction to connected IoT networks to perform a particular action. Aside from hardware connectivity presenting reliability issues in the prototype, the overall predictive performance is accurate and latency between the AWS infrastructure and the final performed action through MQTT is low. Further improvements on this project would include automatically re-training the gesture recognition model, redesigning the hardware for better reliability, and exploration into real-time streamed predictions as opposed to deterministic gesture recognition.