ML Model implementation for accelerometer and gyroscope values

I have a numeric valued dataset, and I need an ML Model implementation based on it. The dataset contains information from the accelerometer and gyroscope of an earable. I am working on a method to quantify the attention/focus of a peer in an online meeting using these accelerometer and gyroscope values collected from the ear. I need an ML model with an algorithmic novelty that can be integrated into a GUI-based system to classify the level of attention in real-time on a google meet screen. Would need more discussions and follow-ups for hyperparameter tuning and algorithm modification to achieve the desired optimum accuracy and precision.