Gesture Recognition

For this project, we were given IMU sensor readings from a cell phone, where the user was instructed to perform six different gestures, such as waving the phone in a circle, figure-8, etc. We were tasked to create a model that would be able to discern between the six different gestures, even if the gestures were made faster or slower than the training data.

To do this, I wrote a Hidden Markov Model that would characterize certain “states” by their profile using the IMU. The model would then recognize different sequences of certain states, and learn to associate them with their respective gestures.

When running the model on the testing data that was given to us, my model performed with 100% accuracy.