Pirouette Recognition LSTM

This started as a class project for Professor Phuong Truong's MAE8 course, but I ended up taking it considerably further than the assignment required. The goal was to classify pirouette movements from IMU time-series data.

I have a background in CS (high school, coding camps, etc) but had never touched MATLAB before this course, so there was a pretty steep learning curve on the tooling side. I used MATLAB's Deep Learning Toolbox to build the full training pipeline from scratch (data preprocessing, model architecture, training and validation) to be used on a multi-dancer dataset I found online. The model achieved over 95% classification accuracy. I documented my methodology and results in a  pseudo-technical paper under Professor Truong's guidance (linked below).

Skills

MATLAB · LSTM · Machine Learning · Signal Processing · IMU · Technical Writing