Back

ML Prediction of Emotions using Biomechanics

Research Project

In addition to performing statistical analyses on the data, we also apply machine learning techniques to be able to predict emotions from biomechanical characteristics. 

First, we applied ML techniques to walking trails in young adults to predict four classes of emotions using spatiotemporal and joint range of motion variables. The ML models were trained using Python using a workflow that utilizes leave-one-participant-out cross validation to ensure the generalizability of the prediction capabilities. The Machine Leaning models used were Multilayer Perceptron, Random Forest, K-nearest Neighbors, and Logistic Regression. The ML models accurately predicted the emotion of a given trial at rates above random chance. I have presented this work at the South-Central America Society of Biomechanics annual conference.

Currently, we are working on predicting emotions from sit-to-walk biomechanics in both young and older adults.