Embedded Files
HHDOW_Software&BIO_DHSFEmotionAIEngineers
School: H.H. Dow High School, Midland, MI
Project: Psychological Evaluation Using AI
Teachers: Lisa S. Tsay & Margaret E. Hitt
Team members: Sanvi Patel, Emma Huang & Katie Cai
Emotion Detection AI/ML Model
Project Description
Astronauts are at risk of feeling isolation, circadian rhythm disturbance & mental health issues. Psychological consulting on well-being, diet, and nutrients might become unavailable for space missions beyond Low Earth Orbit. Recognition of emotions is critical to help astronauts maintain mental health and prevent health issues.
Our Proposal: Develop a trained AI model utilizing facial and vocal recognition to help monitor stress and emotions to maintain mental health and well-being.
Ai MODel Architecture
Ai MODel Architecture
We utilized 90+ datasets from the University of Southern California to train the AI Model. We also used ISS YouTube downlinks to help train the model with real astronauts' voices and facial expressions. We used Python for the AI Model, and utilized many libraries, as well as Logistic Regression and Random Forest Classification to train the AI Model.
Vocal Recognition - Results
Model Correlation & Performance Analysis
Model Correlation & Performance Analysis
For the vocal recognition, the model was trained separately on male and female voices. The heatmap demonstrates the differences in the correlations between each of the features. From there, our group focused on improving our accuracy, eventually reaching 83% accuracy.
Facial Recognition Results
Performance Analysis
Performance Analysis
The model was trained using DeepFace and OpenCV. The model portrayed accurate results on astronaut faces, and can give real-time emotion as well as emotion distribution.
Testing Videos
Facial Recognition
Voice Recognition
Page updated
Google Sites
Report abuse