Real-time emotion recognition for gaming using deep convolutional network features

TL;DR

This paper demonstrates that deep convolutional network features can effectively recognize emotions from single images with high accuracy, enabling real-time affective feedback in gaming applications.

Contribution

It introduces a novel approach using deep convolutional features for emotion recognition from still images and applies it to real-time gaming feedback.

Findings

Achieved 94.4% recognition accuracy

Performed emotion recognition using single still images

Implemented real-time facial expression tracking in a game

Abstract

The goal of the present study is to explore the application of deep convolutional network features to emotion recognition. Results indicate that they perform similarly to other published models at a best recognition rate of 94.4%, and do so with a single still image rather than a video stream. An implementation of an affective feedback game is also described, where a classifier using these features tracks the facial expressions of a player in real-time.