Technology

Facial Expression Recognition

Global market for sentiment and emotion analysis is estimated to reach $3000 million in revenue by 2025, and we’re ready for it. Unlike our competitors, our technology is usable in real world environments, such as analyzing candidates in video interviews to measuring customer satisfaction in your stores. More than that, we offer live streaming as well as support for any camera types.

Try the tech
Markets

Real-world application

Face detection in non-frontal angles and suboptimal lighting

Live streaming

The only real-time analysis solution in the market

Group recognition

Simultaneous analysis of multiple faces or crowds

All camera types

Including webcams, mobile and security cameras

Seven basic emotions

Happiness, sadness, fear, anger, surprise, disgust and boredom

Academic foundation

Build on Paul Ekman’s emotion classification and the Big Five model

1

Face preprocessing

– Face features and eyes detection with Dlib HAAR cascade detector
– Face cropping

2

Feature extraction

– Face alignment, rotation and scaling
– Lighting normalization
– Normalized image is fed to Local Binary Pattern histogram filter

3

Emotion classification

– Vast collection of facial samples
– Classification using one-vs-one array of SVMs, trained with our dataset
– Output summed with weights from academic correlations (’NEOAC’)