Russia Develops a Next-Generation AI System for Emotion Recognition
The software suite can identify seven basic human emotional responses with high accuracy.
Russian researchers are advancing neural-network-based emotion recognition technologies. A system made up of several software modules has already learned to identify human reactions with a high degree of accuracy. Alexei Karpov, the project’s lead at the St. Petersburg Federal Research Center of the Russian Academy of Sciences, described the MASAI system in comments to the Nauchnaya Rossiya portal.
According to Karpov, the team is developing what he calls emotional artificial intelligence — technology that enables machines to mimic human communication. The system analyzes multiple types of data, including speech, voice, facial expressions, and behavior. Each data stream is handled by its own neural network built on transformer principles, but optimized for a specific task.
Multiple Methods, One System
The MASAI system — short for Intelligent System for Multimodal Affective States Analysis — can recognize joy, sadness, surprise, calmness, anger, fear, and disgust. It can also capture a speaker’s attitude and identify acute emotional states.
Karpov said the technology is expected to be used in the development of empathetic voice assistants and in contact centers. MASAI can detect emotions without physical contact, using video or audio data. In the future, it is also expected to identify emotions based on heart rate data from electrocardiograms.
