International Partner Search

Innovation & Technology Offer

New method for personalized mood recognition through video analysis of facial expressions

Country of Origin: Spain
Reference Number: TOES20170606001
Publication Date: 7 June 2017

Summary

A research group in mathematics and the computer science department of a Spanish university have developed a new method capable of recognizing the mood of a person through the video analysis of facial expressions. 

The research group is looking for large companies or SMEs in video game or health sectors that want to carry out a license agreement to exploit their technology.

Description

Facial expression recognition has been widely applied in the field of psychology, video games, health, learning and man-machine interactions in general, and is a very active research area. 



The concepts of "mood" and "emotion" are often confused in colloquial language and its formal definitions. There is a consensus that marks at least two major differences between these concepts:



• Moods have a longer duration than emotions.

• Moods are related to emotions because a person who is in a certain mood is prone to experience emotions in its facial expression.



There are currently technologies for analyzing emotions, but none is known to be able to analyze mood. 



A research group of mathematics and computer Science department of Spanish university have developed a new method capable of recognizing the person’s mood by analyzing video sequences that capture the subject facial expressions. In addition, the method is personalized for each subject by means of a short learning process. The method is based on the dynamic evaluation of the contribution of the recognized Action Units (AUs) to the mood.



The process for conducting this evaluation is as follows:



1.  Definition of data and general criteria: AUs are defined as sufficient to describe and recognize the mood of the subject: euphoria, anxiety, boredom, docility, hostility, relaxation, dependence and condescension.

2.  Definition of custom patterns of facial expressions: a personalized criterion of activation of each AU that is used to determine if a gesture or AU has been made by the subject

3.  Mood evaluation: of the AUs recognized, the subject's mood is estimated in a predefined time window.



This invention could be implemented in different computer-based systems or in mobile devices and, it could perform in near real-time. Currently, the group has developed the method in C++ standard libraries under the Linux OS that works using standard videos such as input.  



This technology can be useful for companies in health sector and for companies dedicated to video games. The research group is looking for large companies or SMEs in these sectors interested in carrying out an agreement to exploit their patent.

Advantages and Innovations

Currently there are tools that can analyze people's emotions, but they are not able to analyze and evaluate a person's mood.



The existing tools use procedures that focus on the recognition and processing of snapshots, while this tool is based on video sequences, which allows to dynamically evaluate the contribution of the AUs to the mood.



Another advantage and innovation is that the system offers a personalization of the subject to minimize errors of detection of AUs, for this reason, the method of recognition is more precise.



Existing methods are restricted to the identification of emotions (happiness, sadness, etc.) but do not allow the detection of complex constructs such as moods, the activation of which can at the same time comprise different configurations of emotions, sometimes even opposing ones (for example, anxiety can occur in a sad or in happy person). This invention solves this problem, giving way to a much more precise analysis and recognition.

Stage Of Development

Already on the market

Requested partner

Large companies and SMEs are sought to carry out a license agreement. 



It is interesting for the research group to find large companies or SMEs that are dedicated to video games or the health sector interested in exploiting the patent about facial recognition.