Detecting Emotions by Computer
November 2, 2015
Polish scientists have created a new computer system that can recognize emotions; it can tell if the user in front of the computer screen is joyful, sad or disgusted.
The technology, known as Xpress Engine, uses computer vision and artificial intelligence to analyze standard video output with a human face. From that information, it can extract such features as emotions—happiness, surprise, sadness, anger, or disgust—along with gender and age group. Created by software engineers Krzysztof Cywiński and Dariusz Kamowski, with the assistance of psychologist Konrad Bocian, the system makes it possible to monitor changes in the intensity of human emotions from one second to the next. Cywiński, Kamowski and Bocian run a research and development company called Quantum Lab Poland.
Users can employ Xpress Engine as an embedded system in their own devices or use Quantum Lab Poland’s application program interface (API) service to send video and receive data in real time via the internet. API is a set of routines, protocols and tools for building software applications.
Xpress Engine technology is designed to appeal to users in market and scientific research, robotics, medicine, retail, security, outdoor marketing and gaming. Its prime targets include market research companies, marketing agencies and advertising companies. Such companies usually use eye-tracking technology to measure consumer engagement. The method developed by Quantum Lab Poland is more efficient—and cheaper. The fact that someone is looking at an advertisement—as evaluated by eye-tracking systems—does not necessarily mean that the consumer is interested in it.
The Xpress Engine system comes with a “research platform” called Ellen. Those wanting to use the platform need to have internet access and a basic camera, for example in a laptop; they are asked to upload material that they want to test, such as a completed commercial, says Quantum Lab Poland’s Bocian.
The system will soon be made available for use by paralyzed people. “Soon we will be able to expand the tool so that it can be used in areas such as medicine, psychotherapy and social robotics,” says Bocian. “The aim is to build personal robots capable of analyzing people’s emotional state.”
Quantum Lab Poland is the only Polish company to have developed and gone commercial with this kind of technology. Internationally two other companies, both associated with the Massachusetts Institute of Technology, offer similar technology, according to Bocian. “We took two years to develop the same kind of product, but with much less funding,” he says.
How can a computer system be taught to recognize human emotions? First, it must be taught to detect faces using appropriate algorithms. The program then starts to analyze the changes that appear in the face, the muscle tone and facial expressions. On this basis, it tries to determine the probability of the face expressing a specific emotion.
To teach their computer system to recognize joy, for example, the scientists showed it a thousand images of smiling faces. Learning to detect a smile is about the easiest thing for the system to do; all that’s needed is to observe the muscles lifting the corners of the mouth and the muscles around the eye.
Quantum Lab Poland developed its Xpress Engine technology using a grant of 50,000 euros from the European Union’s Horizon 2020 program. Quantum Lab Poland is one of three Polish companies to have qualified for funding intended for small and medium-sized enterprises under the Horizon 2020 program.