Smart computers recognize human characteristics

Today's computers can perform countless calculations, but follow a cold, emotionless manner. However, computer engineers are changing that as they try to create human characteristics for computers to deliver

Today's computers can perform countless calculations, but follow a cold, emotionless manner. However, computer engineers are changing that as they try to create human characteristics for computers for human-machine communication to become more friendly and natural.

By combining audio and video data, Yongjin Wang, University of Toronto and Ling Guan, Ryerson University, Toronto has developed a system that identifies 6 types of human emotional states: happiness, sadness, anger data, fear, surprise and disgust. Their system can identify the emotions of people from different cultures and use different languages ​​with a success rate of 82%.

Wang said: 'Human-centered computers focus on understanding people, including facial expressions, emotions, gestures, speech, body movements, etc. Emotional recognition systems help computers understand the emotional state of users and from there computers can react based on that awareness. '

The researchers' system extracts a large number of voice characteristics, such as ' rhyming characteristics'.

As expressed by Wang and Guan, expressing emotions is very diverse: voice features and facial expressions can play an important role in specifying certain emotions, but not significant for feelings. Other contact. A common example such as happiness is better discovered thanks to some visual characteristics (smiling), while anger is due to auditory characteristics (screaming). Researchers found that not just one trait has an important role in all six emotions. While this finding shows that there are no clear boundaries between types of emotions, it also makes it difficult to distinguish the types of emotions of a computer.

Picture 1 of Smart computers recognize human characteristics

Researchers are developing a human-centered computer system that can recognize features such as the emotional state and age of people.(Photo: Guillaume Duchenne)

To overcome this problem, Wang and Guan used a step-by-step method. The method only adds one trait at a time and then gradually moves away to find related traits to recognize the emotional state. They then used a multi-classification system such as 'separation and resolution' to separate possible emotions based on a combination of facial features and voice.

'The hardest part of helping a computer detect human emotions is a huge difference and diversity of speech expression and facial expressions due to various factors such as language, culture and personal character. , etc. In addition, according to our study, there is no clear separation between different emotions. Identifying exactly different patterns is a challenging issue. '

Researchers believe that human emotion-sensing computers could one day be used in customer service, computer games, security monitoring and educational software.

Estimated age

While most people are quite good at distinguishing someone who is sad or happy, determining the age of others just by looking at them is a much bigger challenge, at least for humans. Computer engineers Yun Fu and Thomas Huang, of the University of Illinois, created a computer system that judged a person's age based on their facial features.

'Life expectancy is one of the most important characteristics that deduces personal circumstances, anthropometric information and social circumstances. People-oriented computer systems can capture more information of people in social groups than normal computer systems, by focusing on how people organize and improve their lives around computer technology.'

Fu and Huang based their research on face image data of more than 1600 subjects, half of them women, and the rest of men. Using the results from the previous age judgment system, as well as the age characteristics they identified, the researchers trained a computer system that was able to estimate age between 0 and 93 years old. Their most successful algorithm can predict the age of people - based on only a few images - within 5 years or less than 50% of the time, within 10 years or less 83% of the time.

Researchers predict that age-aware math can be used in preventing children from accessing adult websites, blocking vending machines that sell alcohol to people who are not old enough, and determining their age group. those who spend more time watching certain ads.

By integrating human factors into computers, researchers will continue to bridge the gap between technology and people. No doubt, these technologies will have a lot of applications in the future.

Update 14 December 2018
« PREV
NEXT »
Category

Technology

Life

Discover science

Medicine - Health

Event

Entertainment