4D vision - enabling computers to detect human emotions
The Centre for Machine Vision (CMV) is taking a leading role in developing human-computer interaction technology for application in the real world.
Detecting illness or security threats
Students and staff at UWE Bristol use machine vision to tackle a huge range of challenges. It can measure the growth of skin cancers, find concealed weapons and enable computers to detect changes in human emotions.
The photometric stereo technology used in machine vision creates 3D and 4D images taken with light emitted from several different sources simultaneously in order to create deeper insights.
UWE’s Centre for Machine Vision, part of the Bristol Robotics Laboratory, is one of only a handful of centres in the world dedicated to exploring the potential of machine vision. It has received major funding from the Technology Strategy Board, the EPSRC, Medical Research Council and the Ministry of Defence.
Speeding up automatic face recognition
The Centre has invented a Photoface device which captures a 3D image of their face as they walk through it. The project has built a database of over 450 faces captured in 3,000 sessions using photometric stereo technology.
Research associate Dr Mark Hansen said, “This database allows us to experiment with reducing the amount of measurements a computer has to make to recognise a person.
"Currently automatic face recognition is a slow and resource-hungry process, but we know that humans are very good at recognising faces from just a few details, such as from low resolution images and caricatures. Using novel methods inspired by psychology, we can achieve recognition rates of 95.75% at far greater speeds.”
What lies beneath
This type of automatic face recognition could be used to identify people allowed to access secure areas and is far less intrusive than other biometric techniques. Seeing what lies beneath the surface using photometric stereo can also detect hidden weapons or recognise people who have tried to disguise their features.
Watch the short film below to get a glimpse of the projects currently underway at the Centre for Machine Vision.
Next generation photometric technology
The Centre’s 4D Vision project is taking existing 3D photometric stereo technology to the next level. New feature extraction and classification methods enable fast detection of tiny micro-movements from moving 3D faces.
Visions of health
Work has begun on a project with the University of Bristol’s School of Experimental Psychology on detecting underlying changes in mood. For example, people suffering from depression are thought to display certain facial characteristics that can be measured. This could also have major benefits on caring for the elderly or ill in their homes.
Outdoor advertising - but not as we know it
Already, adverts relevant to our interests pop up on our phones or computer screens. Soon, targeted outdoor advertising will also be available.
Digital advertising panels with photometric stereo equipment embedded are being tested by the Centre and Aralia Systems Ltd. The panels can detect generic information about the people looking at the adverts, such as age, gender and the length of time they look at the ad.
This information is not stored, but used to change the display to show ads relevant to the demographics of the people viewing it. A family group could be shown information about local attractions for days out, while a group of young people would see an advert for a nightclub.
At the heart of human-computer interaction
Professor Melvyn Smith, Director of CMV, said, “Automatic face recognition is a very active area of research for good reasons. Interacting through a touch screen, voice or keyboard is not the only way for humans to communicate with computers.
"Research in UWE’s Centre for Machine Vision is at the heart of how technology of the future will find out what people want and how to interact with them."
Humans are capable of recognising people from low resolution images and caricatures. How long before computers catch up?
Did you know?
- The 3D Photoface database is one of the largest in the world.
- Automatic face recognition is promising research area due to its non-intrusive nature.
- Machines consistently outperform humans in many face recognition experiments.