Those of us who have used the face unlock system in smartphones have many times faced the problem where the system simply can't recognise the owner's photo. Though the face unlock system is present in Samsung Galaxy phones and it is really good at keeping most impostors at bay, when a team of researchers at University of York examined the performance of the feature, they found that it failed to recognise genuine owner of the phone. Dr. David Robertson of the Department of Psychology’s FaceVar lab from the University of York led a research team to find out a solution to this problem. They found that, instead of a single target image, if the system used a combination of different pictures of the user, the smartphone's face recognition is significantly improved.
The team published their research work in the journal PLOS ONE and described how the face recognition systems in our gadgets can be improved to a great extent if users stored an ‘average’ of their own photos – formed by morphing together several different photos. The concept behind this is that the automatic face recognition systems benefit from incorporating the advantages of familiarity. In their experiments, they found that users could unlock their phones more reliably when the device stored an average of the user’s face than when they stored a single image. Not only did this result in consistency across a wide variety of everyday viewing in various lighting conditions, but it also did not affect the rejection of imposter faces.
Smartphone-face-recognition-improved-university-york-research. Left: In a normal ‘face-unlock’ system, smartphone stores a single image of the owner. Right: Owner’s face ‘average’ – an image derived from many photos of a person – being encoded into the system. When it comes to recognising faces in a computational system, it remains a significant challenge for the cognitive sciences. Humans have been recognising faces using a variety of factors such as eye gaze, facial expression and attractiveness. Not many processes are available for processing the person’s identity. Depending upon the camera's specifications, lighting of the area, age, pose etc., a person's face varies considerably and therefore it is possible that two images of the same person could be less similar than two images of different people.
This where the proposition from York researchers comes into picture. They want to tackle this problem of varying face structure of the same person by storing many different images of the same person in memory and then using an 'average' of those photos to be used when recognisng it for letting access to a smartphone or tablet. What are your thoughts on the proposed system for improving face recognition? Have you tried our the face unlock feature available in Samsung Galaxy Android smartphones?
Share your thoughts with us in comments below.
0 Comments:
Post a Comment