The shop that knows your face

By Hatice Kilicer Jul 15, 2010, 10:07 GMT

Berlin - A researcher in Karlsruhe, Germany is working on facial recognition software that could be used by retailers to make assumptions about shopping habits. There are many examples of how companies use technology to find out as much as possible about their customers.

Personalized internet advertisements, customer cards that store data about a shopper's buying habits and countless market surveys are just a few. They typically use the data they collect to place their products as precisely as possible in the market to capture their buyers' interest.

In a new twist on data collection used on shoppers, researchers in Karlsruhe have developed software that could be useful in department stores. The software can recognize faces and tell the gender and age of a customer. It also detects what a shopper looks at.

The sales staff of the department store could use the software to find out, for example, how long a woman spends in the shoe department and where she goes after that.

People who care about the protection of personal data will certainly cry foul, but the scientists working on the software are prepared to assure them that customers will remain anonymous.

The software would, for example, let the store staff know that customer number 300 bought shoes and then chocolate, said Rainer Stiefelhagen, one of the researchers working on the project. Airports could also use the software. They could use image- and video- analysis in access control and thereby improve security.

Stiefelhagen is among a group of researchers at the Institute for Anthropomatics at the University of Karlsruhe who are working in the area of artificial intelligence and the learning capacity of machines. He and eight colleagues are developing facial-recognition software that recognizes, finds and sorts faces.

The software works using a simple camera focused on a face's forehead, nose, eyes and chin. The image is compared with a database and searches for the one with the same characteristics. It then shows the result on a screen. The software recognizes not only the face, but it also shows where the person is looking. The demo computer recognizes all employees at the institute, showing not only the face, but also the person's name.

'We are working on teaching the machine to be more natural,' said Steifelhagen. It should be able to determine who is speaking and who is standing in front of it.

Facial recognition is a subject many companies are interested in. Millions are being spent on promoting it and there is also plenty of money going into research. The goal is not a machine that has just facial recognition abilities, but also the ability to register movement.

'The machine should recognize not only that I am moving my hand, but that I am reaching for a particular bottle,' said Stiefelhagen.