The Principal (Acting)
23 March 2009
Dear Mr Baker
Facial recognition biometrics
According to the Cambridge News , Ely are testing the use of facial recognition technology to determine which pupils are on the premises.
The business case in favour of using biometrics depends on the reliability of that technology.
We all have confidence in the expert use of fingerprinting by the police, and in the forensic use of DNA to identify people. Quite rightly. Traditional fingerprinting and DNA work. But that is not the technology being tested at St Neots. Biometrics based on facial recognition have a long history of not working .
Historically, you should expect this trial to fail – a subject currently being pursued with the Home Office Scientific Development Branch . Typically, about 30% of people cannot be recognised by systems using facial geometry. That is on the same day that they are registered/enrolled into the system. Two months later, typically, facial geometry systems fail to recognise about 60% of people . It is important to run your trial for at least three months.
As an early adopter of the technology in the education sector, Ely may be offered reduced rates for acquisition, installation, training and support. But if the equipment fails to recognise 60% of students after two months, then the business case fails at any price.
Ely is engaged in a technology trial. The technology may or may not be cheaper than the daily 90 minutes of a teacher's time that is expected to be saved. It may or may not make people safer in case of a fire. These matters cannot be determined in advance. They cannot be prejudged. The school needs to conduct the trial first to find out.
You are quoted as
saying: "As for the students, they
love the idea of taking responsibility for their own registration
Please be sceptical, Mr Baker.
 Jordan Day, 5 March 2009, Cambridge News, 'Students register delight at new face-fit check-in', http://www.cambridge-news.co.uk/cn_news_home/displayarticle.asp?id=396794
 http://dematerialisedid.com/PDFs/feasibility_study031111_v2.pdf, see para.52(c)