British researchers are developing a voice recognition system that can identify a speaker if he or she is drunk or even has a cold. This could help reduce fraud involving cash machines and credit cards, which British banks say costs them over f,150 million annually.

The new system, called Time-Encoded Signal Processing or Recognition (Tespar), will record a customer's standard phrase - such as cash voice identification - which will be encoded onto their card. To withdraw cash, they will have to repeat the phrase into an adapted cash point machine capable of comparing the archtype on the card and the spoken word.