Speaker recognition in telecom applications

Now that speech driven information services are developing into transaction systems the interest in speaker recognition has been increasing. Based on research carried out in the EU funded projects CAVE and PICASSO a number of issues are identified that have a direct impact on real-world applications of the speaker recognition (SR). In this paper technical issues are addressed first, followed by human factors issues. Then we identify a number of factors that are critical to the success of applications of SR in telematics transaction services. It is explained why SR, as a behavioural biometric protection technique, cannot provide very high security when it is used on its own. Therefore, due attention must be paid to a proper design of the services, taking security into account from the very start. In addition, it appears that SR can play a major role in making spoken language interfaces more user friendly. The paper ends with some conclusions and recommendations.