摘要 |
In a telephone banking service, initial log-on (before the service has identified the user) must use speaker-independent recognition (ie the speech recognizer 2 operates in accordance with word templates drawn from the population at large). The first digit of a sequence identifying the user is compared with certain templates of two sets of templates to determine which set most closely resembles the user's speech. The two sets of templates may comprise standard templates formed by male and female speakers respectively. <IMAGE> |