On the use of high-level information in speaker and language recognition

Automatic Speaker Recognition systems have been largely dominated by acoustic-spectral based systems, relying in proper modelling of the short-term vocal tract of speakers. However, there is scientific and intuitive evidence that speaker specific information is embedded in the speech signal in multiple shortand long-term characteristics. In this work, a multilevel speaker recognition system combining acoustic, phonotactic and prosodic subsystems is presented and assessed using NIST 2005 Speaker Recognition Evaluation data. For language recognition systems, the NIST 2005 Language Recognition Evaluation was selected to measure performance of a high-level language recognition systems.