An Overview of Statistical Pattern Recognition Techniques for Speaker Verification

Even though the subject of speaker verification has been investigated for several decades, numerous challenges and new opportunities in robust recognition techniques are still being explored. In this overview paper we first provide a brief introduction to statistical pattern recognition techniques that are commonly used for speaker verification. The second part of the paper presents traditional and modern techniques which make real-world speaker verification systems robust in degradation due to the presence of ambient noise; channel variations, aging effects, and availability of limited training samples. The paper concludes with discussions on future trends and research opportunities in this area.