Central and Eastern European Survey

---

Type of Organisation: public
Number of Employees: Less than 10
Activities developed: Research, Education, Software.

At our department the problem of speech and speaker recognition is investigated. Linear prediction (LPC) model parameters and cepstral parameters (LPCC) of vocal tract and residue signal are used as features for speaker recognition. They are extracted from pseudostationary segments of voiced sounds. Likelihood ratio and cepstral distances are used for comparing of feature vectors. Three methods were developed for speaker recognition: method based on an average distance between vocal tracts of speakers; method based on vector quantization; method based on an average distance between vocal tracts and residue signals of speakers. We have developed isolated word recognition system prototipe based on dynamic time warping (DTW). Now our attention is concentrated on use of Hidden Markov Model (HMM) for isolated word recognition. Computer programms on C language that implement the three methods of speaker recognition are created. We use the methods of speaker identification for forensic examination in collaborating with Lithuanian Institute of Forensic Examination . Isolated word recognition modeling, based on DTW, software is developed. We use PC Pentium III and higher, C/C++ language, operating system WINDOWS 98. In fields of natural language processing problems of computational morphology are investigated. We are participating in national program "Lithuanian language in information society"(2000- 2006). Researchers of our department deliver lectures on signal processing,speech processing, pattern recognition and informatics at Vilnius Technical University, Vilnius Pedagogical University and Kaunas Vytautas Magnus University

This page is no longer maintained. Please visit http://www.elsnet.org/survey/quests to find out how to update your organisation profile or to find information about this organisation