Abstract:

Maximum a posteriori adapted Gaussian mixture model (GMM-MAP) is widely used in speaker verification. GMMs have three sets of parameters to be adapted: means, covariances and weights. However, practice has shown that it is sufficient to adapt the means only. Motivated by this, we formulate maximum a posteriori vector quantization (VQ-MAP) procedure which stores and adapts the mean vectors (centroids) only. Experiments on the NIST 2001 and NIST 2006 corpora indicate that VQ-MAP gives comparable accuracy with GMM-MAP with simpler implementation and faster adaptation.

Abstract:

A so-called modulation spectrogram is obtained from the conventional speech spectrogram by short-term spectral analysis along the temporal trajectories of the frequency bins. In its original definition, the modulation spectrogram is a high-dimensional representation and it is not clear how to extract features from it. In this paper, we define a low-dimensional feature which captures the shape of the modulation spectra. The recognition accuracy of the modulation spectrogram based classifier is improved from our previous result of EER=25.1 % to EER=17.4 % on the NIST 2001 speaker recognition task.

Abstract:

Institute of Mathematics and Informatics (www.mii.lt) has been established in 1956. Currently it is the largest in Lithuania independent state research institute. Fields of activities of the institute: scientific research and experimental development, doctoral studies, scientific organizational work, publishing, education The main research areas: data analysis; differential equations; optimization; system analysis; numerical analysis; mathematical logic; mathematical statistics; multimedia systems; probability theory; process control; software engineering; signal processing; informatics methodology. Six international journals are published. Two of them deal with the methodology of informatics: "Informatics in Education", and "Olympiads in Informatics". There are about 50 Ph.D. students at the institute. Example of a dissertation will be given. The research interests of the System Analysis Department and examples of applications will be presented.

Abstract:

More and more applications of intelligent recogniton have been witnessed in recent years. Among these applications, Biometrics and character recognition are typical examples. Biometrics has been widely employed for personal identification, such as in access control, based on acquired physiological or behavioral characteristics. Character recognition has many interesting applications, e.g., recognizing postcode or addresses for automatic sorting of post-mails and packages, character detection and analysis of electronically scanned documents, handwriting recognition for inputting words into computers or digital terminals. In this talk, traditional biometric methods as well as some new methods in biometrics, e.g., palm print, ear, tongue, gender recognition, will be presented. At the same time, novel methods for recognition of Chinese characters, digital numbers, and vehicle license plates will also be discussed. Some existing problems, new progress, and potential applications for intelligent recognition will be discussed as well.