Publications

sort by:  author date

Hybrid method of feature extraction from biometric signals

MATOUŠEK, V., MAUTNER, P., MUSIL, M., ROHLÍK, O. Hybrid method of feature extraction from biometric signals. In 48. Internationales Wissenschaftliches Kolloquium. Ilmenau: Technische Universität Ilmenau, 2003. s. 127-128.
Abstract PDF BibTeX

The article describes a hybrid approach to the feature extraction from signals generated by the special biometric pen with which the high recognition rates were achieved. The method is based on the combination of structural and statistic approaches; it computes the feature values as coefficients of the transformation of predefined wave shopes to the signal wave. The developed feature extraction method achieves much better recognition rates as all still developed techniques.

Detection of relevant speech features using driven spectral analysis

EKŠTEIN, K., MOUČEK, R. Detection of relevant speech features using driven spectral analysis. In Information Technologies & Control. Prague: Academy of Sciences of the Czech Republic, 2003. s. 25-25. ISBN: 80-239-1333-6
Abstract PDF BibTeX

The paper introduces a novel approach to the problem of speech signal analysis. Contemporary Automatic Speech Recognition systems do not involve an extensive intelligence into the task of deriving the feature vectors from frames of incoming speech. Although few ways to adopt the data to the needs of consecutive processing has been already propose, they are more or less static regarding the character of a speech signal.

Utterance models in dialogue systems

MOUČEK, R., EKŠTEIN, K. Utterance models in dialogue systems. In Information Technologies & Control. Prague: Academy of Sciences of the Czech Republic, 2003. s. 26-26. ISBN: 80-239-1333-6
Abstract PDF BibTeX

The paper deals with a building of semantic used as an interface between the module of linguistic analysis and dialogue manager. The syntactic and semantic analysis is based on the theory of microsituations and utterance models.

Signature verification using unsupervised learned neural networks

MAUTNER, P., MATOUŠEK, V., ROHLÍK, O., KEMPF, J. Signature verification using unsupervised learned neural networks. In Artificial Neural Networks in Pattern Recognition. Florence: University of Florence, 2003. s. 71-75.
Abstract PDF BibTeX

The Carpenter-Grosberg`s ART-2 and Kohonen`s Selforganizing Feature Map (SOFM) have been developed for the clustering of input vectors and have been commonly used as unsuperised learned classifiers. In this paper we describe the use of these neural network models for signature verification. The architecture of the verifiers and achieved results are discussed here and ideas for future research are also suggested.

Text, Speech and Dialogue, 6th International Conference, TSD 2003

Matoušek, V., Mautner, P. Text, Speech and Dialogue, 6th International Conference, TSD 2003 . České Budějovice, hotel Gomel, 08-SEP-03 - 12-SEP-03.
Abstract PDF BibTeX

TSD 2003 was concerned with topics in the field of natural language processing, in particular:corpora, texts and transcription; speech analysis, recognition and synthesis; their intertwining within NL dialog systems

Our partners