International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 1999)
In the present paper a set of artificial intelligence methods are presented, with special focus on the pattern recognition algorithms and neural networks techniques, applied to the evaluation of deformed speech signal. The principal idea of the paper is the statement that in the presented problems the standard methods of signal processing and classification, widely applied for analysis and recognition of normal speech, are totally ineffective. In the present work particular attention has been focused on the evaluation of structure for the feature space describing the pathological speech signal. The main original result presented in the paper is the choice of proper vectors of acoustic features adapted for description of those properties of the speech signal, which turned out to be useful for the medical diagnosis, as the ultimate goal of the study is a construction of a diagnostic system for a wide variety of pathological speech signals.
Bibliographic reference. Tadeusiewicz, Ryszard / Wszolek, Wieslaw / Izworski, Andrzej / Wszolek, Tadeusz (1999): "Methods of deformed speech analysis", In MAVEBA-1999, 132-139.