Second International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2001)
In the present work we introduce a new approach to pathological speech processing methods and recognition of various speech pathologies. We no longer attempt to recognise forms of pathological speech deformations, because it is impossible to show any compact template of a normal speech signal (as reference) and it is also impossible to show a standard form of any deformation. Almost any sample of speech, both normal and pathological, is unique and we must introduce a new concept for the treatment of pathological speech; artificial understanding of the nature (and source) of the pathology of speech instead of traditional classification or recognition of signals. We try to implement the process for the artificial understanding of the observed forms of speech deformations with a very close look at vocal tract pathologies. Selected excerpts of research are presented concerning the application of (modified) acoustic signal processing methods and (in particular) the neural network techniques specially designed for solving the problem of "understanding" selected pathologies of vocal tract.
Index Terms. speech analysis, pathological speech, vocal tract pathology, artificial intelligence
Bibliographic reference. Tadeusiewicz, Ryszard / Wszolek, Wieslaw / Izworski, Andrzej / Wszolek, Tadeusz (2001): "Understanding of vocal tract pathology using speech signals analysis", In MAVEBA-2001, 125-133.