| Tartalmi leírás: |
This PhD thesis presented methods for exploiting the non-verbal communication of individuals suffering from specific diseases or health conditions aiming to reach an automatic screening of them. More specifically, we employed one of the pillars of non-verbal communication, paralanguage, to explore techniques that could be utilized to model the speech of subjects. Paralanguage is a non-lexical component of communication that relies on intonation, pitch, speed of talking, and others, which can be processed and analyzed in an automatic manner. This is called Computational Paralinguistics, which can be defined as the study of modeling non-verbal latent patterns within the speech of a speaker by means of computational algorithms; these patterns go beyond the linguistic) approach.
By means of machine learning, we present models from distinct scenarios of both paralinguistics and pathological speech which are capable of estimating the health status of a given disease such as Alzheimer's, Parkinson's, and clinical depression, among others, in an automatic manner.
|
