Handwriting and Speech Prototypes for Parkinson Patients : Belief Network Approach
Iyad Zaarour, Asscociate Professor at Lebanease University
jeudi 10 novembre 2011 à 14h00
Articulatory phonetics and handwriting dysfunctions are frequent observations in Parkinson’s disease (PD). In this paper we make an inductive study of speech and handwriting patterns of PD patients by proposing ways for discovering clusters of PD patients, these clusters will share the same writing and speech patterns. For this approach a combined acquisition of electronic pen and speech signals have been performed to ten PD patients that share the same experimental conditions. The acquired signals are preprocessed and were subjected to feature extractor through signal processing algorithms. Our modeling approach is based on unsupervised learning of a probabilistic graphical model, i.e. a Bayesian network based on EM algorithm. By considering that each writing and voice test is represented by its own local prototype and that there exists a global prototype which deals with each local prototype, each prototype is represented by a hidden variable which is a source influencing and acting on both types of patterns (pattern of speech and writing). The discovered prototype serves as a helpful assistant such as a Motor Diagnostic Tool based on Articulatory and Handwriting diagnosis, more specifically for PD.
Key words : Axial Symptoms, Bayesian belief Network, Classification, Clustering, EM Algorithm, Hidden Variables, Hierarchal structure, Parkinson’s disease (PD).