cation methods for autism disordered speech
Authors: Zoubir Abdeslem Benselama, Mohamed A. Bencherif, Abderrezak Guessoum, Mohamed A. Mekhtiche
Number of views: 437
In this paper, we present the results of our investigation on Autism classifi
cation by applying ensemble classi
ers to disordered speech signals. The aim is to distinguish between Autism sub-classes by comparing an ensemble combining three decision methods, the sequential minimization optimization (SMO) algorithm, the random forests (RF), and the feature-subspace aggregating approach (Feating). The conducted experiments allowed a reduction of 30% of the feature space with an accuracy increase over the baseline of 8.66% in the development set and 6.62% in the test set.