Classification of Pathological Speech Using Fusion of Multiple Subsystems

Pathological speech usually refers to the condition of spee ch distortion resulting from atypicalities in voice and/or in the articulatory mechanisms owing to disease, illness or other ph ysical or biological insult to the production system. While au tomatic evaluation of speech intelligibility and quality cou ld come in handy in these scenarios to assist in diagnosis and treatm ent design, the many sources and types of variability often make it a very challenging computational processing problem. In th is work we design multiple subsystems to address different aspects of pathological speech characteristics. These subsy tems are then fused at the binary hard score level (intelligible o r not intelligible) using Bayesian networks. Results show that s ubsystems, such as multiple language phoneme probability sys tem, prosodic and intonational subsystem, and voice qualit y and pronunciation subsystem, have discriminating power for in telligibility (9.8%, 17.1%, 14.6% higher than by-chance respe ctively). Noise-Majority based fusion shows 66.4% accuracy , but the performance improvement by fusion is not made. Also, voice clustering based joint classification is applied to mi ni ze misclassification of the best subsystem, and it shows the bes t classification accuracy (79.9% on dev set, 76.8% on test set) .