Multiple ANN Recognizers for Adaptive Recognition of the Speech of Dysarthric Patients in AAL Systems.
Journal:
Studies in health technology and informatics
PMID:
26294603
Abstract
People suffering from neuromuscular disorders are one of the main target groups of speech-controlled Ambient Assisted Living systems. However, the speech of these patients is often distorted because of the dysarthric symptoms of the disease. The dysarthria is known to become worse as the disease progresses. We propose a framework for an adaptive speech recognition system that may be able to follow the slow deterioration of speech quality without risking the accuracy of the system from incorrect data.