Speech-based technology has the potential to reinforce traditional aphasia therapy through the development of automatic speech-language assessment systems. Such systems can provide clinicians with supplementary information to assist with progress monitoring and treatment planning, and can provide support for on-demand auxiliary treatment. However, current technology cannot support this type of application due to the difficulties associated with aphasic speech processing. The focus of this dissertation is on the development of computational methods that can accurately assess aphasic speech across a range of clinically-relevant dimensions. The first part of the dissertation focuses on novel techniques for assessing aphasic speech intelligibility in constrained contexts. The second part investigates acoustic modeling methods that lead to significant improvement in aphasic speech recognition and allow the system to work with unconstrained speech samples. The final part demonstrates the efficacy of speech recognition-based analysis in automatic paraphasia detection, extraction of clinically-motivated quantitative measures, and estimation of aphasia severity. The methods and results presented in this work will enable robust technologies for accurately recognizing and assessing aphasic speech, and will provide insights into the link between computational methods and clinical understanding of aphasia.
展开▼
