Aphasia is an acquired communication disorder that affects the ability to speak and understand spoken language, reading, and writing. Although common in the elderly, it can occur at all ages regardless of race or gender. With improved health care, people are living longer with aphasia, which has implications for quality of life. Thi trend has influenced approaches to aphasia rehabilitation, spurring researchers and clinicians to find ways to supplement direct impairment-based treatments with methods that maximize the potential of people with aphasia (PWA) to use residual language skils in their own environments. Our research team includes investigators with strong expertise in speech pathology, virtual reality, human-machine interfaces, dialogue systems, and speech recognition. Our long-term goal is to provide progressive functional rehabilitation programs through user-friendly software that will transform healthcare delivery by being both economically feasible and accessible to a broad spectrum of the population.
The specific aims of this application are to: 1) develop software capable of identifying the language and speech patterns common in aphasia, 2) begin testing human-avatar (virtual human) verbal interactions to implement interactive dialogue with human clients, and 3) compare the quantity and quality of language production in dialogues of virtual clinician-human client dyads and human clinician-human client dyads. We hypothesize that training individuals with aphasia with immediate responses from virtual humans will promote verbal output that is comparable to verbal output when conversing with a clinician in a structured conversational setting.
In Aim 1 we extend the customary treatment paradigm of clinician and client role-playing by developing a spoken dialogue system (SDS) that has individually customized acoustic models, language and lexical models, and dialogue structure. The SDS will be integrated with facial animation provided through a virtual clinician. This system will have the capability of recognizing context and sound variation in speech while continuing to present meaningful keywords in order to stimulate the generation of interactive speech by individuals with aphasia. During the software development phase, we will use a 'Wizard of Oz'paradigm that has a computerized image that simulates the speech generated by a hidden human in order to identify a range of conversational variations and key words for the customized software.
In Aim 3, we will apply the SDS with the virtual clinician developed in Aim 1. We will evaluate (1) language output in human clinician-client dyads and virtual clinician-client dyads using quantitative and qualitative measures of content, syntax, and morphology, and (2) receptiveness to interaction with the avatar and human clinician with a laboratory-developed questionnaire. Responses during the conversations will be recorded and digitized, and data will be analyzed during baseline and follow-up sessions will be evaluated for appropriate content and structure. Results of these studies have the potential to develop a virtual clinician- human interaction system that can be used independently by people with aphasia to practice and improve functional communication skills.
Aphasia is an acquired communication disorder resulting from brain damage and affecting the ability to speak and understand spoken and written language. Improvements in healthcare have enabled people to live longer with aphasia, which has implications for their quality of life. This application proposes to develop a computer-based treatment approach that promotes use of residual language skills in everyday living situations through client-virtual clinician conversations.