In Malaysia around 40,000 people suffer from stroke every year. At least one-fourth of stroke survivors experience aphasia, a communication impairment that varies considerably across patients but most involves some form of deficit in language comprehension. Language rehabilitation, especially when it is intense (8-10 hours/week for 12 weeks) and it starts early (after the patient's medical condition has stabilized, often within 24 to 48 hours after the stroke, preferably in acute-care hospitals), has been shown to be beneficial in speeding or inducing recovery. Lack of immediate access to rehabilitation programs can have a dear consequence, and this was the case in 2006, when only 30% of stroke survivors there claimed part recovery.

There are several compounded problems with language rehabilitation for stroke survivors in Malaysia. First is the shortage of speech therapists. Second is the lack of standard assessment of language comprehension and recovery in Malay. The normal practice is for individual speech therapists to translate existing English assessments, e.g., the Porch Index of Communicative Abilities, the Functional Communication Profile and the Western Aphasia Battery, into Malay (despite the difference in language structure). Third is the lack of transportation resources with the problem more pronounced in rural areas, causing patients to miss out on sessions due to inability to get to the care centers in bigger towns.

The proposed project aims at ameliorating the aforementioned problems by developing a mobile virtual speech therapist that can be self-administered by patients in their own homes. We propose mobile device rather than a desktop computer to host the virtual therapist because 90% of stroke survivors are over 60 and our past work found that many older people are intimidated by desktop computers but are willing to use portable devices such as cell phones. In 2007, the penetration of cell phones in Malaysia stands at 84%, and 3G services had been introduced since 2005. We are currently conducting a pilot study on the use of cell phone for phone-based therapy for stroke survivors in Malaysia and therefore have access to the patients and therapists. We propose to use our animated talking agent as a virtual therapist with accurate speech articulation to provide a more anthropomorphic interface than a standard button/menu interface (our work also found that older people do not enjoy interacting with machine and prefer human-like interaction for providing help). Our three-dimensional animated embodied agent provides realistic speech which is almost as accurate as a natural speaker, with lips, tongue and jaw movements to mimic natural oral movements. The agent has a realistic tongue and palate, which can be displayed by making his facial skin transparent. This system has been used successfully as a communication tutor and was shown to significantly improve language comprehension and speech production. The system also contains a language wizard that allows the easy creation of language learning lessons and exercises that can be tailored to the need of each patient.

2009 Update: