A Conversational Agent for People with Dementia: Supporting and Assessing the Patients

With the goal of supporting people with Dementia and reducing the caregivers’ physical and mental burden, this paper proposes a conversational agent that can serve as a conversation partner for people with Dementia. First, we developed an autonomous conversational agent that can generate backchannel feedback during the user's speech, such as head nod and verbal acknowledgement based on the acoustic information in user’s speech: pitch and intensity. Then, we propose an assessing system for people with Dementia which recognizes the users’ speech and head movements and records their verbal and nonverbal behaviors during the conversation with the agent. The behavioral records are expected to be useful for the caregivers to know the mental status of the patients