Year: 2003
Type: conference-paper
Abstract: Human developmental learning is capable of\ndealing with the dynamic visual world, speech-based\ndialogue, and their complex real-time association.\nHowever, the architecture that realizes\nthis for robotic cognitive development has\nnot been reported in the past. This paper takes\nup this challenge. The proposed architecture does\nnot require a strict coupling between visual and\nauditory stimuli. Two major operations contribute\nto the “abstraction” process: multiscale temporal\npriming and high-dimensional numeric abstraction\nthrough internal responses with reduced variance.\nAs a basic principle of developmental learning,\nthe programmer does not know the nature\nof the world events at the time of programming\nand, thus, hand-designed task-specific representation\nis not possible. We successfully tested the\narchitecture on the SAIL robot under an unprecedented\nchallenging multimodal interaction mode:\nuse real-time speech dialogue as a teaching source\nfor simultaneous and incremental visual learning\nand language acquisition, while the robot is viewing\na dynamic world that contains a rotating object\nto which the dialogue is referring.
Authors: Yilu Zhang, Juyang Weng
Language: English
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W1551998207
