Open AccessModeling of Action’s Semantic Memory Incorporated with Procedural and Skill Memory to Perform Tasks
Author(s) -
Rahul Shrivastava*,
Prabhat Kumar,
Sudhakar Tripathi
Publication year - 2019
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.c4227.098319
Subject(s) - computer science , action (physics) , context (archaeology) , task (project management) , semantic memory , artificial intelligence , procedural memory , process (computing) , human–computer interaction , cognitive science , psychology , cognition , neuroscience , paleontology , physics , management , quantum mechanics , economics , biology , operating system
In this paper, a computational model is proposed to mimic an action’s semantic, procedural and skill learning’s by an abstract modeling of cortical columns of the Neocortex, Basal ganglia and Cerebellum brain region. In proposed work, the action semantic Learning makes a robot capable to learn an action in terms of their body parts movement sequence that allows it to recognize the learnt action by seeing as well. Whereas in procedural, it allows to learn tasks in the form of action’s hierarchy and makes it capable to capture the environmental features as a context for action’s activations. The skill memory also been added in the proposed work which allows an agent to translate the action as per the current demand of the action. Also, the model has used Vnect model of computer vision to map the human motion into sequence of 3D skeleton of human body, therefore the model can learn by seeing, like humans. In experimental work, the model is tested on vague samples of few actions, where the model is found robust in action recognition task and performed well as per the expectations