This paper presents a real application that adds learning capabilities to a telerobotic system, designed formerly to manipulate everyday objects over a board. The user interaction is based on a restricted natural language and object recognition techniques have been designed specifically in order to acquire a system that responds as fast as possible to the user commands. First of all, the article will introduce the overall description of the system, in order to get a general idea of its main functions. Then, it will focus on a comparison of different recognition procedures that have been already added to the implementation. It will show the algorithm that fits better to the real needs of the project, looking for a balance between efficiency and recognition capabilities.
展开▼
