Home /Research /Object recognition by a 'smart' tactile sensor
HRI

Object recognition by a 'smart' tactile sensor

Robin Andrew Russell

Year
2000
Citations
24

Abstract

This paper describes a new approach to object recognition based on the time evolution of tactile images produced by an array tactile sensor. Tactile sensation from the · fingertip skin conveys a lot of information about the shape of a touched object. It is usually considered that this should be.combined with proprioceptive information to give a haptic interpretation of the data as the result of an active search. The method described in this paper determines the position, orientation and identity of objects from a sequence of tactile images gathered over time. External information such as finger joint angles and finger position are not used. It is envisaged that this algorithm could be incorporated into a self-contained 'smart ' tactile sensor. Such a sensor would provide information for a robotic manipulation system including object identity. This information would not require significant additional processing in order to be useful. Implementation details are described in this paper together with experimental results. 1

Keywords

Tactile sensorComputer visionComputer scienceCognitive neuroscience of visual object recognitionArtificial intelligenceObject (grammar)Robot

Related papers

Browse all HRI papers