Object Recognition Using Tactile Measurements: Kernel Sparse Coding Methods

Huaping Liu,Di Guo,Fuchun Sun

DOI: https://doi.org/10.1109/tim.2016.2514779

IF: 5.6

2016-01-01

IEEE Transactions on Instrumentation and Measurement

Abstract:Dexterous robots have emerged in the last decade in response to the need for fine-motor-control assistance in applications as diverse as surgery, undersea welding, and mechanical manipulation in space. Crucial to the fine operation and contact environmental perception are tactile sensors that are fixed on the robotic fingertips. These can be used to distinguish material texture, roughness, spatial features, compliance, and friction. In this paper, we regard the investigated tactile data as time sequences, of which dissimilarity can be evaluated by the popular dynamic time warping method. A kernel sparse coding method is therefore developed to address the tactile data representation and classification problem. However, the naive use of sparse coding neglects the intrinsic relation between individual fingers, which simultaneously contact the object. To tackle this problem, we develop a joint kernel sparse coding model to solve the multifinger tactile sequence classification problem. In this model, the intrinsic relations between fingers are explicitly taken into account using the joint sparse coding, which encourages all of the coding vectors to share the same sparsity support pattern. The experimental results show that the joint sparse coding achieves better performance than conventional sparse coding.