RT Journal Article
T1 Human and Object Recognition with a High-resolution tactile sensor
A1 Gómez-de-Gabriel, Jesús Manuel
A1 García-Cerezo, Alfonso José
A1 Gandarias, Juan Manuel
K1 Sensores
AB This paper 1 describes the use of two artificial intelligence methods for objectrecognition via pressure images from a high-resolution tactile sensor. Both meth-ods follow the same procedure of feature extraction and posterior classificationbased on a supervised Supported Vector Machine (SVM). The two approachesdiffer on how features are extracted: while the first one uses the Speeded-UpRobust Features (SURF) descriptor, the other one employs a pre-trained DeepConvolutional Neural Network (DCNN). Besides, this work shows its applica-tion to object recognition for rescue robotics, by distinguishing between differ-ent body parts and inert objects. The performance analysis of the proposed methods is carried out with an experiment with 5-class non-human and 3-classhuman classification, providing a comparison in terms of accuracy and compu-tational load. Finally, it is discussed how feature-extraction based on SURF can be obtained up to five times faster compared to DCNN. On the other hand, theaccuracy achieved using DCNN-based feature extraction can be 11.67% superiorto SURF.
PB IEEE
YR 2017
FD 2017-10-29
LK https://hdl.handle.net/10630/14881
UL https://hdl.handle.net/10630/14881
LA eng
NO Proyecto DPI2015-65186-REuropean Commission under grant agreement BES-2016-078237.Universidad de Málaga. Campus de Excelencia Internacional Andalucía Tech.
DS RIUMA. Repositorio Institucional de la Universidad de Málaga
RD 21 ene 2026