TY - GEN
T1 - In-hand object recognition via texture properties with robotic hands, artificial skin, and novel tactile descriptors
AU - Kaboli, Mohsen
AU - De La Rosa, Armando T.
AU - Walker, Rich
AU - Cheng, Gordon
N1 - Publisher Copyright:
© 2015 IEEE.
PY - 2015/12/22
Y1 - 2015/12/22
N2 - This paper, for the first time, proposes a solution for the problem of in-hand object recognition via surface textures. In this study, a robotic hand with an artificial skin on the fingertips was employed to explore the texture properties of various objects. This was conducted via the small sliding movements of the fingertips of the robot over the object surface as a human does. Using our proposed robust tactile descriptors, the robotic system is capable of extracting information-rich data from the raw tactile signals. These features then assist learning algorithms in the construction of robust object discrimination models. The experimental results show that the robotic hand distinguished between different in-hand objects through their texture properties (regardless of the shape of the in-hand objects) with an average recognition rate of 97% and 87% while employing SVM and PA as an online learning algorithm, respectively.
AB - This paper, for the first time, proposes a solution for the problem of in-hand object recognition via surface textures. In this study, a robotic hand with an artificial skin on the fingertips was employed to explore the texture properties of various objects. This was conducted via the small sliding movements of the fingertips of the robot over the object surface as a human does. Using our proposed robust tactile descriptors, the robotic system is capable of extracting information-rich data from the raw tactile signals. These features then assist learning algorithms in the construction of robust object discrimination models. The experimental results show that the robotic hand distinguished between different in-hand objects through their texture properties (regardless of the shape of the in-hand objects) with an average recognition rate of 97% and 87% while employing SVM and PA as an online learning algorithm, respectively.
KW - Electrodes
KW - Robot sensing systems
KW - Skin
KW - Surface impedance
KW - Surface texture
UR - http://www.scopus.com/inward/record.url?scp=84962234426&partnerID=8YFLogxK
U2 - 10.1109/HUMANOIDS.2015.7363508
DO - 10.1109/HUMANOIDS.2015.7363508
M3 - Conference contribution
AN - SCOPUS:84962234426
T3 - IEEE-RAS International Conference on Humanoid Robots
SP - 1155
EP - 1160
BT - Humanoids 2015
PB - IEEE Computer Society
T2 - 15th IEEE RAS International Conference on Humanoid Robots, Humanoids 2015
Y2 - 3 November 2015 through 5 November 2015
ER -