A tactile-based fabric learning and classification architecture

Abstract

Abstract-This paper proposes an architecture for tactile-based fabric learning and classification. The architecture is based on a number of SVM-based learning units, which we call fabric classification cores, specifically trained to discriminate between two fabrics. Each core is based on a specific subset of the fully available set of features, on the basis of their discriminative value, determined using the p-value. During fabric recognition, each core casts a vote. The architecture collects votes and provides an overall classification result. We tested seventeen different fabrics, and the result showed that classification errors are negligible.

Fabric Classification Using a Finger-Shaped Tactile Sensor via Robotic Sliding

“…While the second type of method collects the tactile signals using sensors sensitive to vibrations. Tactile signals are first transformed into the frequency domain and then both temporal and frequency features are extracted to identify textures as in Fishel and Loeb (2012); Khan et al (2016); Kerr et al (2018); Massalim et al (2020). Instead of classifying the exact type of material, the work proposed by Yuan et al (2018) aims at recognizing 11 different properties from 153 varied pieces of clothes using a convolutional neural network (CNN) based architecture.…”

“…Several combinations of sliding speeds and normal forces are also tested to enable a Bayesian inference. Khan et al (2016) described a similar experiment with handcrafted statistical features to identify textures. The research employs a custom finger-shaped capacitive tactile sensor, which is mounted on the probe of a 5-axes machine and controlled to slide on a platform covered with the fabric.…”

“…Manfredi et al ( 2014 ) found that the vibrations elicited during the interaction carry information about the microgeometry of the fabric surface and mechanical properties of the tactile sensor itself. In Fishel and Loeb ( 2012 ); Khan et al ( 2016 ); Kaboli and Cheng ( 2018 ); Kerr et al ( 2018 ) sliding motions are conducted in different manners to collect vibration signals about the fabric textures. Particularly, Fishel and Loeb ( 2012 ) show that changing exploratory actions can affect the received tactile signals and lead to different classification performances.…”

“…Inspired by how humans try to identify fabrics with their skin solely, via sliding and rubbing fingers on the fabric surface, we command a robotic arm equipped with a capacitive tactile sensor on its end-effector to grip the fabric and slide. Different from some existing attempts of texture identification in Fishel and Loeb ( 2012 ); Khan et al ( 2016 ); Kerr et al ( 2018 ) that fix the inspectee material on a motorized platform where the sensor is stationary, our vibration signals are acquired during a dynamic process where fabric stripes are free to stretch and bend. We allow the 7-DOF robotic arm to carry out the exploratory motions in a large area, similar to the experiments in Bauml and Tulbure ( 2019 ) and Taunyazov et al ( 2019 ), which resembles the daily scenario where humans touch to feel the fabrics.…”

Active Haptic Perception in Robots: A Review

“…At the lowest level, tactile sensor arrays can be used to find contact locations with a high resolution, and to track variation of contact points. Model-based approaches can be used to retrieve higher-level information related to contact features, such as contact force distributions (Seminara et al, 2015) or contact shape (Khan et al, 2016; Wasko et al, 2019). In all cases, such algorithms are needed to embed 3D sensor locations into a lower dimensional space representing the robot surface, and a further processing step is required to move toward a higher dimensional space by computing features.…”