Fingertip-Based Feature Analysis for the Push and Stroke Manipulation of Elastic Objects

Area of Science:

• Robotics
• Human-Computer Interaction
• Biomechanical Engineering

Background:

• Understanding human manipulation of elastic objects is crucial for advanced robotics and haptic interfaces.
• Existing methods often lack robustness to variations in operators, finger placement, and object properties.

Purpose of the Study:

• To develop and validate robust fingertip-based feature descriptors for quantitative classification of finger operations.
• To investigate the invariance of these descriptors across different operators, finger positions, and elastic object characteristics.

Main Methods:

• A wearable system was developed for simultaneous measurement of fingertip position and strain during object interaction.
• User experiments involved 10 subjects performing push and stroke operations on nine distinct elastic objects.
• Time-series data of fingertip position and strain were collected and analyzed using binary classification with a support vector machine and cross-validation.

Main Results:

• The proposed two-dimensional features derived from fingertip position and strain demonstrated stable recognition of push and stroke operations.
• Classification accuracy remained high across elastic objects with varying shapes, stiffnesses, and thicknesses.
• Effective recognition was achieved within a 0.9-second time frame.

Conclusions:

• Fingertip position and strain data provide robust features for classifying basic finger operations on elastic objects.
• The developed method offers a promising approach for tactile sensing in human-robot interaction and prosthetic devices.
• This research contributes to quantitatively understanding and replicating human tactile manipulation skills.