The following video shows our Texplorer2 device, the novel recording procedure, parametrization, and display using a novel tactile display device (TCM2).
First, the textures are scanned under controlled conditions. Herein, all parameters are kept constant, except the applied force and scan velocity, which are explicitly controlled to allow for a systematic signal analysis. A specialised setup consisting of a rotary texture attachment and a PID-controlled Phantom Omnni is implemented for this purpose (compare to Fig. 1) . The inclination angle of the Omni is fixed, while either the applied force or the scan velocity is increased linearly. The main purpose of this controlled recordings is to test features and their variance against those two scan-time parameters.
The following video gives an overview of the robot-controlled setup and recording process for the LMT Haptic Texture Database using a PhantomOmni(Sensable) device.
The following figures illustrate the hardware setup used for the texture recording. Left: complete setup with a customized rotary plate used to record the controlled signals. Center: The two marked joints are fixed to ensure upright posture of the stylus during the recording. Right: close-up of the customized stainless steel tool tip and the accelerometer mounted on the stylus.
Figure 1 (adopted from [1])
Unconstrained Recordings during Freehand Surface Material Exploration
The second set of texture signals is recorded under less constrained conditions.
It consists of ten free hand recordings for each texture (five linear and five circular movements). These texture signals, recorded by a single subject following typical human exploratory movements, can serve as a test and training set for texture classification and recognition systems.
Figure 1 (adopted from [2])
The following videos give an overview of the Texplorer device and how a surface material is recorded.
The following image shows the used retrieval engine for surface material classification and retrieval of the most similar surface materials. Note that we use perceptually motivated features for surface material classifiaction. For perceptual evaluation, we use the outcome of a subjective experiment with 30 subjects to determine the perceived similarity between all surface materials.
The following image shows the perceptual similarity between all our surface materials.
The work presented on this web page has been financially supported by
the ERC Starting Grant ProHaptics (grant agreement no. 258941, 2011-2015) and the DFG project STE 1093/6-1 (2016-2018) and STE 1093/8-1.
* The research presented on these web pages has received financial support by the German Research Foundation
(DFG) under the project STE 1093/6-1 and 1093/8-1.
