Texture Signals in Whisker Vibrations

Abstract

. Rodents excel in making texture judgments by sweeping their whiskers across a surface. Here we aimed to identify the signals present in whisker vibrations that give rise to such fine sensory discriminations. First, we used sensors to capture vibration signals in metal whiskers during active whisking of an artificial system and in natural whiskers during whisking of rats in vivo. Then we developed a classification algorithm that successfully matched the vibration frequency spectra of single trials to the texture that induced it. For artificial whiskers, the algorithm correctly identified one texture of eight alternatives on 40% of trials; for in vivo natural whiskers, the algorithm correctly identified one texture of five alternatives on 80% of trials. Finally, we asked which were the key discriminative features of the vibration spectra. Under both artificial and natural conditions, the combination of two features accounted for most of the information: The modulation power-the power of the part of the whisker movement representing the modulation due to the texture surface-increased with the coarseness of the texture; the modulation centroid-a measure related to the center of gravity within the power spectrum-decreased with the coarseness of the texture. Indeed, restricting the signal to these two parameters led to performance three-fourths as high as the full spectra. Because earlier work showed that modulation power and centroid are directly related to neuronal responses in the whisker pathway, we conclude that the biological system optimally extracts vibration features to permit texture classification.

Contact type dependency of texture classification in a whiskered mobile robot

“…The first pair of features is inspired by inspection of sampled whisker signals. The other two pairs are inspired by previous neurophysiological [2] and modeling [17], [23] investigations of the rat vibrissal system. In each case we present raw scatter plots of the data classes under the features, and confusion matrices obtained using a standard Gaussian classifier [7].…”

“…In our analysis, the average over time is implicit as we work with a largewindow DFT containing all time points of interest. [17] proposed that the centroid and energy of the contact period spectrum could correspond to the physical frequency and depths of bumps in the texture surface. Hipp et al then tested this hypothesis in experiments in which actuated 80mm steel whiskers were moved against a variety of textured surfaces.…”

“…The brushing of each vibrissa against the target surface induces vibrations in the whisker shaft that are transduced into neural signals by mechanoreceptors in the whisker follicle [23]. The neural encodings of these vibration patterns, formed across the vibrissal array, therefore provide the basis for the texture discrimination capabilities shown by the rat [1], [16]. High-speed videographic observations suggest that contact on only a small number of macrovibrissae over 3-5 whisking cycles is sufficient for reliable texture discrimination [10].…”

“…Here we will take inspiration from studies of coding in the neural pathways of the rat whisker system as one source of candidate features for resolving texture classifications problems for artificial vibrissae. One previous study [17] has taken a similar approach in which frequency-based features inspired by the analysis of activity in whisker ("barrel") sensory cortex were used to successfully classify signals obtained from an actuated 80mm metal whisker moved against textured surfaces from a known fixed starting position. The features identified by [17], along with others inspired by our own investigations of coding in the vibrissal primary afferent neurons [23], will be evaluated below for their usefulness in discriminating texture in a robotic whisker context.…”

Towards hierarchical blackboard mapping on a whiskered robot

“…Individual components of such a system have previously been investigated in isolation, including whiskered texture recognition [15], [26], [11], [17], [32], surface shape recognition [30], [24], [21], [13], and object recognition [19]. These components have previously been tested under ideal laboratory conditions or in individual mobile settings [41]; here we present steps integrating them into a single platform for hierarchical object recognition, along with results and observations on their performance 'in the wild' in a common arena environment.…”