Vibration Influences Haptic Perception of Surface Compliance During Walking

Yon Visell; Bruno L. Giordano; Guillaume Millet; Jeremy R. Cooperstock

doi:10.1371/journal.pone.0017697

Abstract

Background

The haptic perception of ground compliance is used for stable regulation of dynamic posture and the control of locomotion in diverse natural environments. Although rarely investigated in relation to walking, vibrotactile sensory channels are known to be active in the discrimination of material properties of objects and surfaces through touch. This study investigated how the perception of ground surface compliance is altered by plantar vibration feedback.

Methodology/Principal Findings

Subjects walked in shoes over a rigid floor plate that provided plantar vibration feedback, and responded indicating how compliant it felt, either in subjective magnitude or via pairwise comparisons. In one experiment, the compliance of the floor plate was also varied. Results showed that perceived compliance of the plate increased monotonically with vibration feedback intensity, and depended to a lesser extent on the temporal or frequency distribution of the feedback. When both plate stiffness (inverse compliance) and vibration amplitude were manipulated, the effect persisted, with both factors contributing to compliance perception. A significant influence of vibration was observed even for amplitudes close to psychophysical detection thresholds.

Conclusions/Significance

These findings reveal that vibrotactile sensory channels are highly salient to the perception of surface compliance, and suggest that correlations between vibrotactile sensory information and motor activity may be of broader significance for the control of human locomotion than has been previously acknowledged.

Citation: Visell Y, Giordano BL, Millet G, Cooperstock JR (2011) Vibration Influences Haptic Perception of Surface Compliance During Walking. PLoS ONE 6(3): e17697. https://doi.org/10.1371/journal.pone.0017697

Editor: Justin Harris, University of Sydney, Australia

Received: November 21, 2010; Accepted: February 8, 2011; Published: March 25, 2011

Copyright: © 2011 Visell et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This research was supported by a grant from the Ministère du Développement économique, innovation et exportation of Quebec (http://www.mdeie.gouv.qc.ca/) for the European 7th Framework Programme project Natural Interactive Walking (NIW, no. 222107). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

The goal of this study is to measure empirically the role played by vibrotactile sensory information in the perception of ground surfaces during walking. To this end, we focused on a basic property of walking surfaces that is highly salient to locomotion—their mechanical compliance [1], [2]. We investigated what influence, if any, vibration feedback to the plantar soles may have on the perception of ground surface compliance during walking.

The perception of ground surfaces is instrumental to enabling us to move easily on foot in diverse natural environments. Human locomotor movements are adapted when stepping onto, off of, or moving over soft, irregular, or slippery surfaces in ways that minimize metabolic costs, reduce impact forces, or stabilize vertical center of mass [1]–[7]. Compliant ground surfaces, such as sand or soggy grass, perturb locomotion by degrading proprioceptive cues that are acquired via ground contact and by mechanical perturbations due to the compression of material underfoot. Walkers automatically modulate their gait pattern and biomechanics to compensate for such changes in compliance [2].

When haptic sensation in the feet is impaired, as a result of a disease such as diabetes, or through local anesthesia, it can have detrimental effects on locomotion [8]–[11]. However, knowledge about the influence of different sources of haptic sensory information, such as plantar force or vibromechanical stimuli, on the control of walking is incomplete.

Haptic compliance perception involves discerning the deformability of objects touched with the hand, or of surfaces felt underfoot. Compliance, the inverse of stiffness, is the ratio between displacement and applied force, , and is related to the intrinsic material property of elasticity. Most prior research has investigated compliance perception via manual touch [12]–[19], but the haptic perceptual system is also able to discriminate walking surfaces of different elasticity [20], [21]. Sensitivity is highest when there is direct contact between the surface of the skin and a deformable object. In this setting, cutaneous tactile cues predominate [17]. Conversely, when touch is mediated by a rigid link, such as a stick or a stiff shoe sole, cutaneous force cues are combined with proprioceptive information to form compliance estimates [15], [16], [19], [22]. If cutaneous information is blocked entirely, performance is greatly degraded [16].

While we are not aware of any prior investigation of effects of vibrotactile sensory information on compliance perception, it is well established that high-frequency mechanical vibrations generated during interaction with surfaces via manually tapping or scraping with a probe, or scanning with a finger, can influence the perception of properties such as hardness and texture [23]–[28]. For example, amplifying vibrations generated during manual surface scanning, or imposing sinusoidal vibrations, increases perceived surface roughness [29]. Vibrations produced during frictional sliding are indicative of movement [30], and could contribute to compliance perception. On this basis, it could be hypothesized that an amplification of plantar vibration intensity would lead to an increase in the magnitude of compliance estimates, because displacement and compliance are proportional.

Mechanical signals generated during walking on natural ground surfaces constitute a rich source of haptic sensory information [31]–[33]. The compression of many heterogeneous materials (e.g., wood, snow, gravel) results in inelastic, unrecoverable deformations with energy distributed over a broad frequency band [31]. The pattern of these vibrations is highly correlated with material displacement [34], [35], so it is natural to consider them as potential displacement cues. Giordano et al. found that walkers are able to distinguish between the feel of porous and solid ground surfaces, or rock gravel surfaces of different grades, when walking in shoes [36]. When plantar cutaneous input was masked by mechanical vibrations, in the form of synthesized pseudo-random noise (frequency distribution: 50 Hz to 1 kHz), performance was impaired, suggesting that vibrotaction played a significant role. (Further analysis is provided in: Giordano B, Visell Y, Cooperstock JR, Yao HY, Hayward V, and McAdams S (2010) Audiohaptic identification of ground materials during walking, Submitted.)

Relatively few studies have investigated haptic perception with the feet. However, the foot is serially homologous to the hand, and is highly evolved as a sensory instrument. Its perceptual-motor abilities are involved in the regulation of posture and locomotion [37], and in the estimation of ground slipperiness and slant [38]–[40]. The sensory physiology of the plantar sole is highly developed, with the same type of mechanoreceptor populations as are present in the hand: the fast-adapting (FA) type I and II and slow-adapting (SA) type I and II receptors [41], [42]. The sole is highly sensitive to vibration, with FA receptors comprising about 70% of the cutaneous population. Low-frequency forces are sensed by SA receptors [41], and by Golgi organs, muscle spindles, and joint capsule receptors in the muscles, tendons, and joints.

Vibromechanical stimulation of the plantar sole affects both cutaneous receptors and deeper foot and ankle proprioceptors. Such stimuli can result in real or illusory postural effects resembling those due to an increase in local pressure at the same location of the foot sole [43]–[46]. This could be taken to suggest that amplifying plantar vibration may, by increasing perceived forces, decrease ground compliance estimates—contrary to what is suggested by foot-ground mechanical considerations. However, studies of this type have generally been conducted while subjects stood in place, whereas haptic compliance perception always requires movement. In other experiments on vibration stimulation of the leg muscles or tendons, different effects have been observed to accompany stimulation provided during stance than those induced when it is provided during locomotion. In the former case, it induces whole-body postural tilts (attributed to illusory lengthening of the stimulated muscles), whereas during locomotion it results in modified stepping movements with little overall change in muscle coordination [47]–[49]. Courtine et al. argued that this reflects the fact that sensory inflow is processed depending on both the body segment where it arises and the performed task [48]. As a result, we questioned whether prior results on postural effects of plantar vibration would apply in our study, in which subjects were actively moving.

Our experiments evaluated the influence of vibrotactile information felt during stepping onto a floor surface on the perceived compliance of the latter. The above-referenced studies involved a diverse range of signal types (noise-like or natural textures, and sinusoidal stimuli), amplitudes, and temporal dependencies. Experiment 1 was designed to investigate effects of vibration feedback on perceived compliance, and to clarify their dependency on time- and frequency-domain stimulus properties. To further determine the extent to which vibrotactile sensory information is combined with cutaneous force and proprioceptive information in the perception of ground compliance, Experiment 2 measured the effect of plantar vibration on compliance perception via a novel apparatus that allowed both the mechanical stiffness of a floor plate and vibration feedback presented through it to be manipulated. Psychophysical amplitude detection thresholds for the stimuli were also measured in order to provide an indication of the relative intensity of the stimuli.

Materials and Methods

Ethics Statement

The experiments were conducted in accordance with McGill University ethics guidelines, and was reviewed and approved by the McGill Research Ethics Board in accordance with the requirements of the McGill University Policy on Ethical Conduct of Research involving Human Subjects and with the Tri-Council Policy Statement: Ethical Conduct For Research Involving Humans.

General Methods

During the experiments, subjects crossed a short walkway incorporating an actuated floor plate that provided vibration feedback in response to forces exerted by the foot. The mechanical stiffness of the plate was manipulated in Experiment 2.

Apparatus

The apparatus consisted of a short walking platform (Fig. 1) permitting subjects to take a single step onto a vibration-actuated floor plate. The plate was actuated by a Lorentz force inertial motor (Clark Synthesis model TST429) rigidly coupled to it from beneath. This plate was used to present walkers with vibration feedback and to present a specified mechanical stiffness to the walker, via a servo controlled mechanism (see Experiment 2).

Download:

Figure 1. Apparatus for producing compliance and vibration stimuli.

Subjects stepped from one side of the platform a onto the vibrating plate b, and onto the opposite platform a'. They then turned, stepped on b again, and returned to a. In Experiment 2, the plate also displaced up to 2 cm in the vertical direction, compressing a volume of EVA foam that was controlled by the linear servomechanism c, to produce the commanded compliance (see Fig. 3). Subjects entered their responses after each trial at the keyboard d and received instructions from the large-screen video monitor e.

https://doi.org/10.1371/journal.pone.0017697.g001

To ensure that vibromechanical stimuli could be reproduced accurately across a wide range of frequencies, while assuring the stability of the plate under a human walker, we undertook an extensive redesign of our earlier apparatus [50], as fully described in reference [51]. Through measurements, we determined that the device was able to reproduce arbitrary vibrations accurately at forces of more than 40 N within a flat frequency band from 50 to 750 Hz. Within the range of amplitudes used here, vibrations could be presented with a nonlinear waveform distortion of less than 3% (mean absolute percent error, measured at 300 Hz). The device could sense static or transient loads of more than 1000 N supplied by a human foot. Analog data from the force sensors were conditioned, amplified, and digitized via a 16-bit acquisition card (National Instruments model USB-6218). Digital-to-analog conversion of the vibration signal was performed using a 24-bit, 48 kHz audio interface (Edirol model FA-101). The analog signal was then passed through a power amplifier driving the actuator. In order to assess the accurate reproduction of the vibration stimuli, they were independently recorded with a miniature accelerometer permanently attached to the underside of the plate.

General Procedure

During the experiments, stimuli were presented via the plate as subjects stepped on it. They began on one side of the walkway (see Fig. 1), stepped onto the plate with their dominant foot, and proceeded to the opposite side, turned, stepped on the plate again using their dominant foot, returned to the first side, and entered their responses via a computer terminal. Before each experiment, subjects were instructed in the use of the apparatus and interface.

Both experiments took place in a structurally isolated, soundproofed room with a noise-floor rating of PNC20. Subjects wore foam earplugs with an NRR attenuation rating of 33 dB and wireless headphones playing pink noise at a volume sufficient to mask any sounds produced by the vibrating plate and the motors. The non-vibrating walking platforms were isolated from the actuators via cushioning material, eliminating the transmission of vibrations to users before stimulus presentation.

A steady walking pace was enforced via a 1 Hz metronome sound audible above the pink noise. The experiments were conducted at low light levels to allow subjects to focus on what they felt, but sufficient for the walkway to remain visible. However, subjects were asked to avoid looking down at the plate while walking on it, unless necessary to maintain equilibrium, and instead, were instructed to attend to one of the two static visual markers that were positioned at a height of 1.3 m (above foot level), and a distance 1 m from either end of the walkway.

Subjects were required to wear shoes in the experiment in order to avoid directing their attention to the surface properties of the plate. In order to standardize footwear in all experiments, only male subjects were recruited, with North American shoe size between 7 and 12. Each was given an identical model men's hard soled dress shoe in the appropriate size to wear. All subjects reported normal tactile sensation in the feet, with normal walking ability, and were naive with respect to the purpose of the study. They were presented with and signed informed consent forms at the beginning of the experiment and were paid ten dollars (CAD) per hour for their participation upon completion.

Experiment 1

Experiment 1 was based on ratings of subjective compliance. We investigated the perceived compliance of a rigid plate augmented with nine different types of vibration feedback at two amplitude levels, as well as one condition in which no vibration feedback was provided. Subjects walked across each configuration of the plate, and rated its compliance on a continuous scale.

Recruitment.

Twenty people participated in the experiment (mean age 24.5 years, STD = 6.9 years, average mass 70.9 kg, STD = 11.2 kg). None participated in any other study on compliance or vibration perception. Other details were as described under “General Methods”.

Stimuli.

The stimuli consisted of several different types of vibration feedback and one no-vibration reference condition. The stiffness of the plate was held constant, and was set equal to 90 N/mm, the median stiffness value used in Experiment 2. The 18 vibration stimuli were generated by factorial combination of three parameters: amplitude scale (0.5, 1.0), temporal waveform type (Sinusoidal, White Noise, Textured Noise), and amplitude envelope (Constant, Force-Proportional, Dynamic). Each stimulus can be described as an acceleration signal delivered from the plate, having the form , where were stimulus-dependent peak gain factors.

The vibration signals for the nine stimuli resulting from combining the factors “amplitude envelope” and “waveform type” are shown in Fig. 2. Three different waveforms were used. The first was a sinusoid with frequency . The second was a white noise that was band-limited by filtering to remove frequencies above 700 Hz and below 50 Hz. The third was a noise signal intended to resemble the texture felt when a porous material, such as gravel, is compressed. It was obtained by passing an impulsive noise source, consisting of a random impulse train, through a resonant filter. The impulses were identical in amplitude scale, and occurred at times whose time intervals were sampled from a Poisson stochastic process; the intervals were distributed as . The mean event frequency was . Each impulse was rendered as a 1 ms white noise burst beginning at . The impulse train was passed through a second-order infinite impulse response (IIR) bandpass filter with center frequency and bandwidth . The resulting noise had a rough texture with most energy concentrated in a narrow frequency band at which FA type II mechanoreceptors in the foot sole are most sensitive.

Download:

Figure 2. Vibration feedback stimuli.

The thin lines (rows 1, 3, and 5) show force profiles from footsteps of one participant onto the plate, and the darker waveforms are the corresponding vibration feedback stimuli. Vibrations could be felt only during foot-plate contact. Stimuli are labeled with a 2-letter string, with the first encoding waveform type (S, W, T = Sinusoidal, White noise, Textured noise), and the second encoding envelope type (C, P, D = Constant, Proportional, Dynamic). The vibration amplitude range was normalized for display purposes.

https://doi.org/10.1371/journal.pone.0017697.g002

The three different envelopes specified the amplitude profile of the vibration feedback in response to a footstep with normal force profile . The first was a constant function , and the second was a linear force-proportional envelope , with inverse slope . The third was a dynamic envelope derived from an admittance-based simulation of a linear, compressible material, where is the time-derivative of the virtual strain response to , described by(1)The envelope is obtained in real time as the output of a second-order digital IIR bandpass filter computed by solving (1) using the Laplace transform method and the bilinear transform [52], [53]. The filter input was and the output was . could be taken as the mass of a representative volume element, but here it is an arbitrary gain factor. We used frequency , and set for critical damping. This yielded a characteristic envelope response time of .

As explained in reference [51], the combination of the dynamic envelope type with the textured noise waveform (i.e., stimulus TD) can be regarded as a simplified micromechanical model for the production of textured vibrations during the compression of a natural, heterogeneous material such as gravel, sand, or snow.

Stimulus intensity equalization.

In a pilot study, we observed that the vibration feedback stimuli could significantly increase subjective compliance ratings, and that the effect depended primarily on the stimulus amplitude parameter. However, RMS signal energy and subjective stimulus intensity depended on amplitude, waveform, and envelope type parameters. To ensure that the latter two could be manipulated independently of amplitude, a separate procedure was used, prior to the main part of Experiment 1, to equalize the stimuli with respect to the subjective intensity of vibration. Ten subjects that did not participate in the main experiment were recruited for this equalization experiment (mean age 23.1 years, STD = 6.1 years, average mass 71.1 kg, STD = 10.4 kg), which was based on a two-alternative forced-choice adaptive staircase method. On each trial, subjects walked across two configurations of the plate differing in vibration feedback type and amplitude and reported whether the first or second vibration felt stronger. The order of the two stimuli was random from trial to trial. One of the two, the standard, was always the high-amplitude white noise stimulus (WC1). The other, the comparison, was one of the remaining eight stimulus types parametrized by waveform and envelope (type SC, SP, SD, WP, WD, TC, TP, or TD). The amplitude of the comparison was controlled by a staircase method that tracked the point of subjective equality, i.e., the amplification factor for the comparison stimulus that rendered it as intense as the standard. If the subject indicated that the comparison felt stronger (respectively weaker), then its amplitude was reduced (respectively increased) by one step unit. The step size was initially large (10 dB) and became smaller (3 dB) after two reversals in the direction of the threshold-tracking sequence. Each staircase was run for 12 reversals, and the point of subjective intensity equivalence was calculated as the average between the last 8 reversals. A total of 16 staircases (8 interleaved pairs) were completed by each subject. Other details were as described under “General Procedure”.

The results of this procedure were used to assign the values of the gain of the Experiment 1 stimuli. The stimuli were regarded as equal in subjective intensity, within limitations determined by experiment duration and inter-subject variability. Table 1 reports the measured peak and RMS gain values for the equalized stimuli, which are labeled with a two-letter string, with the first encoding waveform (S, W, T = Sinusoidal, White noise, Textured noise), the second encoding envelope (C, P, D = Constant, Proportional, Dynamic). Values are reported for the high-amplitude () stimuli. Both experiments also included stimuli with . Amplitudes were verified by accelerometer measurement while the plate was loaded by a footstep. For the noise stimuli (W and T), since the absolute peak could vary between presentations, a stable measure was obtained as the median of peak amplitudes on a set of 10 ms windows spanning the highest amplitude interval.

Download:

Table 1. Experiments 1 and 2: Peak and RMS amplitudes of plate acceleration,

for the high amplitude

vibration stimuli.

https://doi.org/10.1371/journal.pone.0017697.t001

Procedure.

For each stimulus presentation during the experiment, subjects walked across the plate and rated its compliance using a slider labeled “most compliant” and “least compliant” at the two extremes. Subjects were informed that they might, at times, feel vibrations via their feet, but no further elaboration was given. The first experimental block was a warm-up period in which subjects tried all configurations of the plate that would be presented. In this period, they were instructed to focus on the maximum and minimum compliance within the stimulus set. During the remainder of the experiment, subjects were asked to use the entire range of the slider when rating the stimuli. The 18 equalized stimuli were presented in blocked randomized order, each stimulus being presented once on each of 12 blocks, for a total of 216 trials. The resulting data consisted of twelve compliance ratings per stimulus from each subject. The entire experiment lasted 90 minutes. There was a pause of two minutes between blocks, and a pause of five minutes after the sixth block. Additionally, there was a pause of at least five seconds between stimuli, as subjects entered their responses. Other details were as described under “General Procedure”.

Experiment 2

The experiment investigated the extent to which vibration feedback modified perception of the compliance of the floor plate when both vibration amplitude and plate stiffness were manipulated. The resulting data consisted of the proportion of responses in which the comparison was judged more compliant, for each stiffness and amplitude level.