Wearable Biofeedback System to Induce Desired Walking Speed in Overground Gait Training

Abstract

Biofeedback systems have been extensively used in walking exercises for gait improvement. Past research has focused on modulating the wearer's cadence, gait variability, or symmetry, but none of the previous works has addressed the problem of inducing a desired walking speed in the wearer. In this paper, we present a new, minimally obtrusive wearable biofeedback system (WBS) that uses closed-loop vibrotactile control to elicit desired changes in the wearer's walking speed, based on the predicted user response to anticipatory and delayed feedback. The performance of the proposed control was compared to conventional open-loop rhythmic vibrotactile stimulation with N = 10 healthy individuals who were asked to complete a set of walking tasks along an oval path. The closed-loop vibrotactile control consistently demonstrated better performance than the open-loop control in inducing desired changes in the wearer's walking speed, both with constant and with time-varying target walking speeds. Neither open-loop nor closed-loop stimuli affected natural gait significantly, when the target walking speed was set to the individual's preferred walking speed. Given the importance of walking speed as a summary indicator of health and physical performance, the closed-loop vibrotactile control can pave the way for new technology-enhanced protocols for gait rehabilitation.

Keywords: SportSole; closed-loop control; instrumented footwear; real-time gait parameter estimation; wearable biofeedback system.

Figure 1

(a) The proposed wearable biofeedback system (WBS) consists of two insole modules, a data logger, and a vibration-control unit; (b) The insole module includes a multi-cell piezo-resistive sensor, an IMU, and four eccentric rotating mass (ERM) motors, all embedded in the insole; (c) The vibration-control unit includes a control board and a Li-Po battery. The instrumented insoles are fitted inside regular sneakers; the logic unit is housed inside a customized 3D-printed enclosure and attached to the wearer’s shoes with a clip; and the vibration-control unit is attached to the user’s distal shank through Velcro straps.

Figure 2

System architecture.

Figure 3

Flowchart of closed-loop vibrotactile control. ${\theta }_p$ is the measured foot pitch angle, and a is the foot acceleration. $S{V}_h$, $S{L}_h$, and $S{T}_h$ are the real-time stride-to-stride estimates of stride velocity, stride length, and stride time, respectively. $S{V}_T$ represents the target walking speed. ${\varphi }_h$ and ${\varphi }_{WBS}$ are the user’s current gait phase and target phase, respectively. $\Delta {\varphi }_d$ is the target phase difference. IC, initial contact; FF, foot-flat; TO, toe-off; AFO, adaptive frequency oscillator.

Figure 4

Effects of the closed-loop stimulation on the gait of a representative participant. The black line represents the normalized ground reaction force (GRF) extracted from the multi-cell piezo-resistive sensor. If the user’s current walking speed ($S{V}_h$) is slower than the target speed ($S{V}_T$) set by the experimenter, the stimuli anticipate the IC to encourage a faster pace (a); Conversely, if $S{V}_h$ is faster than $S{V}_T$, the stimuli lag the IC, to elicit a slower pace (b).

Figure 5

(a) Experimental protocol. The sequence of the two rhythmic stimuli (OS = open-loop vibrotactile stimuli, CS = closed-loop vibrotactile stimuli) was randomized. The stimuli in OS mode were triggered at a constant pace corresponding to a target cadence $CA{D}_T$, while the stimuli in CS mode were modulated by the PI controller, given a target velocity $S{V}_T$; (b) For all the tasks, participants were instructed to walk counter-clockwise along a prescribed oval path marked on the floor. BL, baseline.

Figure 6

(a,b) Group averages ($AVG$) and (c,d) coefficient of variation ($CV$) of $SV$ and $CAD$ induced by the two stimulation modes (OS = open-loop vibrotactile stimuli, CS = closed-loop vibrotactile stimuli) during Session 1 (S1), as compared to their baseline (BL) values. Error bars indicate ± 1SE.

Figure 7

(a,b) Percentage changes in $SV$ and $CAD$ (relative to their baseline values $S{V}_P$ and $CA{D}_P$) induced by the two stimulation modes during S1, as functions of the baseline values; (c,d) Changes in the coefficients of variation ($CV$) of $SV$ and $CAD$ induced by the two stimulation methods during S1, as functions of the baseline $CV$. In all plots, each mark represents a participant.

Figure 8

(a,b) Group averages of the percentage mean absolute errors (MAE%) of $SV$ and $CAD$ induced by the two stimulation methods, OS and CS, during S2. MAE% values are computed with respect to the target values $S{V}_T$ and $CA{D}_T$; (c,d) Coefficient of variation ($CV$) of $SV$ and $CAD$ induced by the two stimulation methods during S2. Error bars indicate ± 1SE. •• indicates $p<0.01$.

Figure 9

(a,b) $S{V}_h$ for a representative participant for the two stimulation methods, OS and CS, during S3. The black vertical line represents the time at which the stimulation engine was activated during the walking task; (c) Group averages of the percentage mean absolute errors (MAE%) of $SV$ induced by OS and CS during S3. MAE% values are computed with respect to the time-varying target values $S{V}_T\left(t\right)$. Error bars indicate ± 1SE. • indicates $p<0.05$.

Figure 10

(a,b) $CA{D}_h$ for a representative participant for the two stimulation methods, OS and CS, during S3. The black vertical line represents the time at which the stimulation engine was activated; (c) Group averages of the percentage mean absolute errors (MAE%) of $CAD$ induced by OS and CS during S3. MAE% values are computed with respect to the time-varying target values $CA{D}_T\left(t\right)$. Error bars indicate ± 1SE. •• indicates $p<0.01$.