Estimation of ground reaction forces and joint moments on the basis on plantar pressure insoles and wearable sensors for joint angle measurement

Abstract

BACKGROUND:

Gait analysis is a useful tool medical staff use to support clinical decision making. There is still an urgent need to develop low-cost and unobtrusive mobile health monitoring systems.

OBJECTIVE:

The goal of this study was twofold. Firstly, a wearable sensor system composed of plantar pressure insoles and wearable sensors for joint angle measurement was developed. Secondly, the accuracy of the system in the measurement of ground reaction forces and joint moments was examined.

METHODS:

The measurements included joint angles and plantar pressure distribution. To validate the wearable sensor system and examine the effectiveness of the proposed method for gait analysis, an experimental study on ten volunteer subjects was conducted. The accuracy of measurement of ground reaction forces and joint moments was validated against the results obtained from a reference motion capture system.

RESULTS:

Ground reaction forces and joint moments measured by the wearable sensor system showed a root mean square error of 1% for min. GRF and 27.3% for knee extension moment. The correlation coefficient was over 0.9, in comparison with the stationary motion capture system.

CONCLUSIONS:

The study suggests that the wearable sensor system could be recommended both for research and clinical applications outside a typical gait laboratory.

Keywords

Pressure sensors ground reaction forces joint moments gait wearable system

1. Introduction

Gait analysis is a useful tool in supporting clinical decision making through the delivery of a clinical diagnosis, assessment of pathological gait, rehabilitation, and surgical interventions [1, 2, 3]. The most common stationary system for gait measurement is motion capture integrated with force platforms [4]. It involves observation of human gait with the use of cameras located at fixed points of the laboratory and passive markers located on the patient’s body. However, the main problem with the existing non-wearable systems is their high-costs, the fact that they are time-consuming, their limitation to the laboratory environment, and a requirement for well-trained staff to set up the equipment. In recent years, progress in new technologies has led to the development of a series of devices which enable making measurements with reliable information. They use different sensors: accelerometers, gyroscopic sensors, pressure sensors, force sensors, extensometers, goniometers, and electromyography located on different parts of the human body [4]. Some authors proposed wearable systems in the past, which were usually based on the development of miniaturized sensors and communication systems. Bae et al. [5] presented a mobile gait monitoring system for the diagnosis of abnormal gait and rehabilitation. In [6], the authors proposed a wireless wearable sensor system for successive non-laboratory measurement of ground reaction force, joint angles, joint forces, and joint moments. Liu et al. [7] developed a wearable ground reaction force sensor system to measure CoP and triaxial GRF in a number of walking trials. There is still an urgent need to develop low-cost, unobtrusive mobile health monitoring systems that can provide information about ground reaction forces and joints moments in humans in the real world. The goal of this study was twofold. Firstly, a wearable sensor system composed of plantar pressure insoles and wearable sensors for joint angle measurement was developed. Secondly, the accuracy of the system in the measurement of ground reaction forces and joint moments was examined.

Figure 1.

The wearable sensor system: a) structure of the system with degrees of freedom; b) system components: 1 – measuring card for data acquisition with a power system; 2 – system for wireless data transmission; 3 – sensor for measuring plantar pressure distribution; 4 – angular displacement sensor; 5 – feet mounting system; c) plantar pressure insoles.

2. Methods

2.1 Hardware and architecture of the wearable sensor system

The new wearable sensor system weighs 12 kg and consists of plantar pressure insoles with max. two hundred and forty sensors (T&T medilogic Medizintechnik, GmbH Munich, Germany) (Fig. 1c), and wearable sensors for joint angle measurement (Fig. 1a and b). Measurement of angular displacement was performed by means of six 12-bit (0–10V) resolution hall effect absolute encoder Mab28A (MegaMotive, Germany). Transducer signal from the encoders was processed using a DT9800 series measuring card manufactured by Data Translation [8].

The signal was collected and filtered by moving the average filter (at 60 Hz) using LabVIEW software (National Instruments, USA).

2.2 Estimation of ground reaction forces and joint moments

To estimate ground reaction forces and joint moments of the lower limbs, a 23 degree-of-freedom musculoskeletal model and the inverse dynamics technique were used [9]. The structure of the wearable sensor system, including its dimensions and rotation, the ground reaction forces and joint moments are presented in Fig. 2.

Figure 2.

The structure of the wearable sensor system with a) dimensions and rotation, b) ground reaction forces and joint moments.

The algorithm for estimating the ground reaction forces and joint moments during walking in the sagittal plane (Fig. 3) works in real time [8].

Figure 3.

The algorithm for estimating the ground reaction forces and joint moments during walking in the sagittal plane.

In the first step, angular displacement in joints was performed. Based on the data, a sixth voltage signal representing joint position for the left $\varphi_{1}^{l}$ , $\varphi_{2}^{l}$ , $\varphi{{}_{3}}^{l}$ and for the right $\varphi_{1}^{r}$ , $\varphi_{2}^{r}$ , $\varphi_{3}^{r}$ of the lower limb was collected. The ground reaction forces $F_{1}^{l}$ , $F_{2}^{l}$ , $F_{1}^{r}$ , $F_{2}^{r}$ were estimated in step II based on signals collected from 240 pressure sensors (Fig. 4).

Figure 4.

Position of sensors with the distance from the axis rotation.

To estimate ground reaction forces $F_{1}^{l}$ , $F_{2}^{l}$ , $F_{1}^{r}$ , $F_{2}^{r}$ , Eq. (1) was used:

$\displaystyle F_{1}=\frac{\sum\limits_{i=1}^{12}{G_{i}}w_{i}}{d_{4}},$ (1) $\displaystyle F_{2}=\frac{\sum\limits_{i=13}^{18}{G_{i}}w_{i}}{d_{5}},$

where: $w_{i}$ – distance from the axis of rotation to the center of the sensors, $G_{i}$ – sum of forces in $i$ line collected by the sensors.

To estimate the joint moment of the lower limbs in the step III, Eqs (2) and (2.2) were used:

$\displaystyle\left[\begin{array}[]{c}{M_{1}}\\ {M_{2}}\\ {M_{3}}\\ \end{array}\right]=\left[\begin{array}[]{cc}{Z_{11}}&{Z_{12}}\\ {Z_{21}}&{Z_{22}}\\ {Z_{31}}&{Z_{32}}\\ \end{array}\right][{F_{1}}\ \ {F_{2}}]$ (2)

where:

$\displaystyle Z_{11}=d_{4}+d_{1}\sin(\varphi_{2}+\varphi_{3})-d_{2}\sin(% \varphi_{3}),$ $\displaystyle Z_{12}=-d_{5}+d_{1}\sin(\varphi_{2}+\varphi_{3})-d_{2}\sin(% \varphi_{3}),$ $\displaystyle Z_{21}=d_{4}-d_{2}\sin(\varphi_{3}),$ $\displaystyle Z_{22}=-d_{5}-d_{2}\sin(\varphi_{3}),$ (3) $\displaystyle Z_{31}=d_{4},$ $\displaystyle Z_{32}=-d_{5}.$

The ground reaction forces and joint moments were normalized to the body mass.

2.3 Wearable sensor system validation

To validate the measurements made by the wearable sensor system and to examine the effectiveness of the proposed method for gait analysis, an experimental study on ten volunteers from Bialystok University of Technology, Poland, was conducted. The exclusion criteria were any other disorders that may impact the subject’s gait. All the subjects received full information about the study before giving their signed consent. The data was recorded at 60 Hz with the use of the new wearable sensor system. For data measurement, the subjects walked a distance of approximately 60 meters at their habitual speed outside of a gait laboratory. The accuracy of measurement of ground reaction forces and joint moments was validated against the results obtained from the reference motion capture system (SMART, BTS, Italy).

2.4 Statistical analysis

An ANOVA for repeated measurements was applied to gait variables to evaluate differences in quantitative walking parameters. The accuracy of estimation of ground reaction forces and joint moment was validated against the results obtained from the reference motion capture system by determining the root mean square error of these maximum values, and Spearman’s rank correlation. Statistical tests were performed using Statistica 13.1 (StatSoft, Poland). The significance level was set at $p<$ 0.05.

3. Results

Ten volunteers participating in this study: age 31.5 $\pm$ 2.8 years, height 178 $\pm$ 6.2 cm, body mass 63.4 $\pm$ 5.7 kg. Three moments were analyzed: ankle dorsiflexion/plantarflexion moment, knee flexion/extension moment, and hip flexion/extension moment. The representative plots showing the GRF and the three moments from a single step are shown for a control subject in Fig. 5.

Figure 5.

Representative plot for a control subject showing motion analysis lab and wearable sensor system results for a) ground reaction force, b) ankle d/p moment, c) hip fl/ex moment, and d) knee fl/ex moment.

Table 1 summarizes the root mean square error and the correlation coefficient for the ground reaction forces and joint moments obtained from the wearable sensor system, compared to the motion capture system.

4. Discussion

Recently, many researchers have been proposing new, more affordable, and low-cost, tools to measure gait parameters, for example, gyroscopes, accelerometers, instrumented insoles, etc. [6, 7]. One of the available systems of this type is offered by [7]. To estimate the ground reaction forces and joint moments, the authors developed a custom instrumented insole and a tissue force sensor. In [10], a method for the estimation of total ground reaction forces from pressure insoles during walking was presented. The results indicated a very strong correlation in the anterior-posterior and vertical directions, and a reasonable correlation in the medial-lateral direction. The results are in agreement with our results, where mean Spearman’s correlation coefficient was above 0.9 for all the subjects. The root mean square error was approx. 12%, 5% and 28% of the peak recorded value. The obtained results show that the error for min. GRF and knee extension moment was 1% and 27.3%, respectively. In paper [11], a comparison with data collected simultaneously from a clinical motion analysis laboratory demonstrated that insole results for ground reaction force and ankle moment were highly correlated for all the subjects. The results show that the best performance was achieved for GRF, which is unsurprising, since plantar pressure sensors were used in force measurement.

Table 1
The root mean square error and the correlation coefficient for the ground reaction forces and joint moments

Parameters	Wearable	Motion capture	Correlation	Error
	sensor system	system	coefficient (R)	[%]
GRF (max F) [N/kg] Mean $\pm$ SD	0.93 $\pm$ 0.15	1.12 $\pm$ 0.11	0.92	16.9
GRF (min F) [N/kg] Mean $\pm$ SD	0.76 $\pm$ 0.15	0.77 $\pm$ 0.11	0.96	1.0
Hip flexion/extension moment [Nm/kg] Mean $\pm$ SD	1.03 $\pm$ 0.18/ 0.80 $\pm$ 0.18	1.15 $\pm$ 0.25/ 1.09 $\pm$ 0.24	0.93	10.4/ 26.6
Knee flexion/extension moment [Nm/kg] Mean $\pm$ SD	1.06 $\pm$ 0.18/ 0.24 $\pm$ 0.11	1.16 $\pm$ 0.21/ 0.33 $\pm$ 0.16	0.94	8.6/ 27.3
Ankle dorsiflexion/ plantarflexion moment [Nm/kg] Mean $\pm$ SD	0.95 $\pm$ 0.08/ 0.15 $\pm$ 0.02	1.22 $\pm$ 0.25/ 0.22 $\pm$ 0.11	0.91	24.6/ 22.1

5. Conclusions

This study suggests that the wearable sensor system could be recommended for both research and clinical applications outside a typical gait laboratory. Future work should be performed in the field of the application of this system for rehabilitation purposes.

Footnotes

Acknowledgments

The work has been accomplished under research project No. S/WM/1/2016 financed by the Biaç¡‘stok University of Technology.

Conflict of interest

None to report.

References

Pauk

Minta-Bielecka

. Gait patterns classification based on cluster and bicluster analysis. Biocybernetics and Biomedical Engineering2016; 36(2): 391-396.

Winter

. The biomechanics of motor control of human gait. University of Waterloo Press 1987.

Michnik

Jurkojc

Pauk

. Identification of muscle forces during gait of children with foot disabilities. Mechanika2009; 6(80): 48-51.

Muro-de-la-Herran

Garcia-Zapirain

Mendez-Zorrilla

. Gait Analysis Methods: An Overview of Wearable and Non-Wearable Systems, Highlighting Clinical Applications. Sensors (Basel)2014; 14(2): 3362-3394.

Bae

Kong

Byl

Tomizuka

. A mobile gait monitoring system for gait analysis. IEEE International Conference on Rehabilitation Robotics, (Kyoto, Japan)2009; 50-57.

Liu

Inoue

. Measurement of human gait using a wearable system with force sensors and inertial sensors. Smart Sensors, Measurement and Instrumentation 15, Springer International Publishing, Switzerland 2015.

Liu

Inoue

Shibata

. A wearable ground reaction force sensor system and its application to the measurement of extrinsic gait variability. Sensors2010; 10: 10240-10255.

Pauk

Ostaszewski

Kuzmierowski

. Wearable Mechatronic Motion Capture System for Angle Measurement and its impact on foot loading during walking. Mechanika2015; 21(5): 414-418.

Jacobs

Ferris

. Estimation of ground reaction forces and ankle moment with multiple, low-cost sensors. J Neuroeng Rehabil.2015; 12: 90.

10.

Fong

Chan

Hong

Yung

Fung

Chan

. Estimating the complete ground reaction forces with pressure insoles in walking. J Biomech.2008; 41(11): 2597-601.

11.

Sobotach

Kecskemethy

Steinwender

Zwick

. A simplified approach for rough identification of muscle activation profiles via optimization and smooth profile patches, ECCOMAS, Madrid Spain 2005.

Estimation of ground reaction forces and joint moments on the basis on plantar pressure insoles and wearable sensors for joint angle measurement

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1. Introduction

2.1 Hardware and architecture of the wearable sensor system

2.2 Estimation of ground reaction forces and joint moments

2.4 Statistical analysis

3. Results

Table 1 The root mean square error and the correlation coefficient for the ground reaction forces and joint moments

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
The root mean square error and the correlation coefficient for the ground reaction forces and joint moments