A study on development of fuzzy algorithm based on knee joint anterior displacement for identifying gait intention

Abstract

We divided the gait intention into 3 parts of start, stop and changing walking velocity using the fuzzy algorithm. Decision for gait initiation was determined when the difference of distance was larger than 5 cm between the right and left knee joint anterior displacements. The gait stop was registered when the differences were less than 4 cm after 7 measurements in an 80 m/s. The velocity changed when the distance difference value was more than±10 cm from the mean value of the knee joint anterior displacement values on both sides. Results also showed that the knee joint anterior displacement was efficient as a measurement benchmark for detecting gait intention. Potential use of the algorithm might be expected for older adults.

Keywords

Rollator welfare EMG IR sensor fuzzy algorithm

Nomenclature

P_R, P_L = R, L Pressure

ΔP_R, ΔP_L = (Now – Before) Pressure

L_R, L_L = R, L distance

V_R, V_L = R, L Motor speed

ΔTL_R, ΔTL_L = (Now – Before) distance acceding to Motor speed

U_R, U_L = Output variables

F_e, ΔFⁱ_e, C_i = Functions that belongs to fuzzy set

DF : Deceleration Fast

DS : Deceleration Slow

ZE : Zero

AS : Acceleration Slow

AF : Acceleration Fast

1 Introduction

The main motor activity of a human body in daily life involving the human gait is a complex, dynamic and unforeseeable unpredictable process. It involves multiple joints movements, demanding sensory-motor integration and synchronization with the skeleton via a neurological system [11]. There is also the necessary continuous and repetitive motion to move a body frame forward while another body frame maintains a stable stance [10, 16]. Physical sense and movement ability, however, deteriorate over time due to aging. Such changes cause diverse problems in daily life. That is, they reduce physical activities and gait ability. Various aids for elderly people who have difficulty maintaining a normal gait that may include crutches, canes and walkers. These gait aids reduce pain and injury by mitigating the burden on lower extremity during a gait. That is, they help maintain a safe gait by increasing gait stability [1].

Rollator is another tool to assist users maintains their gait. A low-tech passive rollator is also evolving into a high-tech walk assist robot [15]. There have also been studies that have had a wheel-typed rollator leveraging robot technology [1, 8]. With such applications and with the advent of computer technology, intelligent algorithm has gained a lot of attention. The prominent examples of intelligent algorithm include fuzzy algorithm, neural network, k-NN(k Nearest Neighbor) and SVM(Support Vector Machine)[2, 13].

For the purposes of analysis, gait intention is defined as identifying the behavior of a pedestrian to start gait in from a stationary state, then change the gait with a gait velocity as a pedestrian. In addition, there is also the gait environment of a pedestrian during gait, and lastly, the intention of a pedestrian to stop the gait. In regard to the existing studies on gait intention, a majority of these studies have focused on gait intention in a linear place as to normal people by utilizing three-dimensional motion analysis and change in plantar pressure [5].

However, only a handful of studies have examined gait intention using electric gait assist [7, 12]. Concerning the domestic studies for identifying gait intention, the study of Lee et al. [4, 9] examined the method of judging gait velocity based on pressure intensity and by installing a pressure sensor on the handle of a rollator. From studies in other countries, the study of Martins et al. [14] examined the method of identifying gait intention by utilizing a joystick-type handle. However, such methods have several limitations, and one of these limitations is that pedestrians should have to adjust their gait velocity to according to their rollator.

This study aimed to propose a fuzzy algorithm to identify the gait intention of a pedestrian in relation to an electric rollator. The controller developed in this study was applied to an existing product, for which, the fuzzy algorithm was applied. The final purpose of the study was to validate the application of the algorithm. In the future, we intend to improve the proposed fuzzy algorithm for not only the aged people with normal gait but also those aged people with an abnormal gait.

2 Methods

2.1 Subjects

The study was performed in the Biomechanics and Rehabilitation Laboratory at Chonbuk National University. Conforming to the Declaration of Helsinki (1964), written informed consent was obtained from all subjects. Subject characteristics are 10 young subjects (10 man: mean age 23.7±0.5 years; mean weight 70.6±5.6 kg; mean height 170.4±4.3 cm), and 10 elderly subjects (8 man and 2 women: mean age 74.8±2.5 years; mean weight 61.6±3.6 kg; mean height 168.7±7.3 cm). All subjects were screened with a detailed medical history, physical activity questionnaire, electrocardiogram, and were not treated for any systemic disease.

2.2 System configuration

The electric rollator used in this study was mounted on a frame developed by the product and had a removable driving motor and driver. The input variable of fuzzy algorithm for identifying gait intention was the handle pressure value for knee joint displacement and gait assist for pedestrians. In the electric rollator as shown in Fig. 1(a), the distance sensor was mounted in order to measure changes in knee joint anterior displacement. Moreover, the pressure sensor was also installed on both handles of rollator in order to check whether the subjects held on to their rollator. The sensors used in this study included IR sensor (GP2Y0A02YK0F, SHARP Co., Japan) whose IR based measurement range was 150 cm at maximum and force sensitive resistor(FSR) sensor (A201-100, TECSCAN Co., USA) that could measure up to 444.8 N of force. FSR sensor on both handles of rollator was used as a safety device that was designed to drive the motor of rollator when measuring a signal above a certain size on both hands. Two IR sensors attached at the center of rollator measured both knee joint anterior displacements. The measurements of both knee joint distance, rollator handle pressure, gait assist velocity, etc. were designed to be transmitted wirelessly to the computer.

2.3 Experiment and analysis

The experiment was conducted with straight gait on round trip in a 10 M corridor. The gait protocol developed in this study was applied to the fuzzy algorithm developed to identify gait intention at the time when a pedestrian started the gait, conducted the gait and terminated the gait. The input variables applied to the fuzzy algorithm developed in this study were measured in real-time. They were also designed to control the electric rollator developed in thisstudy.

Figure 1(b) is the block diagram of fuzzy logic controller to identify the gait intention of the electric rollator user. The input variables included the gait velocity of the knee joint on both ends of a pedestrian and the pressure value of both hands on the rollator. The output variable was the motor velocity value of driving motor of electric rollator. The input and output variables of controller were expressed by the linguistic rule of “IF – THEN format”. Such rule serves effectively to identify the gait intention of rollator user. In this study, the control rules were set on the basis of the gait velocity and pattern of theelderly.

The control rule is divided into the following two fuzzy rules. The first fuzzy rule is about judging whether a pedestrian holds the handle of rollator and how large the pressure of holding the handle is. The second fuzzy rule is about the range rate between knee joints on both ends of a pedestrian, the time for a change to take place at the knee joints on both ends and the change in distance between the initial knee joint and the current knee joint. Controlling rule is the fuzzy rule with a total of “IF – THEN format”.

In regard to the fuzzy rule as shown in Fig. 2(a), a pedestrian proceeds with the maximum acceleration when the range rate of both knee joints and the range rate between both knee joints are at the maximum positive value. In contrast, a pedestrian proceeds with the maximum deceleration when they are at the maximum negative value. The rate of change within a certain range was designed to maintain the current velocity. To express the vagueness of linguistic variables used in the fuzzy rule as shown above within the range of 0 to 1 in this study, the range rate between the range rate of each knee joint and the range rate between both knee joints were expressed as an affiliation of the sets used in the designedcontroller.

In regard to the membership function, Gaussian and triangular shapes were used depending on each case as shown in Fig. 2(b). First, the triangular function was used as the membership function for the range rate of the knee joints of pedestrians (the range rate of an electric rollator and pedestrians), and used as an input variable to control the speed of an electric rollator. Second, the Gaussian function was used as the membership function for the range rate of both knee joints of pedestrians (the range rate of both legs of pedestrians) to identity the gait intention. The membership function of the fuzzy algorithm proposed in this study was not a uniform membership function but a non-uniform membership function, because it was reported to have a faster convergence speed of time and more repetition degrees required to reach the optimization than a uniform membership function.

In regard to the inference method, Mandani’s max-min inference method was utilized [6]. To convert the fuzzy inference value derived from the fuzzy set into a scalar value used as a process input, centroid method was utilized in the defuzzification method. As for the defuzzification of centroid method, the center was first obtained from the area of fuzzy set. Then, the point corresponding to the center was utilized as a control value.

In this study, the look-up table was stored and used inside the developed controller in order to process fuzzy logic in real time. In regard to the data measured in real time, noise was removed through IIR(Infinite Impulse Response) filter and low pass filter (50 Hz). Moreover, the relevant data was designed to be transmitted to the PC in real time in order to verify the process inside the developed controller and evaluate the validity of applied fuzzy algorithm.

In order to use the developed fuzzy algorithm, the measurement value of knee joint anterior displacement was found. In case of a gait start, the gait assist judged it as a gait start when there was a difference in the knee joint anterior displacement in a stationary state [3]. In regard to the judgment on gait termination, the rollator judged it as a gait termination when there was no change in the difference of knee joint anterior displacement on the left and right sides during gait for a certain period of time. Lastly, the velocity of rollator was designed to be adjusted if there was a change above a certain distance back and forth on the basis of the mean value of the knee joint anterior displacement values on both sides when the motor of rollator was turned on in relation to the change in the gait environment. The recognition distance for judging the start and termination of gait was adjusted in the units of 1 cm from 2 cm to 8 cm. In regard to the judgment intention of gait environment, the range was adjusted in the units of 5 cm from±20 cm to±5 cm in order to find the optimal recognition value. The rollator was also designed to stop its operation for the safety of a pedestrian if the value of either of the pressure sensors installed on the left and right of the handle of gait assist fell below 3 kg during gait.

3 Results and discussion

In this study, the control algorithm was developed to control a drive by identifying the gait intention of a pedestrian with the electric rollator. It was confirmed that the proposed fuzzy algorithm was a very effective algorithm responding to drivers input and output variables. Also, the optimal knee joint anterior displacement value was found in the gait intention algorithm developed in this study. For the study, the results, as listed in Table 1, were obtained.

To identify the intention of gait start, the start distance was measured from 2 cm to 8 cm. As a result, the distance difference value was too short at 2 cm, 3 cm and 4 cm. The rollator was operated even with a slight movement when the rollator judged it as a gait termination. As for the value of 6 cm or above, there was no problem with general gait. However, there was a difficulty in recognition when the elderly people walked slowly with short and quick steps. To remedy this, the optimal distance for identifying the intention of gait start was set at 5 cm.

In regard to the analysis on the intention of gait termination, the highest accuracy was achieved when the knee joint distance was recognized 5 times at 80 m/s with 4 cm or below during gait. The recognition distance is short when the knee joint distance is less than 4 cm. As a result, it becomes impossible to stop during a gait. In contrast, the recognition distance becomes wider with a larger distance. As a result, the gait assists stop while starting gait.

If the recognition time is faster than 80 ms and the frequency less than 7 times, the rollator will stop while starting gait. If the recognition time is slower than 80 m/s and the frequency more than 7 times, the rollator will continue to operate ceaselessly. astly, this study examined the change in the velocity by judging the distance between the gait assists and the pedestrian during gait. As a result, the initial distance for holding both hands was set as a reference value. The velocity of rollator was changed based on the judgment for the distance between the pedestrian and the rollator. The experiment was conducted in the units of 5 cm from±20 cm to±5 cm based on the reference value. As a result of the experiment, there was no change in the velocity with±20 cm and±15 cm because the distance was too far. In the case of±5 cm, the width was too narrow. The velocity of rollator changed too frequently. Hence, the optimal reference width was set at±10 cm. The algorithm for judging gait intention was developed in accordance with the set values.

Figure 3(a) depicts the measurement values for the distance of knee joint based on the fuzzy algorithm. When the right or left foot enters inside the rollator after holding the rollator in a stationary state, the rollator recognizes the knee joint anterior displacement distance. As a result, the rollator starts operating accordingly. If the knee joint anterior displacement value is kept constant during gait, then the gait will be terminated. In regard to the intention of judging the initiation and termination of gait, the results as shown in Fig. 3(b) were obtained.

In the case of gait initiation, the rollator was designed to start if there was a distance difference of 5 cm or more for the knee joint anterior displacement in a stationary state as the rollator judged it as intention to initiate gait. In contrast, the rollator was designed to suspend its operation when there was not a difference of 5 cm or more for the knee joint anterior displacement. In the case of judgment of gaittermination, the rollator judged it as gait termination if the knee joint anterior displacement difference on the left and right during gait was less than 4 cm and measured 7 times every 80 ms. However, the values would be reset when at least one of the 7 measurements deviated from 4 cm. In such case, the rollator judges that the gait is in progress. Therefore, the rollator continues its motion.

Figure 4(a) is the case of gait intention analysis depending on the gait environment. The rollator judges that the distance between itself and a pedestrian becomes close when the knee joint base distance became within 10 cm when the gait starts. As a result, the velocity at the initial phase of gait increases. The rollator judges that the distance between itself and a pedestrian becomes far when the distance is farther than 10 cm. As a result, the rollator decelerates. When entering inside the base line, the velocity returns to the normal velocity. From this figure, it can be confirmed that the proposed fuzzy controller is adequately controlled.

Figure (4b) shows the result of On/Off system for the motor of rollator thorough the pressure sensor installed on the handle. To operate the rollator, pressure of 3 kg or higher should be set at the pressure sensor of both handles of the rollator. If either the right (R) or the left (L) handle pressure sensor of the gait assists falls below 3 kg during gait, the rollator judges that a pedestrian no longer uses the rollator. As a result, it stops its operation.

4 Conclusion

In any article it is unnecessary to have an arrangement statement at the beginning (or end) of every (sub-) section. Rather, a single overall arrangement statement about the whole paper can be made at the end of the Introduction section. The purpose of the study was to perceive pedestrians’ gait intentions only through the range changes of their knee joints. The gait intentions of pedestrians are an extremely uncertain variable to predict. A fuzzy algorithm was thus suggested to respond to such an uncertain and unpredictable situation. To identify the gait intention of a pedestrian as in this study, the pedestrian’s mental and physical states need to be predicted, which requires the use of a very uncertain variable as a controllable variable. The fuzzy algorithm is known to be very useful in inducing controllable variables in such uncertain situations. To utilize such advantage of the fuzzy algorithm, this study adopted the fuzzy algorithm.

This study showed that the designed fuzzy algorithm was very effective in controlling the velocity of an electric rollator by identifying the gait intention. In regard to judging gait initiation intention, the rollator decides that it is a gait start when there is a change of 5 cm or higher for the knee joint on the left and right in a stationary state. In regard to the judgment of gait termination intention, the rollator judges it as a gait termination when there is a difference of 4 cm or less in the knee joint anterior displacement on the left and right with 7 measurements every 80 ms during gait. Lastly, in regard to the judgment of gait environment intention, the velocity of rollator was designed to vary when there was a change of±10 cm or more in the base line during gait. The input variables of the fuzzy algorithm proposed in this study were sufficient for identifying the gait intention based on the changes in the knee joint anterior displacement.

However, it was confirmed that the proposed fuzzy algorithm was a very effective for identifying the gait intention of an electric rollator if the problem associated with sudden braking resulting from the error from IR sensor distance measurement was solved. In a future study, we intend to improve the proposed fuzzy algorithm for not only the aged people with normal gait but also those aged people with an abnormalgait.

Footnotes

Acknowledgments

This work was supported by Ministry of Trade, Industry & Energy (MOTIE) (QoLT Technology Development, No. 10048001) and supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning (No. 2014R1A1A1006266).

References

Lopes

A.C.

, Pires

and Nunes

, Assisted navigation for a brain-actuated intelligent wheelchair, Robotics and Autonomous Systems61 (2013), 245–258.

Graf

, Hans

and Schraft

, Care-O-bot II - Development of a next generation robotic home assistant, Autonomous Robots16 (2004), 193–205.

C.H.

, Kang

S.R.

, Yang

G.T.

, Hong

C.U.

, Lee

H.J.

and Oh

D.Y.

, Preliminary study on verifying the detection of gait intention based on knee joint anterior displacement of gait slopes, Bio-Medical Materials and Engineering26 (2015), 583–592.

Lee

D.K.

, Kong

J.S.

, Goh

M.S.

, Lee

S.M.

, Kang

S.J.

and Lee

E.H.

, Design of the control algorithm for improvement of the convenience the active-type walking aid, Journal of Rehabilitation Welfare Engineering & Assistive Technology5 (2011), 17–25.

Wentinka

E.C.

, Schuta

V.G.H.

, Prinsenc

E.C.

, Rietmanb

J.S.

and Veltinka

P.H.

, Detection of the onset of gait initiation using kinematic sensors and EMG in transfemoral amputees, Gait & Posture39 (2014), 391–396.

Mandani

E.H.

, Fuzzy Reasoning Its Applications, Academic Press, 1981, pp. 310–323.

Lee

G.H.

, Ohnuma

, Nak

Y.C.

and Lee

S.G.

, Walking intent-based movement control for JAIST active robotic walker, IEEE Transactions on Systems, Man, and Cybernetics: Systems44 (2014), 665–672.

Lacey

and Dawson-Howe

K.M.

, The application of robotics to a mobility aid for the elderly blind, Intelligent Robotics Systems23 (1998), 245–252.

Kong

J.S.

, Lee

D.K.

, Nam

Y.S.

, Lee

B.H.

and Lee

E.H.

, Optimized walking will recognizing system of the walking aid with the fuzzy algorithm, Journal of Korean Institute of Intelligent Systems18 (2008), 692–699.

10.

Wilson

, Effects of walking pole on lower extermity gate mechanics, Medicine & Science in Sports & Exercise33 (1993), 142–147.

11.

Laura

M.D.

, Adrian

and Doru

, Gait intention analysis for controlling virtual reality walking platforms, Technological Innovation for Sustainability349 (2011), 131–138.

12.

Park

M.J.

and Kim

J.Y.

, Development of intelligent walking assistive robot using stereo cameras, Transactions of the Korean Society of Mechanical Engineers A38 (2014), 837–848.

13.

Martins

M.M.

, Santos

C.P.

, Frizera-Neto

and Ramón

, Assistive mobility devices focusing on smart walkers: Classification and review, Robotics and Autonomous Systems60 (2012), 548–562.

14.

Martins

, Santos

, Frizera

and Ceres

, Real time control of the ASBGo walker through a physical human–robot interface, Measurement48 (2014), 77–86.

15.

Chuy

, Hirata

and Kosuge

, A new control approach for a robotic walking support system in adapting user characteristics, IEEE Systems, Man, and Cybernetics Society36 (2006), 725–733.

16.

Cromwell

R.L.

, Newton

R.A.

and Forrest

, Influence of vision on head stabilization strategies in older adults during walking, Journals of Gerontology. Series A: Biological Sciences and Medical Sciences57 (2002), 442–448.