You are on page 1of 5

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 6, November December 2013 ISSN 2278-6856

Classification of Gait- Phases for FES using Neuro-Fuzzy approach


Swaroop R1, Pradip Kumar Sadhu2
1 Amity University-Dubai Campus, Department of Electronics and Telecommunication, DIAC, UAE 2

Department of Electrical Engineering, Indian School of Mines Dhanbad, India.

Abstract: Paraplegic is due to the paralysis of the lower


part of the body due to injury in spinal cord which affects the normal activity like standing, sitting and walking. This lack of physical movements would result in various other diseases. To improve the physical activity FES is used to give various movements to the muscles. The control strategy is usually difficult due to the complex and nonlinear model of muscle and leg movements. FES rehabilitative system stimulation is normally done based on movements in human muscle. In paraplegic patients with upper motor neuron lesions the signal path from the central nervous system to the muscles is interrupted. Functional electrical stimulation applied to the lower motor neurons can replace the lacking signals. Various researches have been carried out on FES to restore the motor functions. This proposed paper is to develop a Fuzzy-Neuro model to find the various phases of gait and to give the corresponding signals. In this project Fuzzy Logic is used only to differentiate the three activities Swing, Stance and Transformation(Sitting-standing). Fuzzy logic also categorize the stance, swing in to three groups. It also gives the information on the force transformation carried out during sitting and standing. This knowledge is fed to the neural network model where it gives signal to the electrodes depending on the output at each phase.

across the foot as shown in the block diagram (Refer Fig4). Here eight strain gauges are connected diagonally and the summation is used to calibrate the forces.

2. LITERATURE REVIEW
The application of electrical stimulation in a rehabilitative setting was initiated nearly half a century ago. In 1960, W.T. Liberson, a physical rehab specialist and medical researcher, began developing a heel switchtriggered personal electronic stimulator device to correct foot drop (weakness or paralysis of the muscles involved in lifting the front part of the foot). Liberson provided the basic groundwork for future rehabilitative techniques involving functional electrical stimulation (FES). His work has led to the present explosion of FES-related research engaging countless scientists, propelling it to a legitimate and viable treatment option for many patients with spinal cord injuries and disorders. FES utilizes electrical currents to activate nerves in areas of a patients body (i.e., arm and leg muscles) affected by paralysis, stroke, traumatic brain injuries, and other neurological disorders to restore some movement and function. Most FES devices have similar componentselectronic stimulator, a feedback or control unit, leads/sensors, and electrodes- and can range in size from as small as a calculator to as large as a computer workstation. Perhaps the best-known FES system is the pacemaker for the heart. ES treatment is accomplished by placing electrodes on-or surgically implanting them under-the patients skin close to peripheral nerves that control specific muscles or muscle groups. [13] The electrodes are connected by leads to an FES device that generates low-voltage electrical impulses. These impulses excite the nerves causing the paralyzed muscles to contract, thereby facilitating basic movement. In some FES applications, tissues instead of muscles may be targeted. There are number of technologies incorporated for the functional electrical stimulation to analyze the gait phases and give signal. Here neuro-fuzzy techniques is implemented to solve the complexity involved.

Keywords: Gait Phase, Neuro-Fuzzy, FES,

1. INTRODUCTION
There are already various models developed in neural network to differentiate the phases and give signals which may prone to errors due to different types of signals applied to electrodes in different phases. Most of the research papers concentrate either on the phases of walking or standing-sitting. In this project Fuzzy Logic is used only to differentiate the three activities Swing, Stance and Transformation (Sitting-standing). Fuzzy logic also categorizes the stance, swing in to three groups. It also gives the information on the force transformation carried out during sitting and standing. This knowledge is fed to the neural network model where it gives signal to the electrodes depending on the output at each phase. This minimizes the error. Another new approach in this proposed project is the LVDT used to find the knee angle at various stages where goniometer and accelerometers are commonly used to get the knee angle. [1] Here strain gauge is used to identify the forces applied at each stance phase. The strain gauge here is connected diagonally Volume 2, Issue 6 November December 2013

3. GAIT CYCLE
The gait cycle of each leg is divided into the stance phase and the swing phase. The stance phase is the period of time during which the foot is in contact with the ground. Page 275

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 6, November December 2013 ISSN 2278-6856
The swing phase is the period of time in which the foot is off the ground and swinging forward. In walking, the stance phase comprises approximately 60% of the gait cycle and the swing phase about 40%.[3] The proportion of swing to stance phase changes as the speed of walking increases.This image summarizes the stance and swing phases of gait and the interrelationships between both limbs. wheatstone bridges, amplified, filtered for noises and could be transmitted through wireless transmitter to PC.

Figure 2: Block diagram on the Hardware part. Figure 1: Various phases of Gait 3.1. Stance Phase Contact Subphase The stance phase can further be subdivided into its three component phases. The first portion of the stance phase is contact.[2] This phase begins with with the contact of the heel to the ground. This phase is completed when the remainder of the foot touches the ground. 3.2 Stance Phase: Midstance Subphase Midstance Phase begins when the entire foot has contacted the ground. The body weight is passing over the foot as the tibia and the rest of the body are moving forward. The opposite leg is off the ground and the foot, in this phase, is bearing body weight alone. 3.3 Stance Phase: Propulsion Subphase Propulsion begins after heel off and ends with toe off. This phase constitutes the final 35% of stance phase. The body is propelled forward during this phase as weight is shifted to the opposite foot as it makes ground contact. 3.4.Swing Phase The swing phase begins immediately after toe off. The first portion of the swing phase is the forward swing which occurs as the foot is being carried forward. The knee is flexed and the foot is dorsiflexed at this time. The next segment of the swing phase is foot descent as the foot is being positioned in preparation for weight bearing and the muscles are stabilizing the body to absorb the shock of heel contact. [4] At heel contact the swing phase ends and a new gait cycle begins. The project focuses on finding the forces beneath the foot which we cannot visually see. It also measures the timing of the phases and subphases of gait that we have reviewed and greatly increases our knowledge of what is occurring during the gait cycle. Normal pressure flow through the foot starts slightly lateral in the heel and flows forward to between the first and second metarsal and exits through the great toe.[5] The strain gauges are placed in the following fashion as shown in the figure 2. The summation of all the strain gauges are taken through Volume 2, Issue 6 November December 2013 Stress is measured as Force per area (in Pascals) while strain is the amount of deformation of the material or change in length (l /l) (or change in length/original length). Strain energy is the type of energy stored in tendons. The formula that describes this energy is SE = (stress2/E) SE= Strain Energy E=Youngs modulous of tendon (fibrous part connecting muscle and bone) Strain Energy is measured in Joules/meter3 . The resilience of a material, in this case a tendon, is defined as the amount of energy that can be absorbed by the material without causing a permanent deformation or damage to the material. Thus some of the work done by the muscle when contracting, is actually done by the stored energy in the tendon, in which stored energy is returned to the locomotion system. Potential energy and kinetic energy are considered to be out of phase in walking and in phase in running. During the airborne phase energy is at a constant (with minimal air resistance), after contact the energy rapidly decreases, then potential energy is again stored as the body moves forward over the limbs and the hip and leg are in positions of tension.

4. MEASUREMENT OF KNEE ANGLE


The block diagram above also shows an novel approach to find the knee angle at various phases. Goniometer or accelerometer is commonly used for this purpose. Here LVDT is connected between the two ends of knee as shown in the figure 2. When a knee bends the piston moves in and the amplitude of voltage decreases and while standing the core moves away from the core and the amplitude of the output voltage increases. [6,7] The change in voltage is calibrated for the knee angle. The voltage is then amplified, converted to DC using rectifier and then fed to Analog to Digital converter. The digital output is then fed to microcontroller before transmitting to PC. The PC receives both the signal where it is fed to Fuzzy Neuro module. The Neuro Fuzzy Module separates the various phases of gait cycle and then send signal to Neural Network model. The neural network Page 276

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 6, November December 2013 ISSN 2278-6856
decides signal to be send to electrodes. The output from the gait cycle for a normal person is shown below.

5. MUSCLES TO BE ACTIVATED AT EACH PHASE


The flexor muscles are more active in swing and extensors are more active during the stance phase. The activation patterns of individual muscles are more elaborate. [7, 8] Some muscles are activated twice per cycle and some only a portion per phase. The pattern of activation is shown in Fig3

Figure 4: Block diagram for the software to be developed for identifying the phases and the output to the electrodes THE RULE BASE The table I below shows the rule base for various phases. Zero input at the strain gauge indicates that the leg is floating and could come under the swing stage. The knee angle directly measures the various phases of swing which is categorized as small medium and large.[11] The zero input at the knee angle indicates that the foot is on the ground and would come under stance phase and the strain gauge output directly measures different phase of stance which is contact (stance small), mid stance (stance ) and the propulsion.(Stance large) TABLE I. Rule Base

Figure 3: Muscle to be activated at each phase.

6. BLOCK DIAGRAM
The Block diagram(refer Fig4) for the proposed Fuzzy Neuro model is shown in Fig4. Here the Fuzzy Logic is used for classifying the phases of gait. Two inputs are given to Fuzzy Logic Controller. One is Input from the strain gauge and the other the Knee angle from the capacitive sensors. The Fuzzy classifies the various stages of input stance phase, swing phase, and the transformation between standing and sitting. The fuzzy sends the phase along with the output for each phase in terms of percentage output (example: Swing Phase output = 60%). This output is fed as input to the neural networks. [9]The neural network sends the signals to each electrode which corresponds to each phase and the output. (Refer Fig 1 for the signal to be send for each phase. The input from strain gauge is divided in to four ranges [ zero, Strain gauge Low(SG-L) Strain gauge Medium (SG-M), Strain gauge High (SG-H). [9] Similarly the another input knee angle is also divided to four ranges Knee Angle Low, Knee Angle Medium (KA-M) and Knee Angle High (KA-H) and Zero (Z).This is known as fuzzification. (Ref Fig 6 and 7) The rule base as shown in the table I below categorizes the swing and the stance phase. The Defuzzification block has four outputs ie Stance Phase, Swing Phase, transformation phase. The defuzzified output gives the phase as well as the % output in each phase (Refer 8 and 9). The transformation phase gives the transformation between sitting and standing. The output is 1 when sitting, 0 when standing and between 0 ans 1 when transformation. The output of each phase is given to the Neural network to give signal to the electrodes. [10] The signal amplitude and the frequency are chosen by the neural network model from the trained set of data. Volume 2, Issue 6 November December 2013

7. FUZZIFICATION AND DEFUZZIFICATION


The diagram below shows the membership function created for the input and output variables created for the Fuzzy Logic. [12] The Fig 4 shows the block diagram and the fig 5 and 6 shows the fuzzification of inputs. The figure 5 shows that there are two inputs and three outputs connected to the fuzzy logic systems. The figure 6 shows that the strain gauge input is divided in to three categories as small, medium and high. Similarly the figure 7 shows that the kneeangle input is divided in to three categories as contact, midstance and propulsion. The fig 7 shows the defuzzification of stance phase to small, medium and high. The figure 9 shows the Page 277

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 6, November December 2013 ISSN 2278-6856
defuzzification of swing phase and the figure 10 shows the defuzzification of the transformation stage as a Gaussian function.

Figure 5: Fuzzy model developed in Matlab to identify the phases

Fig10: Defuzzification for the standing and sitting phases.

Fig 6: The classifying the various ranges of strain gauge (Fuzzification)

Figure:11: sample input and outputs given to the fuzzy controller

8. CONCLUSION
The above project was simulated in MATLAB software by applying random signals for strain gauge and knee angle. It can be seen from the figure 11 that the fuzzy controller coule receive the information from the two sensors straingauge and LVDT and could give the defuzzified result. It can also be observed that for an input of 0.1 and 0.2 the output is 0.223. It has been found that the neuro-fuzzy system could defuzzify the various phases of the system like swing, stance and the transition of sitting-standing postures and the neural network was implemented to send the corresponding signals to each of the flexors and extensors.

Figure 7: The classifying the various ranges of Knee angle (Fuzzification)

Reference
[1] Daisuke Nishikawa, Wenwei Yu, Hiroshi Yokoi, and Yukinori Kakazu On-Line Learning Method for EMG Prosthetic Hand Control Electronics and Communications in Japan, Part 3, Vol. 84, No. 10, 2001 [2] Daisuke Nishikawa, Wenwei Yu, Masaharu Marushi, Ichiro watanbe, Hiroshi Yokoi, Yukio Mano, and Yukinori Kakazu Online Learning Based Electromyogram to Forearm Motion Classifier with Motor Skill Eveluation JSME International Journal, Series C, Vol 43, No.4, 2000 [3] Saso Jezernik*, Ruben G. V. Wassink, and Thierry Keller Sliding Mode Closed-Loop Control of FES: Controlling the Shank Movement IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 51, NO. 2, FEBRUARY 2004 263 Page 278

Figure 8: Classifying the various stages of output Stance (Defuzzification)

Fig 9: Classifying the various stages of output Swing (Defuzzification) Volume 2, Issue 6 November December 2013

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)


Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 2, Issue 6, November December 2013 ISSN 2278-6856
[4] K. Y. Tong and M. H. Granat Reliability of neuralnetwork functional electrical stimulation gait-control system Medical and Biological Engineering and Computing Volume 37, Number 5, 633-638, DOI: 10.1007/BF02513359 [5] R. Riener and T. Fuhr, Patient-driven control of FES supported standing-up: a simulation study, IEEE Trans. Rehab. Eng., vol. 6, pp.113124, June 1998. [6] R. Riener, M. Ferrarin, E. E. Pavan, and C. A. Frigo, Patient-driven control of FES-supported standing up and sitting down: Experimental results, IEEE Trans. Rehab. Eng., vol. 8, pp. 523529, Dec. 2000. [7] T. Fuhr and G. Schmidt, A supervisory control approach to restore locomotion in paraplegics with WALK!, in Proc. 7th Ann. IFESS Conf., Ljubljana, Slovenia, June 2002, pp. 310312. [8] Nammond Yaikwawongs MD, Noppachart Limpaphayom MD, Vajara Wilairatana MD Reliability of Digital Compass Goniometer in Knee Joint Range of Motion Measurement J Med Assoc Thai 2009; 92 (4): 517-22 [9] Deanna J. Fish, CPO Jean-Paul Nielsen, CP Clinical Assessment of Human Gait Journal of Prosthetics and Orthotics 1993 Vol. 5, Num. 2 [10] Stacy J. Morris Bamberg, Ari Y. Benbasat, Donna Moxley David E. Krebs, and Joseph A. Paradiso, Gait Analysis Using a Shoe-Integrated Wireless Sensor System IEEE Transactions on Information Technology in Biomedicine. VOL. 12, NO. 4, JULY 2008 [11] Yu-Luen Chen1, Shih-Ching Chen4 , Weoi-Luen Chen1 , Chin-Chih Hsiao, Te-Son Kuo and Jin-Shin Lai Neural network and fuzzy control in FES-assisted locomotion for the hemiplegic 2004, Vol. 28, No. 1 , Pages 32-38 (doi:10.1080/03091900310001211523) [12] Griffin TM, Roberts TJ, Kram R. Metabolic cost of generating muscular force in human walking. J Appl Physiol 95: 172-183, 2003; 10.1152. [13] Functional Electrical Stimulation: The Future of Rehabilitation? By Tom Scott. www.unitedspinal.org.

Volume 2, Issue 6 November December 2013

Page 279

You might also like