|Year : 2019 | Volume
| Issue : 3 | Page : 240-246
|A mathematical method for electromyography analysis of muscle functions during yogasana
V Devaraju1, Ashitha Besagarahalli Ramesh2, K Kshamith Alva2, V Ramesh Debur3, SN Omkar2
1 Deparment of Mechanical, National Institute of Technology, Surathkal, Karnataka, India
2 Deparment of Aerospace, Indian Institute of Science, Bengaluru, Karnataka, India
3 Department of Physiotherapy, M.S. Ramaiah Medical College, Bengaluru, Karnataka, India
Click here for correspondence address and email
|Date of Web Publication||3-Sep-2019|
| Abstract|| |
Context: For the past few decades, the number of people practicing yoga is increasing in number. Yogasanas need smooth body movements in the process of attaining defined postures that the person must hold on to activate specific muscles of the body related to that asana. Yogasanas should be performed with perfection to derive maximum benefits. Objective: The objective of this study was to introduce a mathematical method to understand muscle functionalities while doing Yogasanas. Materials and Methods: Used Delsys surface electromyography (sEMG) – TrignoTM (Delsys Inc.) sensors for data recording and analyzing muscle activation patterns. Results: Performance analysis was quantified using normalized sEMG signals. The sEMG data during final posture were fit to a straight line using linear regression analysis. Conclusion: The results suggested that the slope of the best fit line is a good metric for monitoring the muscle activity during Yoga performance. The advantages of this method are the slope of the line is a good indicator for monitoring the muscle activity while doing Yogasana and the method suggested in this study can be extended for analyzing other asanas as well.
Keywords: Abdominal obliques, accelerometer, inertial measurement unit, Linear Least Square Fitting, muscle activation, sternocleidomastoid, surface electromyography
|How to cite this article:|
Devaraju V, Ramesh AB, Alva K K, Debur V R, Omkar S N. A mathematical method for electromyography analysis of muscle functions during yogasana. Int J Yoga 2019;12:240-6
|How to cite this URL:|
Devaraju V, Ramesh AB, Alva K K, Debur V R, Omkar S N. A mathematical method for electromyography analysis of muscle functions during yogasana. Int J Yoga [serial online] 2019 [cited 2023 Mar 20];12:240-6. Available from: https://www.ijoy.org.in/text.asp?2019/12/3/240/265741
| Introduction|| |
Yoga is known to have immense physical and mental health benefits. It is important to understand muscle activation patterns and performance in terms of the kinematics of the body and duration of stay in final posture during Yogasanas. Standardization and quantification of Yogasanas are important as different schools of Yoga have developed its own guidelines. Moving the body in various positions and orientations needs the support of muscles. Substantial muscle activity can be observed during muscle contraction or relaxation, which can be used for monitoring the correctness of Yogasana in terms of stability and steadiness.
In the field of Yogasana analysis, people have done significant research on performance analysis by obtaining biomechanical signals using Inertial Measurement Units, a combination of accelerometers (ACCs) and gyroscopes, mounted on various parts of the human body. Most of the signals obtained through body motion analysis have been shown to be both nonlinear and nonstationary, but few efforts have been made in the past to quantify the performance in terms of muscle behavior. The surface electromyography (sEMG) analysis methods suggested for other exercises cannot be used to analyze the Yogasana because, in most asanas, the final posture involves no motion at all, i.e., once the final posture is attained, the body is expected to be a steady and stable state. Thus, the need for a method to quantify the performance of Yoga in terms of muscle activity has been identified.
While performing Yogasana, the body should be in correct posture and it can be measured with ACC or gyroscope-based sensors. At present, ACCs are the most commonly used sensors because of their versatility, portability, and ease of use. ACC sensors provide kinematic information based on measurement of acceleration. ACC signal features are used to distinguish between relative motion of body parts for each transformation pose and to analyze the smoothness with which the exercise was performed.
ACCs are effective in differentiating between static activities such as sitting and standing from remarkably dynamic ones such as running, jumping, and dancing. However, ACC sensors have inherent limitations in differentiating between an active versus passive performance of a movement. On the other hand, sEMG sensors are sensitive to muscular activity and thus help in differentiating an active movement from a passive one. The technique of sEMG study in the field of sports and ergonomics is gaining in popularity. EMG is currently the most practical way of assessing the extent to which a muscle is working (although it has some limitations).
In this study, a yogic exercise named Trikonasana [Figure 1] was chosen for comparative analysis based on sEMG and ACC signals recorded from an expert. The expert chosen as the reference has been practising Yoga since 3 decades. The repeatability of the result was checked by taking sEMG signal data three times in a month having 10 days' interval for 3 trials every time.
The purpose of this study was to identify a mathematical method to analyze the muscle performance during Yoga. For validating the method suggested, Trikonasana was considered as it does not require a lot of expertise to perform. People with limited or no prior Yoga practice can perform this asana without much difficulty.
| Materials and Methods|| |
The study was performed in Biomechanics lab, Indian Institute of Science (IISc), Bangalore. Participants of this study were 31 healthy untrained volunteers (22 males and 9 females) and 1 trained expert. Individuals with a history of back pain, shoulder, or neck injuries were excluded from this study. Written consent from all 32 participants was taken as approved by Human Ethics Committee, IISc. [Table 1] lists the details of the participants.
The primary muscles involved while doing this asana are sternocleidomastoid (SCM), abdominal oblique, latissimus dorsi, quadriceps femoris, and medial hamstring. To obtain quality sEMG signals, skin preparation is essential as mentioned in [Figure 2]. Sensors were placed at 4 anatomical centers as shown in [Figure 3], corresponding to SCM and abdominal oblique major muscles to be constantly active. The sensors were placed so that the electrodes ran parallel to the muscle fibers.
|Figure 2: The sequence of steps followed while electromyography data acquisition|
Click here to view
|Figure 3: Schematic representation of the location of surface electromyography and accelerometer sensors, where X and Y represent the axis direction of the accelerometer. Z axis is into the plane of the paper|
Click here to view
Delsys EMG Trigno™ Wireless Systems equipment was used for the data collection with Smart Sensors connected to a PC running EMGworks® 4.3.2 software (DELSYS). The material of electrodes is made up of 99.9% silver. Sampling rates of the EMG and the ACC were 2000 Hz and 148.1 Hz, respectively. Sensors should be placed on correct muscle group to avoid cross talk. Before placing the sensors, the skin area was cleaned with 70% v/v isopropyl alcohol, and if necessary, the area was also shaved to minimize skin impedance.
Methodology flow chart
The steps involved while performing Trikonasana are as follows [Figure 4]: stand upright and maintain 3–4 ft between the feet. Raise both hands to shoulder level, with palms facing the ground. Parallelly turn right foot 90° outward and left foot to 45° in. Bend upper body toward the right side without bending the back forward or backward and rest the right hand on shin/ankle. Turn head gently and keep gazing at the tip of the left palm. Maintain the position for at least 30 s while breathing normally. Repeat the same procedure on the left side. In this study, Trikonasana was first performed to the right side of the body and then to the left side. The neck is turned to the left during the right side posture and vice versa [Figure 1].
Before conducting the experiments, each individual was given clear instructions about Trikonasana and a video clip of expert performing the asana was shown to all participants. They were asked to practice the asana until they were comfortable with the asana, and necessary feedback was given. Each participant was made to perform 3 trails, and the data were recorded. The data were recorded for approximately 90 s, and individuals held their final position for 20 s.
The behavior of SCM and abdominal oblique muscles was analyzed with sEMG and ACC data. Initially, raw sEMG signals were filtered with sixth-order bandpass Butterworth filter (cutoff frequency 20–150 Hz) to attenuate noises, followed by calculation of root mean square (RMS)., RMS value of sEMG signal represents overall muscle activity and muscular tension (provided muscle is not fatigued). The amplitude of sEMG signal was approximately proportional to force produced by the underlying muscle group, and it is used to compare the muscle effort applied by individuals to perform Trikonasana.
For comparing sEMG signals among different individuals, the signal should be normalized. sEMG signals were normalized with respect to different parameters such as height, weight, limb girth, and neck girth of the person. Most suitable parameter on which the data to be normalized was chosen by considering the relative shift of plots from zero references. From the parameters mentioned above, the weight of the person gave good results, as the shift of graph from zero was within 10% for all the individuals. sEMG signal for Trikonasana performed by an expert is as shown in [Figure 5].
|Figure 5: (a) Example of raw surface electromyography and root mean square signals of the left sternocleidomastoid obtained for expert performed on the right. (b) Example of raw surface electromyography and root mean square signals right sternocleidomastoid obtained for expert performed on the left. (c) Example of raw surface electromyography and root mean square signals of the left abdominal obliques obtained for expert performed on the right. (d) Example of raw surface electromyography and root mean square signals of the right abdominal obliques obtained for expert performed on the left|
Click here to view
For analysis, the data recorded for 20 sec hold time during final posture was fit to a line using Linear Regression analysis with Statistics and Machine Learning Toolbox in MATLAB. This analysis uses the linear least square fitting method to find the best fitting line for a given set of data points by minimizing the sum of squares of the offsets, i.e., minimize (R2) refer equation (1).
For a sample data containing n data pairs (xi, yi), i = 1, 2,…. n and model function f (xi, a0, a1), the squared residual (R2) is given by,
The equation of straight line in x-y plane for linear fit is given by,
Where, a1= slope of the line (regression coefficient), a0= y-intercept. y-axis data are that of sEMG signal and x-axis contains corresponding time [Figure 5].
The slope (a1) of the line directly corresponds to variations in amplitude while performing Trikonasana, through which muscle firing patterns can be analyzed. Ideally, the expected value of the slope is zero, which means that amplitude of sEMG signal is constant [Figure 6]. Constant amplitude suggests that individuals maintained the asana in steady and stable state because these are the main characteristics of correct posture. If the slope is >0, it indicates that the amplitude of muscle firing is increasing with time, if it is −1 or <0, it indicates that the amplitude of muscle firing is decreasing with time, and if it is 0, it indicates that the muscle effort is constant. This method helps us interpret the muscle effort applied along with steadiness and stability of the individuals during hold time in final posture.
|Figure 6: Root mean square signal of surface electromyography signal obtained during Trikonasana for expert (a and b) and individual 4 (c and d). Data points in 20 sec hold time are fit to a straight line. Bar graph represents the percentage of muscle activity|
Click here to view
| Results|| |
The sEMG signals observed for untrained individuals were significantly different from that of the expert. As discussed previously, the amplitude of the sEMG signal increases with an increase in active tension in the underlying muscles. For neck muscles, by observing amplitude of the sEMG signals, we can infer significantly higher muscle activity for the right SCM [[Figure 5]b, time: 30–50 s] than for the left side [[Figure 5]a, time: 30–50 s]. This indicates that muscle effort in the right SCM muscle was high when the neck was turned to the left. Opposite results are observed when the neck was turned to the right side [[Figure 5]a time: 65–85 s].
However, in case of abdominal oblique muscles, the right oblique activity is observed [[Figure 5]c time: 30–50 s] while bending to the right side as right obliques was in contraction and the left oblique activity was observed [Figure 5]d time: 65–85 s] while bending to the left side. Qualitatively, we can infer from the graph that muscles were active continuously over the 20 s period when the body was in final posture [[Figure 5] time: 30–50 s and 65–85 s].
The amplitude of the graph differs from person to person as it depends on how strong or weak that muscle is. For the same body motion, a person with strong muscles generates sEMG signals with higher amplitude than that of a person with weak muscles., Furthermore, the presence of body fat decreases the amplitude of sEMG.
It is difficult to maintain the ideal value of zero slopes in the final static position because even slight movements generate sEMG signals. For an expert, the slope value falls within the range of −0.4 to +0.4, which means the best fit line makes an angle in the range of ±22° with respect to x-axis [[Figure 6] bar graph]. To be able to say that the performance of the asana was smooth, the value of slope for the other individuals was expected to lie around or within this range. Along with slope, we also considered average of sEMG amplitude for analysis [[Figure 6] bar graph]. The slope values for expert and individual 4 are given in [Table 2]. Similarly, such slope values are available for all individuals and are not presented in [Table 2].
|Table 2: Value of slopes for the best fit line during 20 s hold in final posture|
Click here to view
Mean amplitude value of sEMG signal was taken to find the percentage of muscle activation while performing Trikonasana to the left and right side of the body. This method was used only for interpreting sEMG signals of neck muscle. In [Table 3], column 1 denotes the muscle activity in the left SCM muscle as a percentage of that of the right SCM when the neck was turned to the right. Similarly, column 2 denotes the muscle activity in the right SCM muscle as a percentage of that of the left SCM when the neck was turned to the left.
For expert (individual 1), when the neck was turned to the left if the mean amplitude of sEMG in the right SCM is taken as 100% [[Figure 6] - bar graph], only 40% of muscle activity (mean amplitude) was observed in the left SCM muscle. On the other hand, when the neck was turned to the right, if the mean amplitude of sEMG in the left SCM is taken as 100% [[Figure 6] - bar graph], only 38% of muscle activity (mean amplitude) was observed in the right SCM muscle.
[Table 3] gives a ratio of muscle activity (in terms of the mean amplitude of sEMG) observed when the neck was turned to the left and right side provided by Expert and individual 4; however, such values are available for all individuals and are not presented in the table. From [Table 3], we can comment on the proportionate use of the left and right SCM for turning the neck.
We can infer from the result of above method that untrained individuals are using more muscle effort (based on amplitude analysis) to turn their neck. The steep rise of sEMG signals during initial stages of asana gives information about more muscle effort used by the individual to turn the neck [[Figure 6]d time: 20–25 s]. The above inference can be validated by comparing the results of individual 4 with that of the expert. In [Figure 6], sEMG signals of neck muscle (SCM) for expert and individual 4 are shown. Let us look at the signals corresponding to the 20 s hold time; when the neck is turned left, the stretch is felt (range: 20–40 s). However, for individual 4 during left neck turn, instead of right side muscle, the left muscle is also active [Figure 6]c and d time range: 20–40 s]. The mean value of sEMG signal is taken for entire 20 s hold time in final posture and for muscle present on the same side (either left or right SCM); percentage of mean is calculated. In turning the neck right, the effort applied by the expert through his left SCM was 40% of that applied by his right SCM whereas the effort applied by individual 4 was 110% [Figure 6] bar graph]. Individual 4 is not using correct muscles to turn the neck while in final posture. The negative slope of the best fit line describes gradual decrease of muscle activity [[Figure 6]d time: 20–40 s and [[Figure 6]c time: 50–70 s].
The muscle effort used by individual 4 during the initial stage of asana is more, and it has decreased in magnitude over time, so the negative slope of the line. This individual must optimize his movements and use of muscles while performing asana.
Referring to the ACC plots from [Figure 7], we observe that ripples of significantly higher strength are observed with individual 4 when compared with those from the expert. This behavior is attributed to the wiggling movement of the body. Furthermore, a decrease in slope is observed in the signal from individual 4 which is not the case with that from the expert. Both observations solidify the point that nonexperts tend to deliver a performance that is neither as stable nor as smooth as that from an expert.
|Figure 7: Acceleration signal obtained during Trikonasana for expert (a and b) and individual 4 (c and d). Red = Acceleration in the x-axis, Green = Acceleration in y-axis, Blue = Acceleration in the z-axis|
Click here to view
ACC data were used for verifying the results obtained from analyzing sEMG signal. The objective of this paper was to introduce a method for analyzing sEMG while performing Yogasanas. Hence, we limit the ACC data analysis qualitatively.
| Discussion|| |
In Yoga sutras, “asana” is defined as “sthirasukhamasanam” which means that posture should be steady, stable, and comfortable. In this paper, our study examines a method to analyze the sEMG signal while performing Trikonasana. The steadiness and stability of the person were quantified based on muscle activity. Linear regression analysis was used to fit the data points of final posture to a straight line. The value of the slope of the line [Table 2] along with the percentage of muscle activity [Table 3] can be used to distinguish a good asana from poorly performed one. The slope values in [Table 2] can be used to comment on stability and steadiness of the person while performing asana. Percentage of muscle activity represented in [Table 3] gives an idea about the effective use of left and right SCM muscle while turning the neck. The changes in the amplitude of sEMG plot indicate variations in muscular activity and unbalance in the body while performing asana. These types of changes lead to deviation in the slope of the line from zero. Videos of all the experimental trials were collected to compare the kinematics of the person during final posture with that of the ACC data.
From [Table 3], we observe that individual 3 and individual 10 have optimally utilized the SCM muscle while turning neck to the left and right side. However, the magnitude of the slope is greater than zero, which means amplitude of sEMG signal is varying with time. This variation of slope suggests lack of stability of the individuals while performing the asana. From the sEMG plots for a few individuals, we observe the presence of impulse kind of response prevalent for a few seconds during the final posture. This may be due to forceful hold of the body.
The collaborative use of sEMG and ACC sensors for identifying the muscle performance while doing Yogasana has rarely been reported in the literature. In this study, we considered sEMG sensor as primary sensors for identifying whether the asana is performed and maintained correctly or not. We can use sEMG plot, the slope of the best fit line, and muscle activation pattern (percentage of muscle activity) for differentiating results among different individuals. However, there is also a need for ACC sensor to monitor the body kinematics.
Muscle recruitment pertaining to the asana was not stable during the hold of the final posture in novices whereas the muscle was expected to be in active state. The expert seems to be able to optimize the muscle recruitment at the right time and amplitude. This could reflect the learning patterns of the brain in optimizing the muscle activity during yoga keeping in mind the energy conservation principle.
The stability of the muscle recruitment reflects the ability of the nervous system to sustain a contraction at optimal levels for the duration of the asana. It may also reflect the muscular changes with respect to the fiber type and plastic changes that occur with long-term practice.
We see that the slope of the muscle recruitment is a distinguishing factor between expert and novice. We also see that achievement of the end points of the pose (Trikonasana) has been successful in all the participants of the study. However, the varying patterns of movement suggest different motor strategies in each individual. We propose based on the study a hypothesis that as the training progresses and the expertise increases, we should be able to see a predictable trend in the recruitment pattern of muscles. This would also suggest that for a given asana, it would be possible to predict the time and a normalized amplitude of movement that suggests expert achievement.
We suggest further studies to be conducted to test these hypotheses as well as correlate expert achievement with their stated therapeutic effects of Yoga. This study aimed to study the mechanics of Yoga both from a biomechanical as well as a motor control points of view and has succeeded in laying a foundational framework for future research in this area.
Our study provides information on a limited set of muscles that were mounted with sensors. Although the behavior of muscles such as latissimus dorsi, quadriceps femoris, and medial hamstring was not included as part of this study, multiplying the number of sensors (both ACC and sEMG) placed on the body would help obtain the additional data required to integrate body kinematics with the sEMG data. Furthermore, we believe that an analysis of variations in muscle activity among individuals with muscular disorders can provide information on how they differ from normal behavior. Every possible precaution was taken to avoid cross talk among the muscles associated with the respective body movements. However, due to the inherent nature of the human muscular system to involuntarily work in a collaborative fashion, it cannot be assured that it was eliminated. As the muscles chosen in this study are quite large in size, it would be reasonable to claim that the effect of cross talk was minimal.
| Conclusion|| |
The results suggested that the slope of the best fit line is a good metric for monitoring the muscle activity during Yoga performance. The slope can be used to differentiate a correct asana from one that is not performed correctly. This is a valid method for distinguishing an expert from a novice. This method can be extended for analyzing other asanas as well.
Financial support and sponsorship
I am thankful to The Department of Science & Technology for funding this project.
Conflicts of interest
There are no conflicts of interest.
| References|| |
Omkar SN, Badri Narayanan B, Rao A, Bhaskar R, Kumar S. A comparative study on performance analysis of sun salutation using fast Fourier transform, wavelet transform and HilbertHuang transform; 2011.
Roy SH, Cheng MS, Chang SS, Moore J, De Luca G, Nawab SH. A combined sEMG and accelerometer system for monitoring functional activity in stroke. IEEE Trans Neural Syst Rehabil Eng 2009;17:585-94.
Basmajian JV, De Luca CJ. Muscles Alive. 5th
ed. Baltimore, MD: Williams and Wilkins; 1985.
Andreia Sousa SP, João Manuel RS. Tavares, “Surface electromyographic amplitude normalization methods: A review”, Electromyography: New Developments, Procedures and Applications, Nova Science Publishers, Inc.; 2012. p. 85-102.
Leslie K. Yoga anatomy/Leslie Kaminoff; illustrated by Sharon Ellis. p. cm.
Zaheer F, Roy SH, De Luca CJ. Preferred sensor sites for surface EMG signal decomposition. Physiol Meas 2012;33:195-206.
De Luca CJ. Surface Electromyography: Detection and Recording. DelSys Incorporated. 2002; p. 2-10.
Lee J, Kagamihara Y, Kakei S. Quantitative evaluation of movement disorders in neurological diseases based on EMG signals. Conf Proc IEEE Eng Med Biol Soc 2008;2008:181-4.
Visser A, McCarroll RS, Oosting J, Naeije M. Masticatory electromyographic activity in healthy young adults and myogenous craniomandibular disorder patients. J Oral Rehabil 1994;21:67-76.
De Luca CJ, Gilmore LD, Kuznetsov M, Roy SH. Filtering the surface EMG signal: Movement artifact and baseline noise contamination. J Biomech 2010;43:1573-9.
Lawrence JH, De Luca CJ. Myoelectric signal versus force relationship in different human muscles. J Appl Physiol Respir Environ Exerc Physiol 1983;54:1653-9.
Milner-Brown HS, Stein RB. The relation between the surface electromyogram and muscular force. J Physiol 1975;246:549-69.
Woods J, Bigland-Ritchie B. Linear and Non-linear surface EMG/force relationships in human muscles. Am J Phys Med 1983;62:287-99.
Gretchen D. Oliver, Audrey J. Stone and Hillary Plummer, Electromyographic examination of selected muscle activation during isometric core exercises. 2010;20:6.
Ad Hoc Committee, International Society of Electrophysiological Kinesiology. Units, Terms, and Standards in the Reporting of EMG Research. International Society of Electrophysiological Kinesiology; [Goteborg, Sweden]: I.S.E.K.; 1980.
Ekstrom RA, Donatelli RA, Carp KC. Electromyographic analysis of core trunk, hip, and thigh muscles during 9 rehabilitation exercises. J Orthop Sports Phys Ther 2007;37:754-62.
Visser A, Kroon GW, Naeije M, Hansson TL. EMG differences between weak and strong myogenous CMD patients and healthy controls. J Oral Rehabil 1995;22:429-34.
Ms Ashitha Besagarahalli Ramesh
Indian Institute of Science, Bengaluru, Karnataka
Source of Support: None, Conflict of Interest: None
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]
[Table 1], [Table 2], [Table 3]
|This article has been cited by|
||A Survey on Yogic Posture Recognition
| ||Arun Kumar Rajendran, Sibi Chakkaravarthy Sethuraman |
| ||IEEE Access. 2023; 11: 11183 |
|[Pubmed] | [DOI]|
| Article Access Statistics|
| Viewed||3910 |
| Printed||119 |
| Emailed||0 |
| PDF Downloaded||180 |
| Comments ||[Add] |
| Cited by others ||1 |