analyzing-bicep-emg
- 1. Analyzing bicep EMG signals to quantify a relationship between muscle electric potential and object weight Austin Lefebvre1,*, Jericho Hallare1,*, and Bruce Wheeler1 1Department of Bioengineering, University of California, San Diego, United States *Assignment submitted by these students ABSTRACT Electromyography (EMG) allows for the quantification of electrical signal activity produced by skeletal muscle tissue by using the electrical potential of muscle cells. Here, a biopotential amplifier is built and utilized on a subject to correlate bicep contraction magnitude to peak-to-peak voltage of the myocyte potential signal. Running the signal through a MATLAB code allows for a quantitative prediction of the average force generated by a muscle’s contraction by utilizing this correlation. The EMG’s noise was minimized, and desired frequencies were amplified using various instrumentation amplifiers, operational amplifiers, and bandpass filters. The results confirmed a correlation between muscle contraction and amount of weight lifted, allowing for possible innovative and individualized physical therapy and strength conditioning techniques by extrapolation of this linear correlation. Introduction EMGs have proven useful in biomedical engineering applications as it provides visualization into the physiological events of the electrical activity occurring within the muscle cells. When a muscle contracts, multiple action potentials are being sent through the body, from the neurons in the nervous system, to the muscle fibers receiving the electrical signals.1 Based on the strength and health of these muscle fibers, contraction potential ability is varying, and is of great interest to physical therapists.2 We hypothesize that the analysis of the contraction of specific muscles can be analyzed and correlated to the weight of an object, which could provide detailed therapy tailored to specific muscles of a patient, in contrast to most physical therapy where muscular problems are treated in a generalized manner. The electrical potential of localized muscle cells is difficult to isolate, as there are many extraneous electrical signals creating background noise, including those of the heart, other neurons firing in the body, and any ambient signals in the environment and equipment being used. These extrinsic factors can generate signals that remain difficult to quantitatively analyze without proper filtration and amplification techniques. A typical muscle contraction can range anywhere from 0 to 10 mV before amplification, depending on the intensity of the muscle contraction.3 Additionally, the firing rate of the muscle cells generally lies within the 0 to 20 Hz range.3 In this paper, a biopotential amplifier was constructed with the proper filtration and amplification for myocyte potentials. Electrodes were then applied to the human test subject’s bicep to acquire an EMG signal. The signal captured the electrical potential of the muscle cells in the bicep during the subject’s muscle contraction, and data acquisition from this signal was exported through a LabVIEW virtual instrument. This data was then analyzed through MATLAB, rejecting noise and acquiring an average peak-to-peak voltage, creating a model fit between the magnitude of the EMG contractions and voltage differences in the data. This model fit can then be used to derive a relationship of voltage differences to force applied by the bicep. A possible transmission of contraction analysis using Bluetooth can lead to the real-time extrapolation of object weight directly to a smart-phone application, leading to a variety of useful day-to-day implementations. Methods A biopotential amplifier and band-pass filter circuit were created (Figure 1) in order to amplify the desired electrical potential from the bicep, while filtering out any noise generated by the equipment and the rest of the body. Amplification was achieved using an LM741 operational amplifier, while the AD622 helped to amplify differential noise between the electrode inputs on the bicep and reducing any DC offset present. The band-pass filter was customized to allow a passage of typical frequencies generated by muscle cells.
- 2. Figure 1. Biopotential circuit diagram with measured resistance and capacitor values at 5% tolerance. V1 connects with one EMG electrode at the bicep, while V2 connects with the other. Ground is connected to an electrode at the hip. In order to measure the correlation between bared force and bicep contraction, a healthy adult male subject held objects of measured varied weights. The contraction potential was then measured and analyzed, creating a graphical and equational representation of the data. The subject was asked to stand with his arm bent at the elbow, perpendicular to his body and parallel to the ground, with one palm facing upwards. The subject then had electrodes placed approximately 8 centimeters apart, along and parallel to the bicep muscle, and an additional electrode was placed on the hip as a ground reference point (Figure 2). The outputs of the electrodes were then connected to the input pins of the AD622 of the circuit. The LabVIEW Virtual Instrument was then ran for 10 seconds for each weight, capturing 2000 samples with a sampling rate of 200 samples per second. The first 5 seconds were without a weight, to capture a resting myoelectric signal to be used to reduce the peak-to-peak analysis variability. The proceeding 5 seconds were measured while the subject held the chosen weight. The short time length was chosen to limit the effects of muscle fatigue while holding the weight. Varying weights were chosen by placing varying objects inside of a backpack and weighed, giving a wide range of testable weights, and one data set was captured for each weight tested. The subject was required to wait a minimum of 120 seconds before testing another weight in order to minimize the effect of muscle fatigue from previous tests. Figure 2. Set-up of experimental design. Two electrode inputs are placed on the bicep, and one electrode is placed on the hip as ground. The user holds a weight in the hand of the arm being measured. The data was then exported to Matlab, where it was analyzed by the constructed Matlab script. The output waveform peak-to-peak average was then exported from Matlab to Microsoft Excel, where they were plotted against their respective weight. A trendline, linear equation, and R-squared value were then generated to form a model of and simultaneously quantitatively analyze the correlation 2/5
- 3. Results The circuit and subsequent user setup was successfully implemented and allowed for accurate myocyte potential signal collection using the LabVIEW VI (Figure 3). Figure 3. A graph of collected data points through LabVIEW. 2000 samples were captured from the biopotential amplifier over 10 seconds while the subject held a weight of 31.5 pounds. By quantifying the relationship between bicep myocyte potential and the force of the object being held, a high linear correlation coefficient was found in the results (Figure 4). Figure 4. A graphical representation of the average peak-to-peak values captured from the constructed biopotential amplifier plotted against the respective weight of the object held by the subject. For the subject that was tested, each additional pound of weight added to the backpack caused a contraction signal increase of approximately 3.24 mV after amplification. With such a high R-squared value, the correlation between contraction signal and object weight cannot be discounted. Discussion As hypothesized, there was a strong correlation between bicep myocyte potential and the amount of weight carried. By using an EMG, and simple extrapolation of a resulting line equation modeled from a preliminary data acquisition period, an estimation of the weight of an object held by a subject can be measured. 3/5
- 4. Several issues arose with the experimentation procedure. The weight of the backpack was measured to the best of our abilities, using a simple bathroom scale and taking the difference between the subject carrying the backpack, and the subject without the backpack. As this is quite inaccurate, a better model would surely result in the use of a professional procedure. Additionally, the electrodes used were faulty at many points, and are a big source of unreliable data, leading to potential deviations evident on the graph. Furthermore, additional experimentation would involve multiple subjects, and of varying genders and fitness levels in order to establish more individualized quantitative models. A Matlab script was created for the purposes of this experiment in order to analyze average peak-to-peak values of a highly variable signal, and in an attempt to limit as much noise as possible. Further research into existing programs or scripts would likely generate a more accurate way of producing desired results, and should not be ignored. The efficacy of implementation in a physical therapy patient’s day-to-day life would be most based upon the portability of such a device. Fortunately there are many ways to introduce a portable aspect to the EMG. A bluetooth transmitter can be attached to an arduino device using a breadboard, and used to transmit the analyzed data to a user’s smartphone (Figure 5). Throughout the experiment, multiple attempts were sought out using this method, but unfortunately failed in the end, achieving only acquisition of signals by the Arduino, but no transmission from the Arduino were received by the computer. Future experiments could likely overcome this barrier, presenting a viable way for users to easily and portably use this device. Figure 5. Wiring a bluetooth module (HC-05) to an Arduino using a breadboard. In physical therapy, a common technique for analyzing muscle failure is to use varying set weights to analyze and test the capabilities of the patient. Unfortunately there are dangers presented with increasing the set weight by too much, as a recovering patient’s muscles are vulnerable to injury. Using instead a rubber arm band would allow for a continuous range of forces to be applied to the patient’s muscle, allowing him to dictate when his capability limit is reached. Unfortunately, the acquisition of the weight data for the armband is difficult, however this can be facilitated using an EMG attached to the patient’s muscle as was done in this experiment. Conclusion This experiment demonstrated the applicability and usefulness of acquiring EMG signals. The results revealed the ability of the EMG to filter out unwanted signals while simultaneously amplifying the EMG signal. To possess the ability to acquire an EMG signal and correlate to the amount of weight an individual can tolerate is a very powerful tool in the biomedical world. This project allowed for the correlation of the bicep contraction to the weight that the subject is holding, but further experimentation is needed to fully outline this correlation in a wider, and more reliable set of data, as well as creating a feasible method of producing a usable, and useful device. References 1. Saladin, K. S. Anatomy and Physiology: the Unity of Form and Function., vol. 1 (Dubuque: McGraw-Hill, 2010). 4/5
- 5. 2. GILMORE, K. L. & MEYERS, J. E. Using Surface Electromyography in Physiotherapy Research. Australian Jour- nal of Physiotherapy 29, 3–9 (1983). URL http://www.sciencedirect.com/science/article/pii/ S0004951414606590. 3. Raez, M., M.S., H. & F., M.-Y. Tehniques of EMG Signal Analysis: Detection, Processing, Classification and Applications. Biological Procedures Online 8, 11–35 (2006). Acknowledgements Thank you to the TA’s Edward Catoiu, Rui Wang, and Hillary Lam for the extensive help throughout the entire experiment, and especially to Professor Wheeler who helped us through frustrating experimentation procedures. 5/5