Relating surface electromyogram (EMG) activity to force/torque models is used in many areas including: prosthesis control systems, to regulate direction and speed of movement in reaching and matching tasks; clinical biomechanics, to assess muscle deficiency and effort levels; and ergonomics analysis, to assess risk of work-related injury such as back pain, fatigue and skill tests. This thesis work concentrated on improving the performance of dynamic EMG-to-force models for the hand-wrist and multiple fingers. My contributions include: 1) rapid calibration of dynamic hand-wrist EMG-force models using a minimum number of electrodes, 2) efficiently training two degree of freedom (DoF) hand-wrist EMG-force models, and 3) estimating individual and combined fingertip forces from forearm EMG during constant-pose, force-varying tasks. My calibration approach for hand-wrist EMG-force models optimized three main factors for 1-DoF and 2-DoF tasks: training duration (14, 22, 30, 38, 44, 52, 60, 68, 76 s), number of electrodes (2 through 16), and model forms (subject-specific, DoF-specific, universal). The results show that training duration can be reduced from historical 76 s to 40–60 s without statistically affecting the average error for both 1-DoF and 2-DoF tasks. Reducing the number of electrodes depended on the number of DoFs. One-DoF models can be reduced to 2 electrodes with average test error range of 8.3–9.2% maximum voluntary contraction (MVC), depending on the DoF (e.g., flexion-extension, radial-ulnar deviation, pronation-supination, open-close). Additionally, 2-DoF models can be reduced to 6 electrodes with average error of 7.17–9.21 %MVC. Subject-specific models had the lowest error for 1-DoF tasks while DoF-specific and universal were the lowest for 2-DoF tasks. In the EMG-finger project, we studied independent contraction of one, two, three or four fingers (thumb excluded), as well as contraction of four fingers in unison. Using regression, we found that a pseudo-inverse tolerance (ratio of largest to smallest singular value) of 0.01 was optimal. Lower values produced erratic models and higher values produced models with higher errors. EMG-force errors using one finger ranged from 2.5–3.8 %MVC, using the optimal pseudoinverse tolerance. With additional fingers (two, three or four), the average error ranged from 5–8 %MVC. When four fingers contracted in unison, the average error was 4.3 %MVC. Additionally, I participated in two team projects—EMG-force dynamic models about the elbow and relating forearm muscle EMG to finger force during slowly force varying contractions. This work is also described herein.

Electromyogram signals generated by human muscles can be measured on the surface of the skin and then processed for use in applications such as prostheses control, kinesiology and diagnostic medicine. Most EMG applications extract an estimate of the EMG amplitude, defined as the time-varying standard deviation of EMG, EMGσ. To improve the quality of EMGσ, additional signal processing techniques, such as whitening, noise reduction and additional signal features can be incorporated into the EMGσ processing. Implementation of these additional processing techniques improve the quality of the processed signal but at the cost of increased computational complexity and required calibration contractions. Whitening filters are employed to temporally decorrelate data so that the samples are statistically independent. Different types of whitening filters, linear and adaptive, and their performance have been previously studied in (Clancy and Hogan) and (Clancy and Farry). The linear filter fails at low effort levels and the adaptive filter requires a calibration every time electrodes are removed and reapplied. With the goal of avoiding the disadvantages of the previous whitening filter approaches, the first signal processing technique studied herein developed a universal fixed whitening filter using the ensemble mean of the power spectrum density of EMG recordings from the 64 subjects available in an existing data set. Performance of the EMG to torque model with the universal fixed whitening filter was computed to be 4.8% maximum voluntary contraction (MVC); this is comparable to the 4.84 %MVC error computed for the adaptive whitening filter. The universal fixed whitening filter preserves the performance of the adaptive filter but need not be calibrated for each electrode. To optimize noise reduction, the second signal processing technique studied derived analytical models using the resting EMG data. The probability density function of the rest contractions was observed to be very close to a Gaussian distribution, showing only a 1.6% difference when compared to a Gaussian distribution. Once the models were developed, they were used to prove that the optimal subtraction of the noise variance is to compute the root of the difference between the signal squared and noise variance (RDS). If this result would lead to a negative value, it must be set to zero; EMGσ cannot contain negative components. Once the RDS was proven to be the optimal noise subtraction, it was implemented on 0% MVC and 50% MVC data. The RDS processing has a considerable impact on lower level contractions (0% MVC), but not on higher level contractions (50% MVC), as expected. The third signal processing technique involved the creation of a new EMG feature from four individual signal features. Different techniques were used to combine EMGσ, zero crossings (ZC), slope sign changes (SSC) and waveform length (WL) into a single new EMG feature that would be used in an end application, such as the modeling of torque about the elbow or prosthesis control. The new EMG feature was developed to reduce the variance of the traditional EMGσ only feature and to eliminate the need for calibration contractions. Five different methods of combination were attempted, but none of the new EMG features improved performance in EMG to torque model.

This thesis reports research on analysis of the variance of surface electromyogram (sEMG) for healthy participants and people suffering with Lower Back Pain (LBP) when they are walking and running. SEMG signal recorded when the participants were walking and running on a treadmill. The strength and duration of the muscle activity for each heel strike were the features. The results indicate that there was no significant difference in the variance and in the change of variance over time of the amplitude between the two groups when the participants were walking. However when the participants were running, there was a significant difference in the two cohorts. While there was an increase in the total variance over the duration of the exercise for both the groups, the increase in variance of the LBP group was much greater (order of ten times) compared with the participants with healthy backs. The difference between the two groups was also very significant when observing the change of variance over the duration of the exercise. From these results, it is suggested that variance of sEMG of the muscles of the lower back, recorded when the participants are running, can be used to identify LBP patients.

No description available.

This thesis describes the system design of EMG Site, a MATLAB-based application for collection and visualization of surface electromyograms (EMGs) and the real-time control of an upper limb prosthesis, including details pertaining to the design of the software and the graphical user interface (GUI). The application consists of features that aid in the visualization of the collected EMG data and the control of a prosthesis. Visualization of the collected EMG data is handled in one of two ways: an oscilloscope-like view showing the raw EMG data collected with respect to time, or a radial plot showing the processed EMG data collected with respect to the site of EMG data collection on the arm. The control of a hand-wrist prosthesis is primarily regulated through the use of signal processing designed to relate EMG to torque and is visualized in the tracking window - a plotting window showing both a user-control cursor and an either static (or dynamic) computer-controlled target. This thesis concludes with a description of the real-time capabilities of the application regarding both the visualization of the collected EMG data as well as the control of a prosthesis.

Extensive studies about neural mechanisms involved in insect flight control have been carried out. Adaptive control of locomotion requires integration of salient sensory cues with ongoing motor activity. During flight, inputs received by an organism through sensory organs are processed by the central nervous system (CNS) and the integrated output thus obtained plays a significant role in controlling the wing phase shifts and flight muscle depressor asymmetries associated with adaptive flight maneuvers. The resulting maneuvers, in turn, bring a change in the insects sensory environment, thereby closing the feedback loop. Research on insect flight has been carried out using immobile preparations (tethered) and mobile preparations (free flight untethered). There are pros and cons associated with the tethered and the untethered approach. The tethered approach, however, provides an easier way to study the CNS and its role in motor control of flight. Insects such as locusts and moths exhibit pertinent wing phase shifts and asymmetries in depressor muscles. For locusts constant wing phase shifts and m97 (forewing first basalar depressor muscle) depressor asymmetries have been observed during adaptive flight maneuvers making this a useful system for creation of behaviorally appropriate visual feedback. A method that utilizes asymmetrical timing of bilateral depressor muscles, the forewing first basalars (m97), of the locust to close a visual feedback loop in a computer-generated flight simulator is presented here. The method converts the time difference between left and right m97s to analog voltage values. Analog voltage values can be acquired using an open-loop experimental protocol (visual motion controlled by the experimenter), or can be used to control closed-loop experiments (muscle activity controls the visual stimuli) experiments. We recorded electromyographic (EMG) activity from right and left m97 muscles. On testing this circuit with real animals, we were able to detect the spike time difference and convert it to voltage values. These voltage values were utilized to control the presentation of a stimulus in a closed-loop environment. The feedback circuit presented here may be used in conjunction with the flight simulator(s) to understand the neural mechanisms involved in control of insect flight and provide further understanding of general mechanisms of neural control of behaviour.

Extensive studies about neural mechanisms involved in insect flight control have been carried out. Adaptive control of locomotion requires integration of salient sensory cues with ongoing motor activity. During flight, inputs received by an organism through sensory organs are processed by the central nervous system (CNS) and the integrated output thus obtained plays a significant role in controlling the wing phase shifts and flight muscle depressor asymmetries associated with adaptive flight maneuvers. The resulting maneuvers, in turn, bring a change in the insects sensory environment, thereby closing the feedback loop. Research on insect flight has been carried out using immobile preparations (tethered) and mobile preparations (free flight untethered). There are pros and cons associated with the tethered and the untethered approach. The tethered approach, however, provides an easier way to study the CNS and its role in motor control of flight. Insects such as locusts and moths exhibit pertinent wing phase shifts and asymmetries in depressor muscles. For locusts constant wing phase shifts and m97 (forewing first basalar depressor muscle) depressor asymmetries have been observed during adaptive flight maneuvers making this a useful system for creation of behaviorally appropriate visual feedback. A method that utilizes asymmetrical timing of bilateral depressor muscles, the forewing first basalars (m97), of the locust to close a visual feedback loop in a computer-generated flight simulator is presented here. The method converts the time difference between left and right m97s to analog voltage values. Analog voltage values can be acquired using an open-loop experimental protocol (visual motion controlled by the experimenter), or can be used to control closed-loop experiments (muscle activity controls the visual stimuli) experiments. We recorded electromyographic (EMG) activity from right and left m97 muscles. On testing this circuit with real animals, we were able to detect the spike time difference and convert it to voltage values. These voltage values were utilized to control the presentation of a stimulus in a closed-loop environment. The feedback circuit presented here may be used in conjunction with the flight simulator(s) to understand the neural mechanisms involved in control of insect flight and provide further understanding of general mechanisms of neural control of behaviour.

Estimates of the time-varying standard deviation of the surface EMG signal (EMGσ) are extensively used in the field of EMG-torque estimation. The use of a whitening filter can substantially improve the accuracy of EMGσ estimation by removing the signal correlation and increasing the statistical bandwidth. However, a subject-specific whitening filter which is calibrated to each subject, is quite complex and inconvenient. To solve this problem, we first calibrated a 60th-order “Universal” FIR whitening filter by using the ensemble mean of the inverse of the square root of the power spectral density (PSD) of the noise-free EMG signal. Pre-existing data from elbow contraction of 64 subjects, providing 512 recording trials were used. The test error on an EMG-torque task based on the “Universal” FIR whitening filter had a mean error of 4.80% maximum voluntary contraction (MVC) with a standard deviation of 2.03% MVC. Meanwhile the subject-specific whitening filter had performance of 4.84±1.98% MVC (both have a whitening band limit at 600 Hz). These two methods had no statistical difference. Furthermore, a 2nd-order IIR whitening filter was designed based on the magnitude response of the “Universal” FIR whitening filter, via the differential evolution algorithm. The performance of this IIR whitening filter was very similar to the FIR filter, with a performance of 4.81±2.12% MVC. A statistical test showed that these two methods had no significant difference either. Additionally, a complete theory of EMG in additive measured noise contraction modeling is described. Results show that subtracting the variance of whitened noise by computing the root difference of the square (RDS) is the correct way to remove noise from the EMG signal.

Soft material-enabled electronics can demonstrate extreme mechanical flexibility and stretchability. Such compliant, comfortable electronics allow continuous, long-term measurement of biopotentials on the skin. Manufacturing of the stretchable electronic devices is enabled by the recent development combining materials transfer printing and microfabrication. However, the existing method using inorganic materials and multi-layered polymers requires long material preparation time and expensive processing cost due to the requirement of microfabrication tools and complicated transfer printing steps. Here, this study develops a new fabrication method of soft electronics via a micro-replica-molding technique, which allows fast production, multiple use, and low cost by avoiding microfabrication and multiple transfer printing. The core materials, carbon nanomaterials integrated with soft elastomers, further reduces the entire production cost, compared to costly metals such as gold and silver, while offering mechanical compliance. Collectively, skin-wearable electrodes, designed by optimized materials and fabrication method enable a high-fidelity measurement of non-invasive electromyograms on the skin for advanced human-machine interface, targeting prosthesis.

The standard deviation of surface EMG (EMGσ) is often related to muscle force; the accuracy of EMGσ estimation is valuable for many application areas such as clinical biomechanics, prostheses control and sports medicine. Numerous researchers have developed methods to optimize EMGσ estimation. Whitening the EMG signal has been proved to improve the statistical efficiency of EMGσ estimation, but conventional linear whitening filters fail at low contraction level. An adaptive whitening filter was proposed by Clancy and Farry[14]. This technique has a better performance than prior whitening methods, however, the adaptive whitening filter needs to be calibrated each time electrodes are applied, which increase the complexity of the implementation. We designed a universal whitening filter which can omit most calibration steps for the adaptive whitening filter in future use. We used the ensemble mean of the power spectrum of 512 EMG recordings to form a general shape of a fixed whitening filter that can applied on any EMG signal. The test error on an EMG-torque task based on universal whitening over 512 subjects had a mean error of 4.80% maximum voluntary contraction (MVC) and standard deviation (std) of 2.03% MVC, compared with an original adaptive whitening filter error of 4.84±1.98% MVC. Additionally, the rest contraction modeling hasn’t received enough attention. Existing RMS estimates of EMGσ subtract noise in either the amplitude or power domain. These procedures have never been modeled analytically. We show that power domain noise subtraction is optimal. But rest contractions which are processed using power domain noise subtraction only estimate a zero-valued EMGσ approximately 50% of the time, which is undesirable in prosthesis-control. The prosthesis has a 50% possibility to slowly drift based on the current RMS estimator. We propose a new estimator to improve the zero-amplitude estimation probability during rest. We used 512 rest contraction recordings to validate the probability distribution of rest EMG signal showing that it only has 1.6% difference compared with Gaussian distribution. We also evaluated the percent of zero-valued EMGσ estimates using power domain noise subtraction and our new estimator, matching experimental findings to the theoretic predictions.