- Open Access
Muscle activation patterns during walking from transtibial amputees recorded within the residual limb-prosthetic interface
© Huang and Ferris; licensee BioMed Central Ltd. 2012
Received: 4 October 2011
Accepted: 5 July 2012
Published: 10 August 2012
Powered lower limb prostheses could be more functional if they had access to feedforward control signals from the user’s nervous system. Myoelectric signals are one potential control source. The purpose of this study was to determine if muscle activation signals could be recorded from residual lower limb muscles within the prosthetic socket-limb interface during walking.
We recorded surface electromyography from three lower leg muscles (tibilias anterior, gastrocnemius medial head, gastrocnemius lateral head) and four upper leg muscles (vastus lateralis, rectus femoris, biceps femoris, and gluteus medius) of 12 unilateral transtibial amputee subjects and 12 non-amputee subjects during treadmill walking at 0.7, 1.0, 1.3, and 1.6 m/s. Muscle signals were recorded from the amputated leg of amputee subjects and the right leg of control subjects. For amputee subjects, lower leg muscle signals were recorded from within the limb-socket interface and from muscles above the knee. We quantified differences in the muscle activation profile between amputee and control groups during treadmill walking using cross-correlation analyses. We also assessed the step-to-step inter-subject variability of these profiles by calculating variance-to-signal ratios.
We found that amputee subjects demonstrated reliable muscle recruitment signals from residual lower leg muscles recorded within the prosthetic socket during walking, which were locked to particular phases of the gait cycle. However, muscle activation profile variability was higher for amputee subjects than for control subjects.
Robotic lower limb prostheses could use myoelectric signals recorded from surface electrodes within the socket-limb interface to derive feedforward commands from the amputee’s nervous system.
Recent advances in robotic technology have allowed for the development of powered lower limb prostheses that improve ambulation for amputees. A major feature of these new devices is the ability to interject mechanical power into the gait cycle to replace the mechanical power that is lost due to missing biological muscles. Hugh Herr’s research group at the Massachusetts’s Institute of Technology has developed a robotic ankle that uses a finite state controller to modulate ankle dynamics during gait and add power to the trailing limb during push off [1–3]. The prosthesis uses intrinsic sensing of kinetics and kinematics (e.g., heel- and toe-contact, ankle angle, and ankle torque) to determine when to transition between gait phases during walking. Their powered prosthesis resulted in lower metabolic cost compared to traditional passive elastic prostheses for level ground walking . In addition to a robotic ankle, they have developed a variable impedance robotic knee that uses intrinsic sensing and a finite state controller to modulate knee stiffness during level ground walking . Michael Goldfarb’s research group at Vanderbilt University has developed a robotic knee and ankle for transfemoral amputees that also uses intrinsic sensing and finite state control [6–8]. Tom Sugar’s research group at Arizona State University developed a powered ankle that relies on elastic elements to store energy and amplify mechanical power generated by the actuator . It uses intrinsic sensing to detect heel strike and then the controller initiates a predetermined gait pattern. This sampling of robotic prostheses is representative of the intrinsic sensing approaches that are beginning to be utilized for prosthetic control [10, 11].
There are advantages and disadvantages of controlling prosthetic lower limbs via intrinsic sensing. An advantage of prosthetics that rely on kinetic and kinematic sensing to infer user intent is that all of the sensors and associated computational hardware are built directly into the prosthetic. The interface with the human is purely mechanical, which simplifies socket design. These prosthetics generally have low step-to-step variability due to the robustness of the finite state controllers and the low sensor noise. Controllers based on intrinsic sensing tend to work well for stereotyped or cyclical tasks, such as gait. One of the inherent drawbacks of these devices is that control based on intrinsic sensing is not very good at aperiodic or highly variable motor tasks. For example, going up on the toes to reach a higher shelf would be very difficult for a state-based controller to perform using intrinsic sensing. Similarly, tasks with highly variable step-to-step kinematics such as traversing obstacles in the terrain, traversing unstable terrain, or negotiating through a crowd of people, or dealing with a variety of natural surfaces like sand and rocks would be difficult to deal with using intrinsic sensing alone.
An alternative to controllers that rely solely on intrinsic kinematic and kinetic sensing is to directly connect the prosthesis dynamics to the user’s nervous system via electromyography [12–14]. Myoelectric control has been implemented for powered upper limb prostheses. High costs have limited widespread acceptance of these devices but cost will continue to fall with continued technological advances. A more lasting obstacle to widespread acceptance of powered upper limb prostheses is the degrees of freedom that must be controlled. The human hand and wrist have more than 20 mechanical degrees of freedom but upper limb prostheses usually rely on fewer than 6 myoelectric control sources. This limits the ability for users to accurately and reliably control prosthesis mechanics. For the lower limb, fewer mechanical degrees of freedom are necessary to provide functional motor ability. For a transtibial amputee, active mechanical plantar flexion/dorsiflexion and passive foot elasticity can provide a huge energetic improvement compared to passive lower limb prostheses .
Controlling a limited number of mechanical degrees of freedom with myoelectric signals is feasible. Transfemoral amputees can learn to volitionally control virtual knee/ankle joint movements using myoelectric control signals from residual thigh muscles while seated and not wearing their prosthesis [15, 16]. In addition, transtibial amputees can learn to volitionally activate residual muscles during the swing phase of walking to switch between level-ground walking and stair-descent locomotion modes . To the best of our knowledge, this is the only case where myoelectric signals have been recorded from within the socket-limb interface during walking and used for user movement intent recognition.
To implement more robust myoelectric controllers for transtibial prostheses, it is important to assess lower leg electromyographic signal quality, variability, and adaptability during amputee gait. In the near future, it may be possible to use intramuscular electromyography sensors (IMES) to transmit electromyographic signals through the socket interface without breaking the skin [17–19]. These IMES would make it feasible to implement a wide range of myoelectric control methods with powered prostheses. However, rather than waiting for these IMES to be approved for human testing, we have recorded electromyography from lower leg muscles of transtibial amputees within the socket interface using surface electrodes. The purposes of this study were 1) to determine if surface electromyography signals can be recorded from residual lower leg muscles inside the prosthetic socket during walking, and 2) to quantify differences in muscle activation patterns between amputee and non-amputee subjects during walking.
Amputee subject details
We recorded ground reaction forces in the vertical, medial-lateral, and fore-aft directions at 1000 Hz using a custom-built instrumented split-belt treadmill . We defined heel-strike and toe-off events from vertical ground reaction force.
The first part of the test protocol assessed the subject’s ability to differentiate plantar flexor and dorsiflexor muscle activation. Subjects performed maximum voluntary activation trials where they tried to isolate the activation of their tibialis anterior (dorsiflexion trial) and gastrocnemii (plantar flexion trial) muscles. Subjects were seated upright on a raised platform so that their feet did not contact the ground during the maximum voluntary activation trials. To obtain maximal activation of the tibialis anterior, we instructed subjects to point their feet and toes towards the ceiling as hard as possible and sustain muscle activation at maximum dorsiflexion. To obtain maximal activation of the gastrocnemii, we instructed subjects to point their feet and toes towards the ground as hard as possible and sustain muscle activation at maximum plantar flexion. All ankle movements were performed bilaterally. We instructed amputee subjects to activate their lower leg muscles as if they had an intact ankle and foot. During practice trials, we displayed real time EMG signals to amputee subjects to provide feedback on the level of muscle activation. Once EMG signals appeared consistent, we recorded three repetitions for each maximum voluntary activation task. For each repetition, we asked the subjects to sustain the maximum voluntary activation for five seconds then rest with muscles fully relaxed for five seconds.
The second part of the test protocol assessed muscle activation patterns during walking. Subjects walked on a treadmill at four speeds (0.7, 1.0, 1.3, and 1.6 m/s) for two minutes at each speed. Not all subjects were able to walk at the two faster speeds. To determine the fastest walking trial that subjects could complete safely, we asked each subject to practice walking on the treadmill starting at the slowest speed. If they could walk comfortably at the given speed, we increased the treadmill speed gradually to the next level. We continued this until the fastest treadmill speed was reached or until the subject could no longer maintain walking speed. All subjects completed the 0.7 and 1.0 m/s trials. Eight of the twelve amputee subjects and eleven of the twelve control subjects completed the 1.3 m/s trial. Seven of the twelve amputee subjects and eleven of the twelve control subjects completed the 1.6 m/s trial.
We performed all signal processing and statistical analyses using the R computing environment (R Development Core Team, 1999). We processed EMG signals using two separate methods. To look at raw EMG, we applied a high-pass filter (bidirectional Butterworth, 4th order, 50 Hz cutoff frequency) and then demeaned the signal. We chose a cutoff frequency of 50 Hz to ensure that motion artifacts were attenuated. To analyze the frequency content of the signal, we calculated a smoothed periodogram estimated by a discrete Fourier transform and filtered using Daniell smoothers (single span of length 5). We calculated an empirical cumulative distribution function of the power spectrum to compare the distribution of frequency content between the amputee and control groups.
For the maximum voluntary activation trials, we performed frequency analysis of the tibilias anterior and gastrocnemii EMG for two seconds of sustained activation. For each subject, we selected the repetition where the maximum amplitude of the rectified signal (high-pass filtered and demeaned) was the greatest across trials. For some amputee subjects, the residual limb tibialis anterior was activated more than the gastrocnemii during the plantar flexion trial and vice versa during the dorsiflexion trial. For 1.0 m/s walking, we performed frequency analysis of the tibilias anterior and gastrocnemii for a single gait cycle. For each subject, we selected the gait cycle where the variance of the signal (high-pass filtered and demeaned) was closest to the mean variance of all cycles.
To quantify muscle activation profiles, we calculated EMG intensity using a wavelet decomposition method . We calculated an intensity curve by summing across wavelets 4 (center frequency = 62.1 Hz) through 11 (center frequency = 395.5 Hz) in time. This method was chosen over other methods (e.g. generating a linear envelope using a low-pass filter) because the intensity curve provided a more distinct profile, specifically at transitions between baseline and activation. We divided the intensity curve into cycles defined by consecutive heel strike events. We normalized time by interpolating over 500 equally spaced points per cycle using cubic splines, and we normalized the amplitude to the maximum amplitude across all walking speeds. We calculated a mean intensity curve from 40 consecutive time- and amplitude-normalized cycles. To quantify the repeatability of the recorded EMG signals, we calculated a variance-to-signal ratio (VSR) as the sum of the signal variance over the sum of the signal mean squared across the 40 consecutive normalized intensity curves: . To quantify differences in EMG shape, we used mean intensity curves to calculate normalized cross-correlations with zero time lag  between: 1) control group grand mean and control subject mean , 2) control group grand mean and amputee subject mean , and 3) amputee group grand mean and amputee subject mean . For cross-correlations and , individual subject data was excluded from the group mean. Normalized cross-correlations were calculated for EMG from all seven muscles using the subset of subjects who completed all four walking speeds.
We performed two separate ANOVAs to determine if there were significant differences in median EMG frequency between subject groups during either maximum voluntary activations or treadmill walking at 1.0 m/s. (model: median frequency ~ muscle + group). We performed another ANOVA to determine if there were significant differences in median EMG frequency between maximum voluntary activation and treadmill walking (factor: task) at 1.0 m/s for lower leg muscles only (model: median frequency ~ muscle + group*task). We performed two ANOVAs to determine if there were significant differences in cross-correlation (R-value) between subject groups (model: R-value ~ muscle + group). For the first ANOVA, the independent variable was for control subjects and for amputee subjects. For the second ANOVA, the independent variable was for control subjects and for amputee subjects. For all ANOVAs, if factors of interest were significant (p < 0.05), we performed a Tukey’s Honestly Significant Difference test to determine which contrasts were significant (p < 0.05).
Maximum voluntary activation of lower Leg muscles
Lower Leg EMG during walking
EMG activation pattern cross-correlations
0.73 (0.11) *°
−0.33 (0.13) °
0.44 (0.21) *
Gastrocnemius Medial Head
0.90 (0.09) *°
0.48 (0.42) °
0.45 (0.32) *
Gastrocnemius Lateral Head
0.79 (0.17) *°
0.37 (0.40) °
0.37 (0.35) *
0.75 (0.10) *°
0.31 (0.48) °
0.35 (0.36) *
0.72 (0.31) *
0.67 (0.28) *
0.80 (0.09) *°
−0.24 (0.08) °
0.32 (0.25) *
Gastrocnemius Medial Head
0.87 (0.08) *°
0.23 (0.40) °
0.20 (0.19) *
Gastrocnemius Lateral Head
0.83 (0.12) *°
0.20 (0.37) °
0.24 (0.37) *
0.86 (0.06) *°
0.38 (0.39) °
0.53 (0.30) *
0.82 (0.14) *
0.61 (0.32) *
0.84 (0.08) *°
−0.09 (0.18) °
0.20 (0.22) *
Gastrocnemius Medial Head
0.88 (0.07) *°
0.32 (0.26) °
0.41 (0.24) *
Gastrocnemius Lateral Head
0.92 (0.07) *°
0.32 (0.26) °
0.22 (0.32) *
0.89 (0.05) *°
0.33 (0.34) °
0.55 (0.26) *
0.77 (0.19) *
0.56 (0.35) *
0.85 (0.07) *°
−0.05 (0.36) °
0.27 (0.28) *
Gastrocnemius Medial Head
0.88 (0.07) *°
0.48 (0.28) °
0.53 (0.15) *
Gastrocnemius Lateral Head
0.91 (0.09) *°
0.40 (0.40) °
0.46 (0.34) *
0.89 (0.07) *°
0.31 (0.26) °
0.72 (0.13) *
0.75 (0.18) *
0.45 (0.30) *
Upper Leg EMG during walking
Inter-stride variability of EMG during walking
Variance-to-signal ratios for 1.0 m/s walking
Controls mean (sd)
Amputees mean (sd)
Gastrocnemius Medial Head
0.8 (0.2) *
5.2 (7.4) *
Gastrocnemius Lateral Head
EMG median frequencies
EMG median frequencies
Maximum voluntary activation
Treadmill walking (1.0 m/s)
Controls mean (sd)
Amputees mean (sd)
Controls mean (sd)
Amputees mean (sd)
153 (14) *°
127 (23) *
115 (22) °
Gastrocnemius Medial Head
174 (23) *°
137 (26) *
131 (18) °
Gastrocnemius Lateral Head
166 (23) *°
124 (34) *
122 (14) °
The main finding of this study is that during walking, most amputee subjects had residual lower leg muscle activation patterns that were entrained to the gait cycle but highly variable across subjects. The residual lower leg muscle activation patterns were very different from the normal control patterns (Figure 5). This is evidenced by the low EMG cross-correlation values between amputee subjects and the control mean for tibialis anterior and gastrocnemii (Table 2). Despite the high variability in residual lower leg EMG patterns across amputee subjects, inter-stride variability was similar to that of control subjects. The gastrocnemius medial head was the only muscle with a variance-to-signal ratio significantly greater in the amputee group compared to the control group. This significant difference in variance-to-signal ratio between groups was due to a single amputee subject whose variance-to-signal noise ratio was magnitudes greater than other amputee subjects (Figure 5, subject A03). Subject A03 had high inter-stride variability for all three residual lower leg muscles (Figure 5). The inter-stride variability could be problematic if it continued when using a powered lower limb prosthesis under myoelectric control. However, it seems reasonable to presume that the inter-stride variability would decrease if the residual muscle activity had a functional purpose during walking (e.g., to control dynamics of a powered prosthesis). Future studies should document the variability in muscle recruitment patterns while subjects learn to use powered prostheses.
Another finding of this study is that many, but not all, amputee subjects had robust volitional control of residual lower leg muscle activation. During maximum voluntary dorsiflexion and plantar flexion, residual muscle activation profiles in several amputee subjects were similar to controls (Figure 2). The maximum activation levels were well above resting baseline, the time to reach maximum activation from resting baseline was short, and the activation levels were well sustained. Some of the amputee subjects were able to differentiate tibialis anterior and gastrocnemii activation and had coactivation levels similar to control subjects (e.g., Figure 2A, subjects A05 and A09). Other amputee subjects were not able to differentiate tibialis anterior and gastrocnemii activation during volitional maximum activation. As a result, there was either complete coactivation for both plantar flexion and dorsiflexion tasks (e.g., Figure 2A, subject A01) or an inability to recruit any muscles strongly during dorsiflexion (e.g., Figure 2A, subjects A02 and A08). For the subjects that demonstrated complete coactivation, synchronous recruitment of residual muscles was not hard-wired because their tibialis anterior and gastrocnemii activation patterns were distinctly different from each other during walking, especially at faster walking speeds (e.g., Figure 2A, subjects A01 and A02). One reason that the amputee subjects may have lost robust volitional control of the residual limb muscles is the lack of proprioceptive or visual feedback of muscle activity. Without an ankle joint to provide sensory information about joint position, there is no clear information reinforcing the consequences of muscle activity. It seems likely that coupling a powered prosthetic limb to the residual limb muscle activity would increase the volitional motor control [14, 24–26].
In the upper leg muscles, our data show that amputee subjects had greater inter-subject variability in their biceps femoris and gluteus medius muscle activation profiles compared to control subjects during walking (Table 2, Figure 6). In addition, our data show that amputee subjects had a different biceps femoris activation profile shape than control subjects (Table 2, Figure 6). Previous studies have suggested that transtibial amputees walk with greater residual leg biceps femoris activation during early stance compared to the intact biceps femoris to stabilize the knee joint [27–29] and/or increase propulsion of the residual leg [30, 31]. In normal walking, the primary function of the gluteus medius is to provide support during early stance to midstance and the biceps femoris has the potential for generating support from early stance to midstance. Ankle dorsiflexors provide support during early stance and ankle plantar flexors provide support during late stance . It is likely that transtibial amputees compensate for the loss of support from ankle muscles by recruiting muscles above the knee to increase walking stability during stance. The inter-subject variability in the biceps femoris and gluteus medius activation shape observed in our amputee subjects suggests that there are differences in compensatory muscle recruitment patterns used by transtibial amputees during walking.
One limitation of our study is that we did not present data from overground walking. Past studies have shown that lower limb EMG patterns and kinematics can be different during treadmill walking compared to overground walking [33, 34]. Biomechanically, treadmill gait and overground gait is identical if the treadmill belt speed is constant . The differences in biological gait measurements occur primarily due to two aspects: differences in visual flow  and treadmill speed fluctuations . We did not include overground walking in this study because our primary focus was to quantify differences in signal patterns and variability between amputee and non-amputee groups and within groups. Now that we have demonstrated that reliable signals can be recorded from residual muscles of transtibial amputees during treadmill walking at constant speeds, we plan to expand our study to include lower limb EMG patterns of transtibial amputees and non-amputees during overground walking at self-selected walking speeds. This will provide a better understanding of how signals recorded from residual muscles in transtibial amputees can be utilized to control robotic lower limb prostheses. Another limitation of our study is that the mean age of our amputee group was greater than our non-amputee group. We do not believe that the results presented in this study would change significantly given more similar ages between groups, but further data could support or refute this assumption.
Several previous studies have presented EMG data from the amputated limb of transtibial amputees during walking [27, 28, 30, 38], but they did not record EMG from residual limb muscles inside the socket. It has traditionally been thought that the mechanics of the socket-limb interface prevent reliable measurements of EMG from the residual limb muscles during walking with surface electrodes. Au et al. recorded EMG from residual limb muscles within the socket, but were only able to get a reliable signal during swing . We were able to record robust and reliable EMG during both stance and swing by using active EMG electrodes to maximize signal-to-noise ratio and using silicone putty to minimize movement and discomfort at the electrode sites.
Although there was the possibility for mechanical artifacts in our EMG recordings, data of EMG median frequencies suggest that we measured muscle activity from the residual limb muscles with little to no motion artifact. The EMG median frequencies recorded from the residual limb muscles during walking were similar to the EMG median frequencies recorded from the residual limb muscles during seated maximum voluntary activation trials (Table 4). In addition, the EMG median frequencies recorded from residual lower leg muscles in amputee subjects during treadmill walking were similar to the EMG median frequencies of the intact lower leg muscles in control subjects during treadmill walking (Table 4). Some of the amputee subjects demonstrated abnormal EMG patterns that had rhythmic, short-duration, and high-amplitude bursts (e.g., Figure 3, subjects A06 and A09). We do not believe that these bursts resulted from mechanical perturbations to the electrodes because of the filtering we used and the frequency content of the resulting signals. Similar EMG patterns have been demonstrated in individuals with spinal cord injury that have had long-term disuse atrophy of the muscles [39, 40]. The short-duration, high-amplitude EMG bursts that occurred around heel-strike and toe-off events may have been a result of reflex activation from muscle fiber stretch (Ia and II afferents) or rapid loading/unloading (Ib afferents).
The unique residual muscle activation patterns seen in our amputee subjects during gait suggest that neural plasticity may have occurred following amputation. Previous studies have demonstrated that neural plasticity in lower limb amputees occurs predominantly at the cortical level [41, 42]. Neural plasticity can be affected by cause of amputation (e.g. traumatic, cancer-related, dysvascular-related), age at amputation, surgical procedure, muscle atrophy, and degeneration of nerves. The long-term cortical reorganization that occurs following injury is also highly use-dependent . Changes in gait-related muscle activity following amputation would have a major impact on use-dependent cortical plasticity. Some amputees may learn to activate their residual muscles to improve stability at the limb-socket interface or to minimize socket discomfort/pain associated with impulsive prosthetic forces. This could alter the activation patterns away from the normal functional pattern seen in intact subjects and could contribute to increased inter-subject variability in amputees.
The results of this study are encouraging for the development of powered lower limb prosthesis under myoelectric control. Coupling an amputee’s nervous system to a robotic prosthesis should provide a strong stimulus for learning to modify residual muscle activation patterns. In past studies, we have found that subjects with intact musculoskeletal systems can quickly adapt their muscle activation patterns to control powered lower-limb orthoses under proportional myoelectric control [44–47]. It seems likely that amputees could also learn to modify their muscle activation patterns to control powered lower-limb prostheses, though it may take longer due to the motor plasticity that has occurred since the amputation. Residual limb muscle activation patterns during dynamic tasks such as walking may function to improve fit and/or minimize discomfort at the socket-limb interface. Learning new residual activation patterns to control lower-limb prostheses may compete with this. Future studies should investigate why amputees adopt specific residual limb muscle activation patterns in order to assess the feasibility of myoelectric control using residual limb muscles during walking. Continued technological advances in intramuscular electrodes that could transmit control EMG signals through the prosthetic socket-limb interface without breaking the skin [17–19] would provide a means for generating feedforward control signals to a robotic prosthesis from the nervous system. Another option is recent technological advances in flexible epidermal electronics that could be mounted directly on the skin within the prosthetic socket-limb interface . Either of these options could provide a long-term means for improving the control of powered lower limb prosthesis using EMG from the residual limb muscles.
It is possible to record artifact-free muscle activation patterns from residual limb muscles within the prosthetic socket-limb interface with surface electromyography electrodes. There is high inter-subject variability in recruitment patterns in amputees, but for each subject EMG patterns are consistent from stride to stride. Our results support the potential use of myoelectric controllers for direct feedforward control of robotic lower limb prostheses.
This research was funded by a Department of Defense Grant (W81XWH-09-2-0142).
- Au S, Berniker M, Herr H: Powered ankle-foot prosthesis to assist level-ground and stair-descent gaits. Neural Netw 2008, 21: 654-666. 10.1016/j.neunet.2008.03.006View ArticlePubMedGoogle Scholar
- Eilenberg MF, Geyer H, Herr H: Control of a powered ankle-foot prosthesis based on a neuromuscular model. IEEE Trans Neural Syst Rehabil Eng 2010, 18: 164-173.View ArticlePubMedGoogle Scholar
- Markowitz J, Krishnaswamy P, Eilenberg MF, Endo K, Barnhart C, Herr H: Speed adaptation in a powered transtibial prosthesis controlled with a neuromuscular model. Philos Trans R Soc Lond B Biol Sci 2011, 366: 1621-1631. 10.1098/rstb.2010.0347PubMed CentralView ArticlePubMedGoogle Scholar
- Herr HM, Grabowski AM: Bionic ankle-foot prosthesis normalizes walking gait for persons with leg amputation. Proc Biol Sci 2011, 279: 457-464.PubMed CentralView ArticlePubMedGoogle Scholar
- Martinez-Villalpando EC, Herr H: Agonist–antagonist active knee prosthesis: a preliminary study in level-ground walking. J Rehabil Res Dev 2009, 46: 361-373. 10.1682/JRRD.2008.09.0131View ArticlePubMedGoogle Scholar
- Sup F, Bohara A, Goldfarb M: Design and Control of a Powered Transfemoral Prosthesis. Int J Rob Res 2008, 27: 263-273. 10.1177/0278364907084588PubMed CentralView ArticlePubMedGoogle Scholar
- Varol HA, Sup F, Goldfarb M: Multiclass real-time intent recognition of a powered lower limb prosthesis. IEEE Trans Biomed Eng 2010, 57: 542-551.PubMed CentralView ArticlePubMedGoogle Scholar
- Sup F, Varol HA, Goldfarb M: Upslope Walking with a Powered Knee and Ankle Prosthesis: Initial Results with an Amputee Subject. Neural Systems and Rehabilitation Engineering, IEEE Transactions on 2011 2011, 1-1.Google Scholar
- Hitt J, Sugar T, Holgate M, Bellman R, Hollander K: Robotic transtibial prosthesis with biomechanical energy regeneration. Industrial Robot Int J 2009, 36: 441-447. 10.1108/01439910910980169View ArticleGoogle Scholar
- Versluys R, Beyl P, Van Damme M, Desomer A, Van Ham R, Lefeber D: Prosthetic feet: State-of-the-art review and the importance of mimicking human ankle-foot biomechanics. Disabil Rehabil Assist Technol 2009, 4: 65-75. 10.1080/17483100802715092View ArticlePubMedGoogle Scholar
- Laferrier JZ, Gailey R: Advances in lower-limb prosthetic technology. Phys Med Rehabil Clin N Am 2010, 21: 87-110. 10.1016/j.pmr.2009.08.003View ArticlePubMedGoogle Scholar
- Parker P, Englehart K, Hudgins B: Myoelectric signal processing for control of powered limb prostheses. J Electromyogr Kinesiol 2006, 16: 541-548. 10.1016/j.jelekin.2006.08.006View ArticlePubMedGoogle Scholar
- Ohnishi K, Weir RF, Kuiken TA: Neural machine interfaces for controlling multifunctional powered upper-limb prostheses. Expert Rev Med Dev 2007, 4: 43-53. 10.1586/17434422.214.171.124View ArticleGoogle Scholar
- Di Pino G, Guglielmelli E, Rossini PM: Neuroplasticity in amputees: main implications on bidirectional interfacing of cybernetic hand prostheses. Prog Neurobiol 2009, 88: 114-126. 10.1016/j.pneurobio.2009.03.001View ArticlePubMedGoogle Scholar
- Hargrove LJ, Simon AM, Lipschutz RD, Finucane SB, Kuiken TA: Real-time myoelectric control of knee and ankle motions for transfemoral amputees. JAMA 2011, 305: 1542-1544. 10.1001/jama.2011.465View ArticlePubMedGoogle Scholar
- Ha KH, Varol HA, Goldfarb M: Volitional control of a prosthetic knee using surface electromyography. IEEE Trans Biomed Eng 2011, 58: 144-151.View ArticlePubMedGoogle Scholar
- Troyk PR, DeMichele GA, Kerns DA, Weir RF: IMES: an implantable myoelectric sensor. Conf Proc IEEE Eng Med Biol Soc 2007, 2007: 1730-1733.PubMedGoogle Scholar
- Weir RF, Troyk PR, DeMichele GA, Kerns DA, Schorsch JF, Maas H: Implantable myoelectric sensors (IMESs) for intramuscular electromyogram recording. IEEE Trans Biomed Eng 2009, 56: 159-171.PubMed CentralView ArticlePubMedGoogle Scholar
- Merrill DR, Lockhart J, Troyk PR, Weir RF, Hankin DL: Development of an implantable myoelectric sensor for advanced prosthesis control. Artif Organs 2011, 35: 249-252. 10.1111/j.1525-1594.2011.01219.xPubMed CentralView ArticlePubMedGoogle Scholar
- Collins SH, Adamczyk PG, Ferris DP, Kuo AD: A simple method for calibrating force plates and force treadmills using an instrumented pole. Gait Posture 2009, 29: 59-64. 10.1016/j.gaitpost.2008.06.010PubMed CentralView ArticlePubMedGoogle Scholar
- von Tscharner V: Intensity analysis in time-frequency space of surface myoelectric signals by wavelets of specified resolution. J Electromyogr Kinesiol 2000, 10: 433-445. 10.1016/S1050-6411(00)00030-4View ArticlePubMedGoogle Scholar
- Merletti R, Parker PA: Electromyography: Physiology, engineering, and noninvasive applications. Wiley-IEEE Press; 2004.View ArticleGoogle Scholar
- Wren TA, Do KP, Rethlefsen SA, Healy B: Cross-correlation as a method for comparing dynamic electromyography signals during gait. J Biomech 2006, 39: 2714-2718. 10.1016/j.jbiomech.2005.09.006View ArticlePubMedGoogle Scholar
- Dhillon GS, Horch KW: Direct neural sensory feedback and control of a prosthetic arm. Neural Syst Rehabil Eng 2005, 13: 468-472. IEEE Transactions on 10.1109/TNSRE.2005.856072View ArticleGoogle Scholar
- Dhillon GS, Kruger TB, Sandhu JS, Horch KW: Effects of short-term training on sensory and motor function in severed nerves of long-term human amputees. J Neurophysiol 2005, 93: 2625-2633. 10.1152/jn.00937.2004View ArticlePubMedGoogle Scholar
- Dhillon GS, Lawrence SM, Hutchinson DT, Horch KW: Residual function in peripheral nerve stumps of amputees: implications for neural control of artificial limbs. J Hand Surg Am 2004, 29: 605-15. discussion 616–8 10.1016/j.jhsa.2004.02.006View ArticlePubMedGoogle Scholar
- Isakov E, Burger H, Krajnik J, Gregoric M, Marincek C: Knee muscle activity during ambulation of trans-tibial amputees. J Rehabil Med 2001, 33: 196-199. 10.1080/165019701750419572View ArticlePubMedGoogle Scholar
- Isakov E, Keren O, Benjuya N: Trans-tibial amputee gait: time-distance parameters and EMG activity. Prosthet Orthot Int 2000, 24: 216-220. 10.1080/03093640008726550View ArticlePubMedGoogle Scholar
- Winter DA, Sienko SE: Biomechanics of below-knee amputee gait. J Biomech 1988, 21: 361-367. 10.1016/0021-9290(88)90142-XView ArticlePubMedGoogle Scholar
- Fey NP, Silverman AK, Neptune RR: The influence of increasing steady-state walking speed on muscle activity in below-knee amputees. J Electromyogr Kinesiol 2010, 20: 155-161. 10.1016/j.jelekin.2009.02.004View ArticlePubMedGoogle Scholar
- Silverman AK, Fey NP, Portillo A, Walden JG, Bosker G, Neptune RR: Compensatory mechanisms in below-knee amputee gait in response to increasing steady-state walking speeds. Gait Posture 2008, 28: 602-609. 10.1016/j.gaitpost.2008.04.005View ArticlePubMedGoogle Scholar
- Anderson FC, Pandy MG: Individual muscle contributions to support in normal walking. Gait Posture 2003, 17: 159-169. 10.1016/S0966-6362(02)00073-5View ArticlePubMedGoogle Scholar
- Kautz SA, Bowden MG, Clark DJ, Neptune RR: Comparison of motor control deficits during treadmill and overground walking poststroke. Neurorehabil Neural Repair 2011, 25: 756-765. 10.1177/1545968311407515PubMed CentralView ArticlePubMedGoogle Scholar
- Lee SJ, Hidler J: Biomechanics of overground vs. treadmill walking in healthy individuals. J Appl Physiol 2008, 104: 747-755. 10.1152/japplphysiol.01380.2006View ArticlePubMedGoogle Scholar
- van Ingen Schenau GJ: Some fundamental aspects of the biomechanics of overground versus treadmill locomotion. Med Sci Sports Exerc 1980, 12: 257-261.View ArticlePubMedGoogle Scholar
- Warren WHJ, Kay BA, Zosh WD, Duchon AP, Sahuc S: Optic flow is used to control human walking. Nat Neurosci 2001, 4: 213-216. 10.1038/84054View ArticlePubMedGoogle Scholar
- Savelberg HH, Vorstenbosch MA, Kamman EH, van de Weijer JG, Schambardt HC: Intra-stride belt-speed variation affects treadmill locomotion. Gait Posture 1998, 7: 26-34. 10.1016/S0966-6362(97)00023-4View ArticlePubMedGoogle Scholar
- Ventura JD, Klute GK, Neptune RR: The effect of prosthetic ankle energy storage and return properties on muscle activity in below-knee amputee walking. Gait Posture 2011, 33: 220-226. 10.1016/j.gaitpost.2010.11.009View ArticlePubMedGoogle Scholar
- Beres-Jones JA, Johnson TD, Harkema SJ: Clonus after human spinal cord injury cannot be attributed solely to recurrent muscle-tendon stretch. Exp Brain Res 2003, 149: 222-236.PubMedGoogle Scholar
- Gordon KE, Wu M, Kahn JH, Dhaher YY, Schmit BD: Ankle load modulates hip kinetics and EMG during human locomotion. J Neurophysiol 2009, 101: 2062-2076. 10.1152/jn.90949.2008PubMed CentralView ArticlePubMedGoogle Scholar
- Chen R, Cohen LG, Hallett M: Nervous system reorganization following injury. Neuroscience 2002, 111: 761-773. 10.1016/S0306-4522(02)00025-8View ArticlePubMedGoogle Scholar
- Chen R, Corwell B, Yaseen Z, Hallett M, Cohen LG: Mechanisms of cortical reorganization in lower-limb amputees. J Neurosci 1998, 18: 3443-3450.PubMedGoogle Scholar
- Churchill JD, Muja N, Myers WA, Besheer J, Garraghty PE: Somatotopic consolidation: a third phase of reorganization after peripheral nerve injury in adult squirrel monkeys. Exp Brain Res 1998, 118: 189-196. 10.1007/s002210050271View ArticlePubMedGoogle Scholar
- Cain SM, Gordon KE, Ferris DP: Locomotor adaptation to a powered ankle-foot orthosis depends on control method. J Neuroeng Rehabil 2007, 4: 48. 10.1186/1743-0003-4-48PubMed CentralView ArticlePubMedGoogle Scholar
- Ferris DP, Gordon KE, Sawicki GS: Peethambaran A: an improved powered ankle-foot orthosis using proportional myoelectric control. Gait Posture 2006, 23: 425-428. 10.1016/j.gaitpost.2005.05.004View ArticlePubMedGoogle Scholar
- Gordon KE, Ferris DP: Learning to walk with a robotic ankle exoskeleton. J Biomech 2007, 40: 2636-2644. 10.1016/j.jbiomech.2006.12.006View ArticlePubMedGoogle Scholar
- Kao PC, Ferris DP: Motor adaptation during dorsiflexion-assisted walking with a powered orthosis. Gait Posture 2009, 29: 230-236. 10.1016/j.gaitpost.2008.08.014PubMed CentralView ArticlePubMedGoogle Scholar
- Kim DH, Lu N, Ma R, Kim YS, Kim RH, Wang S, Wu J, Won SM, Tao H, Islam A, Yu KJ, Kim TI, Chowdhury R, Ying M, Xu L, Li M, Chung HJ, Keum H, McCormick M, Liu P, Zhang YW, Omenetto FG, Huang Y, Coleman T, Rogers JA: Epidermal electronics. Science 2011, 333: 838-843. 10.1126/science.1206157View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.