Force Myography Based Novel Strategy for Locomotion Classification

Human motion recognition with high accuracy and fast response speed has long been considered an essential component in human-machine interactive activities such as assistive robotics, medical prosthesis, and wearable electronics. The force myography (FMG) signal has been the focus of much investigation in the search for a reliable and efficient muscular locomotion recognition system. However, the effect of the sensing system on FMG-based locomotion classification accuracy has yet to be understood. This study proposed a novel FMG sensing strategy for human lower limb locomotion classification based on flexible supercapacitive iontronic sensors. Benefiting from the ultrahigh sensitivity (up to 1 nF/mmHg) and low activation pressure (less than 5 mmHg) of the supercapacitive iontronic pressure sensor, FMG signal can be acquired accurately from 5 iontronic sensors strapped to the thigh. In the experiment with 12 subjects, the real-time classification strategy based on sliding window and SVM model gave an average locomotion classification accuracy of 99% for seven categories, including sitting, standing, walking on level ground, ramp ascent, ramp descent, stair ascent, stair descent. Compared with traditional FSR sensors, the result showed that iontronic sensors improved the classification accuracy by up to 10 percentage points in the case of short time window. The implementation of the high sensitivity flexible iontronic sensors in the wearable system brings a valuable tool for detecting small human body pressure signals and has great potential to improve the performance of the human-machine interface in rehabilitation and medical applications.

Locomotion mode classification is one of the most important aspects for the control of powered lower-limb prostheses. We propose a wearable capacitive sensing system for recognizing locomotion modes as an alternative solution to popular electromyography (EMG)-based systems, aiming to overcome drawbacks of the latter. Eight able-bodied subjects and five transtibial amputees were recruited for automatic classification of six common locomotion modes. The system measured ten channels of capacitance signals from the shank, the thigh, or both. With a phase-dependent linear discriminant analysis classifier and selected time-domain features, the system can achieve a satisfactory classification accuracy of 93.6% ±0.9% and 93.4% ±0.8% for able-bodied subjects and amputee subjects, respectively. The classification accuracy is comparable with that of EMG-based systems. More importantly, we verify that neuro-mechanical delay inherent in capacitive sensing does not affect the timeliness of classification decisions as the system, similar to EMG-based systems, can make multiple judgments during a gait cycle. Experimental results also indicate that capacitance signals from the thigh alone are sufficient for mode classification for both able-bodied and transtibial subjects. Our investigations demonstrate that capacitive sensing is a promising alternative to myoelectric sensing for real-time control of powered lower-limb prostheses.

Given the higher fall risk and the fatal sequelae of falls on stairs, it is worthwhile to investigate the mechanism of dynamic balance control in individuals with knee osteoarthritis during stair negotiation. Whole-body angular momentum () is widely used as a surrogate to reflect dynamic balance and failure to constrain may increase the fall risk. This study aimed to compare the range of between people with and without knee osteoarthritis during stair ascent and descent. Seven participants with symptomatic knee osteoarthritis and eight asymptomatic controls were instructed to ascend and descend an instrumented staircase at a fixed cadence. Kinematic and kinetic data were collected and range of in sagittal, frontal, and transverse planes were computed. The knee osteoarthritis group exhibited greater in the sagittal plane during both stair ascent (P = 0.005, Cohen’s d = 1.7) and descent (P = 0.020, Cohen’s d = 1.3) as well as in the transverse plane during stair descent (P = 0.015, Cohen’s d = 1.3), than the control group. These observations may be explained by greater hip flexion (P < 0.05, Cohen’s d > 1.12) and reduced knee flexion moment (P < 0.001, Cohen’s d<-2.77) during stair ascent and descent, and decreased foot progression angle (P = 0.038, Cohen’s d=-1.2) during stair descent, in individuals with knee osteoarthritis. No significant difference in frontal plane was found between the two groups (P > 0.05). Individuals with knee osteoarthritis exhibited greater whole-body angular momentum during stair negotiation when compared to asymptomatic controls. Our findings may provide mechanistic rationale for a greater fall risk among people with knee osteoarthritis.

Force Myography (FMG) is a method of tracking functional motor activity using volumetric changes associated with muscle function. With comparable accuracy and multiple advantages over traditional methods of functional motor activity tracking, FMG has shown a promising potential in terms of applications in human-machine interfaces, tele-operation and healthcare devices. This paper provides a study that explores the effect of the spatial coverage and placement of the Force Myography (FMG) measurements on the accuracy and predictability of the machine learning models of isometric hand force. Five participants were recruited in this study and were asked to exert isometric force along three perpendicular axes while wearing custom built FMG devices. During the tests, the isometric force was measured using a 6 degree-of-freedom (DOF) load cell whereas the FMG signals were recorded using a total number of 60 FSRs, which were embedded into four bands worn on the arm. General Regression Neural Network (GRNN) model was employed in this study for predicting the hand force in three axes from the recorded FMG signals. The regression model was trained using all possible band combinations to find the optimal placement for the FMG measurements. The results showed that the accuracy significantly improved when increasing the spatial coverage from 1 FMG band to 2 or 3 bands for all axes. While the accuracy slightly improved when the 4 bands used instead of 3. Specifically, the average R2 across all subjects and axes are 0.68 ± 0.12, 0.84 ± 0.04, 0.91 ± 0.02 and 0.95 ± 0.01 using single, double, triple and four bands combination, respectively, in 5-fold cross-validation evaluation. The knowledge generated from this work aims serve as a guide towards the development of portable FMG based technology for widespread deployment in the general population.

Whole-body angular momentum during stair ascent and descent

Joint moment measurements represent an objective biomechemical parameter in joint health assessment. Inverse dynamics based on 3D motion capture data is the current 'gold standard' to estimate joint moments. Recently, machine learning combined with data measured by wearable technologies such electromyography (EMG), inertial measurement units (IMU), and electrogoniometers (GON) has been used to enable fast, easy, and low-cost measurements of joint moments. This study investigates the ability of various deep neural networks to predict lower limb joint moments merely from IMU sensors. The performance of five different deep neural networks (InceptionTimePlus, eXplainable convolutional neural network (XCM), XCMplus, Recurrent neural network (RNNplus), and Time Series Transformer (TSTPlus)) were tested to predict hip, knee, ankle, and subtalar moments using acceleration and gyroscope measurements of four IMU sensors at the trunk, thigh, shank, and foot. Multiple locomotion modes were considered including level-ground walking, treadmill walking, stair ascent, stair descent, ramp ascent, and ramp descent. We show that XCM can accurately predict lower limb joint moments using data of only four IMUs with RMSE of 0.046 ± 0.013 Nm/kg compared to 0.064 ± 0.003 Nm/kg on average for the other architectures. We found that hip, knee, and ankle joint moments predictions had a comparable RMSE with an average of 0.069 Nm/kg, while subtalar joint moments had the lowest RMSE of 0.033 Nm/kg. The real-time feedback that can be derived from the proposed method can be highly valuable for sports scientists and physiotherapists to gain insights into biomechanics, technique, and form to develop personalized training and rehabilitation programs.

Gait analysis post anterior cruciate ligament reconstruction: Knee osteoarthritis perspective

Introduction: Patellofemoral pain is an incessant lower limb musculoskeletal disorder that may be underreported in older adults. During common locomotor activities, such as when negotiating stairs, older adults (over the age of 65 years) adopt knee biomechanics reported to increase patellofemoral pain. Negotiating stairs with a challenging surface, such as uneven or slick, may place greater demand on the knee and further exacerbate joint biomechanics related to PFJ stress. Yet, it is unknown if older adults exhibit increases in patellofemoral joint (PFJ) stress when negotiating stairs with challenging surfaces. Purpose: The purpose of this study was to examine the effect of age (young and older adults) and surface (normal, slick, and uneven) on the magnitude and temporal waveform of patellofemoral joint stress during stair ascent and descent tasks. Methods: Two cohorts (12 young: ages 18-25 years; 12 older: over 65 years) had knee biomechanics quantified after they ascended and descended 18.5 cm stairs on normal, slick, and uneven surfaces at a self-selected speed. Statistical Analysis: Peak of stance (0-100%) PFJ stress and associated components (including PFJ reaction force and contact area, and knee flexion angle and moment) were submitted to a two-way RM ANOVA to test the main effects of and interaction between age (young vs old) and surface (normal, slick, and uneven). A statistical parametric mapping two-way ANOVA was used to determine main effects of and interaction between age and surface for the PFJ stress waveform. Results: During the stair ascent, older adults exhibited greater PFJ stress from 56 to 84% of stance (p &lt; 0.001), which may be attributed to the greater PFJ stress-time integral (p = 0.004) and later peak PFJ stress (p = 0.024) compared to young adults. Additionally, a significant age by surface interaction was observed for time of peak PFJ stress (p = 0.041) during stair ascent, where older adults exhibited a later peak PFJ stress compared to young adults (p = 0.008), and later peak PFJ stress compared to normal and slick surface (both: p = 0.014). Surface impacted PFJ stress waveform (all: p &lt; 0.001), but not magnitude (p &gt; 0.05) during both stair ascent and descent. During stair ascent on the uneven surface, participants exhibited smaller PFJ stress from 8 to 25% of stance compared to normal surface, but greater PFJ stress from 57 to 90% and 49 to 77% of stance compared to the normal and slick surfaces (all: p &lt; 0.001). On the uneven surface, participants exhibited a greater PFJ stress-time integral (both: p = 0.010) compared to the normal and slick surfaces. During stair descent, on the uneven surface, participants only exhibited greater PFJ stress-time integral (p = 0.017) compared to slick surface, while PFJ stress was smaller from 5 to 18% of stance, but greater stress from 92 to 99% of stance (both: p &lt; 0.001) on the slick compared to the normal surface. Conclusion: Older adults are more likely to exhibit knee biomechanics related to PFJ pain development when navigating stairs. Specifically, the larger, later PFJ stress exhibited by older adults when ascending, but not descending the stairs may increase loading of the joint’s articular cartilage and increase risk of developing PFJ pain. Yet, all participants exhibited alterations in knee biomechanics that may lead to greater PFJ stress when negotiating stairs with slick and uneven surfaces.

This study is aimed at exploring a technology that can use the human physiological information, such as Force Myography (FMG) signals to provide sensory feedback to prosthetic hand users. This is based on the principle that with the intent to move the prosthetic hand, the existing limbs in the arm recruit specific group of muscles. These muscles react with a change in the cross-sectional area; piezoelectric sensors placed on these muscles will generate a voltage (FMG signals), in response to the change in muscle volume. The correlation between the amplitude of the FMG signals and intensity of pressure on fingertips during grasping is then computed and a dynamic relation (model) is established through system identification in MATLAB. The estimated models generated a fitting accuracy of more than 80%. The model is then programmed into the Arduino microcontroller, so that a real-time and proportional force feedback is channeled to amputees through a micro actuator. Obtaining such percentages of accuracy in sensory feedback without relying on touch sensors on the prosthetic hand that could be affected by mechanical wear and other interaction factors is promising. Applying advanced signal processing and classification techniques may also refine the findings to better capture and correlate the force variations with the sensory feedback.

Modulation of the Hoffmann reflex in soleus and medial gastrocnemius during stair ascent and descent in young and older adults

One control challenge in prosthetic legs is seamless transition from one gait mode to another. User intent recognition (UIR) is a high-level controller that tells a low-level controller to switch to the identified activity mode, depending on the user’s intent and environment. We propose a new framework to design an optimal UIR system with simultaneous maximum performance and minimum complexity for gait mode recognition. We use multi-objective optimization (MOO) to find an optimal feature subset that creates a trade-off between these two conflicting objectives. The main contribution of this paper is two-fold: (1) a new gradient-based multi-objective feature selection (GMOFS) method for optimal UIR design; and (2) the application of advanced evolutionary MOO methods for UIR. GMOFS is an embedded method that simultaneously performs feature selection and classification by incorporating an elastic net in multilayer perceptron neural network training. Experimental data are collected from six subjects, including three able-bodied subjects and three transfemoral amputees. We implement GMOFS and four variants of multi-objective biogeography-based optimization (MOBBO) for optimal feature subset selection, and we compare their performances using normalized hypervolume and relative coverage. GMOFS demonstrates competitive performance compared to the four MOBBO methods. We achieve a mean classification accuracy of and with the optimal selected subset for able-bodied and amputee subjects, respectively, while using only 23% of the available features. Results thus indicate the potential of advanced optimization methods to simultaneously achieve accurate, reliable, and compact UIR for locomotion mode detection of lower-limb amputees with prostheses.

Robotic prostheses provide new opportunities to better restore lost functions than passive prostheses for trans-femoral amputees. But controlling a prosthesis device automatically for individual users in different task environments is an unsolved problem. Reinforcement learning (RL) is a naturally promising tool. For prosthesis control with a user in the loop, it is desirable that the controlled prosthesis can adapt to different task environments as quickly and smoothly as possible. However, most RL agents learn or relearn from scratch when the environment changes. To address this issue, we propose the knowledge-guided Q-learning (KG-QL) control method as a principled way for the problem. In this report, we collected and used data from two able-bodied (AB) subjects wearing a RL controlled robotic prosthetic limb walking on level ground. Our ultimate goal is to build an efficient RL controller with reduced time and data requirements and transfer knowledge from AB subjects to amputee subjects. Toward this goal, we demonstrate its feasibility by employing OpenSim, a well-established human locomotion simulator. Our results show the OpenSim simulated amputee subject improved control tuning performance over learning from scratch by utilizing knowledge transfer from AB subjects. Also in this paper, we will explore the possibility of information transfer from AB subjects to help tuning for the amputee subjects.

Human-machine-interfaces (HMI) have a key role for translating human intention into control commands to external devices. Different wearable techniques, including surface electromyography (sEMG), have been proposed for acquiring bio-signals that reveal the human intention. In this paper, we explore an easy-to-use wearable sensor device that can be used to measure force-myography (FMG) signals. We assess if FMG signals can be used to estimate isometric torque of hand pronation-supination, wrist flexion-extension or wrist radial-ulnar using a regression model. Results of our investigation report an average accuracy over 90%. The related standard deviation of 0.02 is showing consistency of data among different data collecting sessions. The proposed FMG-based device shows therefore promising performance for different future applications, which may include: monitoring the progress of patients during exercising in arm rehabilitation programs; proportional control of robotic hand prosthesis; and control of robot movements.

Understanding the hip loading environment for daily activities is useful for hip fracture prevention, rehabilitation, and the design of osteogenic exercises. Seventeen older adults (50–70 yrs) and twenty young adults (18–30 yrs) were recruited. A rigid body model combined with a musculoskeletal model was used to estimate lower extremity loading. An elliptical cross-section model of the femoral neck was used to estimate femoral neck stress during stair ascent and descent. Two peaks were identified in the stress curves, corresponding to the peaks in the vertical ground reaction force. During stair ascent, significantly higher tension on the superior femoral neck was found for the young group at peak 1 (young: 13.5±6.1 MPa, older: 4.2±6.5 MPa, p<0.001). Also during stair ascent, significantly higher compression on the posterior femoral neck was found for the older group at peak 2 (young: -11.4±4.9 MPa, old: -18.1±8.6 MPa, p = 0.006). No significant difference was found for stair descent. Components of stress (muscle vs. reaction forces; axial forces vs. bending moments) were also examined for each trial of stair ascent and descent. The stresses and their components provided loading magnitude and locations of higher stress on the femoral neck during stair ascent and descent. Understanding femoral neck stresses may be used to help prevent hip fractures, reduce pain, improve rehabilitation, and design osteogenic exercises.

Kinematics in the Terminal Swing Phase of Unilateral Transfemoral Amputees: Microprocessor-Controlled Versus Swing-Phase Control Prosthetic Knees