Inertial Sensing for Gait Event Detection and Transfemoral Prosthesis Control Strategy

Motion Smoothness Analysis of the Gait Cycle, Segmented by Stride and Associated with the Inertial Sensors' Locations

Portable monitoring devices based on Inertial Measurement Units (IMUs) have the potential to serve as quantitative assessments of human movement. This article proposes a new method to identify the optimal placements of the IMUs and quantify the smoothness of the gait. First, it identifies gait events: foot-strike (FS) and foot-off (FO). Second, it segments the signals of linear acceleration and angular velocities obtained from the IMUs at four locations into steps and strides. Finally, it applies three smoothness metrics (SPARC, PM, and LDLJ) to determine the most reliable metric and the best location for the sensor, using data from 20 healthy subjects who walked an average of 25 steps on a flat surface for this study (117 measurements were processed). All events were identified with less than a 2% difference from those obtained with the photogrammetry system. The smoothness metric with the least variance in all measurements was SPARC. For the smoothness metrics with the least variance, we found significant differences between applying the metrics with the complete signal (C) and the signal segmented by strides (S). This method is practical, time-effective, and low-cost in terms of computation. Furthermore, it is shown that analyzing gait signals segmented by strides provides more information about gait progression.

A Fusion Network With Stacked Denoise Autoencoder and Meta Learning for Lateral Walking Gait Phase Recognition and Multi-Step-Ahead Prediction

Lateral walking gait phase recognition and prediction are the premise of hip exoskeleton application in lateral resistance walk exercise. We presented a fusion network with stacked denoise autoencoder and meta learning (SDA-NN-ML) to recognize gait phase and predict gait percentage from IMU signals. Experiments were conducted to detect the four lateral walking gait phases and predict their percentage across different speeds. The performance of SDA-NN-ML and Support Vector Machine (SVM), Adaptive Boosting (AdaBoost) and Long Short Term Memory (LSTM) were evaluated. The cross-subject recognition accuracy of SDA-NN-ML (89.94%) decreased by 4.62% compared to the training accuracy, which outperformed SVM (8.60%), AdaBoost (5.61%), and LSTM (7.12%). For real-time and cross-subject prediction of gait phase percentage, the RMSE of SDA-NN-ML (0.2043) outperformed that of a single regression network (0.2426). With a signal noise ratio of 100:30, the cross-subject recognition accuracy decreased by a mere 5.70%, while the prediction result (RMSE) of SDA-NN-ML increased by 0.0167 when compared to the noise-free results. SDA-NN-ML demonstrates a stable multi-step-ahead prediction ability with an accuracy higher than 82.50% and an RMSE of less than 0.23 when the ahead time is less than 200 ms. The results demonstrated that the proposed method has high accuracy and robust performance in lateral walking gait recognition and prediction.

Gait event detection accuracy: Effects of amputee gait pattern, terrain and algorithm

Several kinematic-based algorithms have shown accuracy for gait event detection in unimpaired and pathological gait. However, their validation in subjects with lower limb amputation while walking on different terrains is still limited. The aim of this study was to evaluate the accuracy of three kinematic-based algorithms: Coordinate-Based Algorithm (CBA), Velocity-Based Algorithm (VBA) and High-Pass Filtered Algorithms (HPA) for detection of gait events in subjects with unilateral transtibial amputation walking on different terrains. Twelve subjects with unilateral transtibial amputation, using a hydraulic ankle prosthesis, walked at self-selected walking speed, on level ground and up and down a slope. Detection of Initial Contact (IC) and Foot Off (FO) by the three algorithms for intact and prosthetic limbs was compared with detection by force platforms using the True Error (TE) (time difference in detection). Mean TE found for over 100 events analysed per condition were smaller than 40 ms for both events in all conditions (approximately 6 % of stance phase). Significant interactions (p < 0.01) were found between terrain and algorithm, limb and algorithm, and also a main effect for the algorithm. Post-hoc analyses indicate that the algorithm, the limb and the terrain had an effect on the accuracy in detection. If an accuracy of 40 ms is acceptable for the particular application, then all three algorithms can be used for event detection in amputee gait. However, if accuracy in detection of events is crucial for the intended application, an evaluation of the algorithms in pathological gait walking on the terrain of interest is recommended.

Quantifying Asymmetric Gait Pattern Changes Using a Hidden Markov Model Similarity Measure (HMM-SM) on Inertial Sensor Signals

Wearable gait analysis systems using inertial sensors offer the potential for easy-to-use gait assessment in lab and free-living environments. This can enable objective long-term monitoring and decision making for individuals with gait disabilities. This study explores a novel approach that applies a hidden Markov model-based similarity measure (HMM-SM) to assess changes in gait patterns based on the gyroscope and accelerometer signals from just one or two inertial sensors. Eleven able-bodied individuals were equipped with a system which perturbed gait patterns by manipulating stance-time symmetry. Inertial sensor data were collected from various locations on the lower body to train hidden Markov models. The HMM-SM was evaluated to determine whether it corresponded to changes in gait as individuals deviated from their baseline, and whether it could provide a reliable measure of gait similarity. The HMM-SM showed consistent changes in accordance with stance-time symmetry in the following sensor configurations: pelvis, combined upper leg signals, and combined lower leg signals. Additionally, the HMM-SM demonstrated good reliability for the combined upper leg signals (ICC = 0.803) and lower leg signals (ICC = 0.795). These findings provide preliminary evidence that the HMM-SM could be useful in assessing changes in overall gait patterns. This could enable the development of compact, wearable systems for unsupervised gait assessment, without the requirement to pre-identify and measure a set of gait parameters.

Experimental Evaluation of Machine Learning Models for Gait Segmentation

Accurate estimation of the gait phase is extremely important in exoskeleton control, as various stages in the gait cycle require different control objectives. This study involved the evaluation of seven machine learning models utilizing inertial measurement unit and joint angle data from eight healthy subjects wearing the Ekso Indego exoskeleton for the purpose of gait segmentation. During the experiments, subjects walked on a level instrumented treadmill with gravity compensation assistance for three trials and underwent a fourth trial simulating impairment. A six-state model of gait was employed, where bilateral heel strike, toe off, and tibia vertical determined state transitions. True state transitions were determined using optical motion capture and ground reaction force data. Of the evaluated models, the Support Vector Machine achieved the highest performance, with an average accuracy of 94.5% and 94.1% for normal walking and impaired walking, respectively. Future research should focus on assessing the model's real-time performance among exoskeleton users before considering its application as the basis for an exoskeleton controller.

Bilateral Elimination Rule-Based Finite Class Bayesian Inference System for Circular and Linear Walking Prediction

OBJECTIVE

The prediction of upcoming circular walking during linear walking is important for the usability and safety of the interaction between a lower limb assistive device and the wearer. This study aims to build a bilateral elimination rule-based finite class Bayesian inference system (BER-FC-BesIS) with the ability to predict the transition between circular walking and linear walking using inertial measurement units.

METHODS

Bilateral motion data of the human body were used to improve the recognition and prediction accuracy of BER-FC-BesIS.

RESULTS

The mean predicted time of BER-FC-BesIS in predicting the left and right lower limbs' upcoming steady walking activities is 119.32 ± 9.71 ms and 113.75 ± 11.83 ms, respectively. The mean time differences between the predicted time and the real time of BER-FC-BesIS in the left and right lower limbs' prediction are 14.22 ± 3.74 ms and 13.59 ± 4.92 ms, respectively. The prediction accuracy of BER-FC-BesIS is 93.98%.

CONCLUSION

Upcoming steady walking activities (e.g., linear walking and circular walking) can be accurately predicted by BER-FC-BesIS innovatively.

SIGNIFICANCE

This study could be helpful and instructional to improve the lower limb assistive devices' capabilities of walking activity prediction with emphasis on non-linear walking activities in daily living.

A Novel Gait Event Detection Algorithm Using a Thigh-Worn Inertial Measurement Unit and Joint Angle Information

This paper describes the development and evaluation of a novel, threshold-based gait event detection algorithm utilizing only one thigh inertial measurement unit (IMU) and unilateral, sagittal plane hip and knee joint angles. The algorithm was designed to detect heel strike (HS) and toe off (TO) gait events, with the eventual goal of detection in a real-time exoskeletal control system. The data used in the development and evaluation of the algorithm were obtained from two gait databases, each containing synchronized IMU and ground reaction force (GRF) data. All database subjects were healthy individuals walking in either a level-ground, urban environment or a treadmill lab environment. Inertial measurements used were three-dimensional thigh accelerations and three-dimensional thigh angular velocities. Parameters for the TO algorithm were identified on a per-subject basis. The GRF data were utilized to validate the algorithm's timing accuracy and quantify the fidelity of the algorithm, measured by the F1-Score. Across all participants, the algorithm reported a mean timing error of -41±20 ms with an F1-Score of 0.988 for HS. For TO, the algorithm reported a mean timing error of -1.4±21 ms with an F1-Score of 0.991. The results of this evaluation suggest that this algorithm is a promising solution to inertial based gait event detection; however, further refinement and real-time evaluation are required for use in exoskeletal control.

Inertial Measuring System to Evaluate Gait Parameters and Dynamic Alignments for Lower-Limb Amputation Subjects

The study aims to construct an inertial measuring system for the application of amputee subjects wearing a prosthesis. A new computation scheme to process inertial data by installing seven wireless inertial sensors on the lower limbs was implemented and validated by comparing it with an optical motion capture system. We applied this system to amputees to verify its performance for gait analysis. The gait parameters are evaluated to objectively assess the amputees' prosthesis-wearing status. The Madgwick algorithm was used in the study to correct the angular velocity deviation using acceleration data and convert it to quaternion. Further, the zero-velocity update method was applied to reconstruct patients' walking trajectories. The combination of computed walking trajectory with pelvic and lower limb joint motion enables sketching the details of motion via a stickman that helps visualize and animate the walk and gait of a test subject. Five participants with above-knee (n = 2) and below-knee (n = 3) amputations were recruited for gait analysis. Kinematic parameters were evaluated during a walking test to assess joint alignment and overall gait characteristics. Our findings support the feasibility of employing simple algorithms to achieve accurate and precise joint angle estimation and gait parameters based on wireless inertial sensor data.

Determination of Gait Events and Temporal Gait Parameters for Persons with a Knee-Ankle-Foot Orthosis

Gait event detection is essential for controlling an orthosis and assessing the patient's gait. In this study, patients wearing an electromechanical (EM) knee-ankle-foot orthosis (KAFO) with a single IMU embedded in the thigh were subjected to gait event detection. The algorithm detected four essential gait events (initial contact (IC), toe off (TO), opposite initial contact (OIC), and opposite toe off (OTO)) and determined important temporal gait parameters such as stance/swing time, symmetry, and single/double limb support. These gait events were evaluated through gait experiments using four force plates on healthy adults and a hemiplegic patient who wore a one-way clutch KAFO and a pneumatic cylinder KAFO. Results showed that the smallest error in gait event detection was found at IC, and the largest error rate was observed at opposite toe off (OTO) with an error rate of -2.8 ± 1.5% in the patient group. Errors in OTO detection resulted in the largest error in determining the single limb support of the patient with an error of 5.0 ± 1.5%. The present study would be beneficial for the real-time continuous monitoring of gait events and temporal gait parameters for persons with an EM KAFO.

Robust and adaptive terrain classification and gait event detection system

Real-time gait event detection (GED) system can be utilized for gait analysis and tracking fitness activities. GED for various types of terrains (e.g., stair-walk, uneven surfaces, etc.) is still an open research problem. This study presents an inertial sensor-based approach for real-time GED system that works for diverse terrains in an uncontrolled environment. The GED system classifies three types of terrains, i.e., flat-walk, stair-ascend and stair-descend, with an average classification accuracy of 99%. It also accurately detects various gait events, including, toe-strike, heel-rise, toe-off, and heel-strike. It is computationally efficient, implemented on a low-cost microcontroller, works in real-time and can be used in portable rehabilitation devices for use in dynamic environments.

Classifying Changes in Amputee Gait following Physiotherapy Using Machine Learning and Continuous Inertial Sensor Signals

Wearable sensors allow for the objective analysis of gait and motion both in and outside the clinical setting. However, it remains a challenge to apply such systems to highly diverse patient populations, including individuals with lower-limb amputations (LLA) that present with unique gait deviations and rehabilitation goals. This paper presents the development of a novel method using continuous gyroscope data from a single inertial sensor for person-specific classification of gait changes from a physiotherapist-led gait training session. Gyroscope data at the thigh were collected using a wearable gait analysis system for five LLA before, during, and after completing a gait training session. Data from able-bodied participants receiving no intervention were also collected. Models using dynamic time warping (DTW) and Euclidean distance in combination with the nearest neighbor classifier were applied to the gyroscope data to classify the pre- and post-training gait. The model achieved an accuracy of 98.65% ± 0.69 (Euclidean) and 98.98% ± 0.83 (DTW) on pre-training and 95.45% ± 6.20 (Euclidean) and 94.18% ± 5.77 (DTW) on post-training data across the participants whose gait changed significantly during their session. This study provides preliminary evidence that continuous angular velocity data from a single gyroscope could be used to assess changes in amputee gait. This supports future research and the development of wearable gait analysis and feedback systems that are adaptable to a broad range of mobility impairments.

A pilot case series for concurrent validation of inertial measurement units to motion capture in individuals who use unilateral lower-limb prostheses

Introduction

Inertial measurement units (IMUs) may be viable options to collect gait data in clinics. This study compared IMU to motion capture data in individuals who use unilateral lower-limb prostheses.

Methods

Participants walked with lower-body IMUs and reflective markers in a motion analysis space. Sagittal plane hip, knee, and ankle waveforms were extracted for the entire gait cycle. Discrete points of peak flexion, peak extension, and range of motion were extracted from the waveforms. Stance times were also extracted to assess the IMU software's accuracy at detecting gait events. IMU and motion capture-derived data were compared using absolute differences and root mean square error (RMSE).

Results

Five individuals (n = 3 transtibial; n = 2 transfemoral) participated. IMU prosthetic limb data was similar to motion capture (RMSE: waveform ≤4.65°; discrete point ≤9.04°; stance ≤0.03s). However, one transfemoral participant had larger differences at the microprocessor knee joint (RMSE: waveform ≤15.64°; discrete ≤29.21°) from IMU magnetometer interference. Intact limbs tended to have minimal differences between IMU and motion capture data (RMSE: waveform ≤6.33°; discrete ≤9.87°; stance ≤0.04s).

Conclusion

Findings from this pilot study suggest IMUs have the potential to collect data similar to motion capture systems in sagittal plane kinematics and stance time.

Comparison of machine learning and deep learning-based methods for locomotion mode recognition using a single inertial measurement unit

Locomotion mode recognition provides the prosthesis control with the information on when to switch between different walking modes, whereas the gait phase detection indicates where we are in the gait cycle. But powered prostheses often implement a different control strategy for each locomotion mode to improve the functionality of the prosthesis. Existing studies employed several classical machine learning methods for locomotion mode recognition. However, these methods were less effective for data with complex decision boundaries and resulted in misclassifications of motion recognition. Deep learning-based methods potentially resolve these limitations as it is a special type of machine learning method with more sophistication. Therefore, this study evaluated three deep learning-based models for locomotion mode recognition, namely recurrent neural network (RNN), long short-term memory (LSTM) neural network, and convolutional neural network (CNN), and compared the recognition performance of deep learning models to the machine learning model with random forest classifier (RFC). The models are trained from data of one inertial measurement unit (IMU) placed on the lower shanks of four able-bodied subjects to perform four walking modes, including level ground walking (LW), standing (ST), and stair ascent/stair descent (SA/SD). The results indicated that CNN and LSTM models outperformed other models, and these models were promising for applying locomotion mode recognition in real-time for robotic prostheses.

Rules-Based Real-Time Gait Event Detection Algorithm for Lower-Limb Prosthesis Users during Level-Ground and Ramp Walking

Real-time gait event detection (GED) using inertial sensors is important for applications such as remote gait assessments, intelligent assistive devices including microprocessor-based prostheses or exoskeletons, and gait training systems. GED algorithms using acceleration and/or angular velocity signals achieve reasonable performance; however, most are not suited for real-time applications involving clinical populations walking in free-living environments. The aim of this study was to develop and evaluate a real-time rules-based GED algorithm with low latency and high accuracy and sensitivity across different walking states and participant groups. The algorithm was evaluated using gait data collected from seven able-bodied (AB) and seven lower-limb prosthesis user (LLPU) participants for three walking states (level-ground walking (LGW), ramp ascent (RA), ramp descent (RD)). The performance (sensitivity and temporal error) was compared to a validated motion capture system. The overall sensitivity was 98.87% for AB and 97.05% and 93.51% for LLPU intact and prosthetic sides, respectively, across all walking states (LGW, RA, RD). The overall temporal error (in milliseconds) for both FS and FO was 10 (0, 20) for AB and 10 (0, 25) and 10 (0, 20) for the LLPU intact and prosthetic sides, respectively, across all walking states. Finally, the overall error (as a percentage of gait cycle) was 0.96 (0, 1.92) for AB and 0.83 (0, 2.08) and 0.83 (0, 1.66) for the LLPU intact and prosthetic sides, respectively, across all walking states. Compared to other studies and algorithms, the herein-developed algorithm concurrently achieves high sensitivity and low temporal error with near real-time detection of gait in both typical and clinical populations walking over a variety of terrains.

IMU-Based Classification of Locomotion Modes, Transitions, and Gait Phases with Convolutional Recurrent Neural Networks

This paper focuses on the classification of seven locomotion modes (sitting, standing, level ground walking, ramp ascent and descent, stair ascent and descent), the transitions among these modes, and the gait phases within each mode, by only using data in the frequency domain from one or two inertial measurement units. Different deep neural network configurations are investigated and compared by combining convolutional and recurrent layers. The results show that a system composed of a convolutional neural network followed by a long short-term memory network is able to classify with a mean F1-score of 0.89 and 0.91 for ten healthy subjects, and of 0.92 and 0.95 for one osseointegrated transfemoral amputee subject (excluding the gait phases because they are not labeled in the data-set), using one and two inertial measurement units, respectively, with a 5-fold cross-validation. The promising results obtained in this study pave the way for using deep learning for the control of transfemoral prostheses with a minimum number of inertial measurement units.

Estimation of Gait Parameters for Transfemoral Amputees Using Lower Limb Kinematics and Deterministic Algorithms

Accurate estimation of gait parameters depends on the prediction of key gait events of heel strike (HS) and toe-off (TO). Kinematics-based gait event estimation has shown potential in this regard, particularly using leg and foot velocity signals and gyroscopic sensors. However, existing algorithms demonstrate a varying degree of accuracy for different populations. Moreover, the literature lacks evidence for their validity for the amputee population. The purpose of this study is to evaluate this paradigm to predict TO and HS instants and to propose a new algorithm for gait parameter estimation for the amputee population. An open data set containing marker data of 12 subjects with unilateral transfemoral amputation during treadmill walking was used, containing around 3400 gait cycles. Five deterministic algorithms detecting the landmarks (maxima, minima, and zero-crossings [ZC]) in the foot, shank, and thigh angular velocity data indicating HS and TO events were implemented and their results compared against the reference data. Two algorithms based on foot and shank velocity minima performed exceptionally well for the HS prediction, with median accuracy in the range of 6–13 ms. However, both these algorithms produced inferior accuracy for the TO event with consistent early prediction. The peak in the thigh velocity produced the best result for the TO prediction with <25 ms median error. By combining the HS prediction using shank velocity and TO prediction from the thigh velocity, the algorithm produced the best results for temporal gait parameters (step, stride times, stance, and double support timings) with a median error of less than 25 ms. In conclusion, combined shank and thigh velocity-based prediction leads to improved gait parameter estimation than traditional algorithms for the amputee population.

Real-Time Gait Phase Detection Using Wearable Sensors for Transtibial Prosthesis Based on a kNN Algorithm

Those with disabilities who have lost their legs must use a prosthesis to walk. However, traditional prostheses have the disadvantage of being unable to move and support the human gait because there are no mechanisms or algorithms to control them. This makes it difficult for the wearer to walk. To overcome this problem, we developed an insole device with a wearable sensor for real-time gait phase detection based on the kNN (k-nearest neighbor) algorithm for prosthetic control. The kNN algorithm is used with the raw data obtained from the pressure sensors in the insole to predict seven walking phases, i.e., stand, heel strike, foot flat, midstance, heel off, toe-off, and swing. As a result, the predictive decision in each gait cycle to control the ankle movement of the transtibial prosthesis improves with each walk. The results in this study can provide 81.43% accuracy for gait phase detection, and can control the transtibial prosthetic effectively at the maximum walking speed of 6 km/h. Moreover, this insole device is small, lightweight and unaffected by the physical factors of the wearer.

Estimation of gait parameters using leg velocity for amputee population

Quantification of key gait parameters plays an important role in assessing gait deficits in clinical research. Gait parameter estimation using lower-limb kinematics (mainly leg velocity data) has shown promise but lacks validation for the amputee population. The aim of this study is to assess the accuracy of lower-leg angular velocity to predict key gait events (toe-off and heel strike) and associated temporal parameters for the amputee population. An open data set of reflexive markers during treadmill walking from 10 subjects with unilateral transfemoral amputation was used. A rule-based dual-minima algorithm was developed to detect the landmarks in the shank velocity signal indicating toe-off and heel strike events. Four temporal gait parameters were also estimated (step time, stride time, stance and swing duration). These predictions were compared against the force platform data for 3000 walking cycles from 239 walking trials. Considerable accuracy was achieved for the HS event as well as for step and stride timings, with mean errors ranging from 0 to -13ms. The TO prediction exhibited a larger error with its mean ranging from 35-81ms. The algorithm consistently predicted the TO earlier than the actual event, resulting in prediction errors in stance and swing timings. Significant differences were found between the prediction for sound and prosthetic legs, with better TO accuracy on the prosthetic side. The prediction accuracy also appeared to improve with the subjects’ mobility level (K-level). In conclusion, the leg velocity profile, coupled with the dual-minima algorithm, can predict temporal parameters for the transfemoral amputee population with varying degrees of accuracy.

Recent Trends and Practices Toward Assessment and Rehabilitation of Neurodegenerative Disorders: Insights From Human Gait

The study of human movement and biomechanics forms an integral part of various clinical assessments and provides valuable information toward diagnosing neurodegenerative disorders where the motor symptoms predominate. Conventional gait and postural balance analysis techniques like force platforms, motion cameras, etc., are complex, expensive equipment requiring specialist operators, thereby posing a significant challenge toward translation to the clinics. The current manuscript presents an overview and relevant literature summarizing the umbrella of factors associated with neurodegenerative disorder management: from the pathogenesis and motor symptoms of commonly occurring disorders to current alternate practices toward its quantification and mitigation. This article reviews recent advances in technologies and methodologies for managing important neurodegenerative gait and balance disorders, emphasizing assessment and rehabilitation/assistance. The review predominantly focuses on the application of inertial sensors toward various facets of gait analysis, including event detection, spatiotemporal gait parameter measurement, estimation of joint kinematics, and postural balance analysis. In addition, the use of other sensing principles such as foot-force interaction measurement, electromyography techniques, electrogoniometers, force-myography, ultrasonic, piezoelectric, and microphone sensors has also been explored. The review also examined the commercially available wearable gait analysis systems. Additionally, a summary of recent progress in therapeutic approaches, viz., wearables, virtual reality (VR), and phytochemical compounds, has also been presented, explicitly targeting the neuro-motor and functional impairments associated with these disorders. Efforts toward therapeutic and functional rehabilitation through VR, wearables, and different phytochemical compounds are presented using recent examples of research across the commonly occurring neurodegenerative conditions [viz., Parkinson's disease (PD), Alzheimer's disease (AD), multiple sclerosis, Huntington's disease (HD), and amyotrophic lateral sclerosis (ALS)]. Studies exploring the potential role of Phyto compounds in mitigating commonly associated neurodegenerative pathologies such as mitochondrial dysfunction, α-synuclein accumulation, imbalance of free radicals, etc., are also discussed in breadth. Parameters such as joint angles, plantar pressure, and muscle force can be measured using portable and wearable sensors like accelerometers, gyroscopes, footswitches, force sensors, etc. Kinetic foot insoles and inertial measurement tools are widely explored for studying kinematic and kinetic parameters associated with gait. With advanced correlation algorithms and extensive RCTs, such measurement techniques can be an effective clinical and home-based monitoring and rehabilitation tool for neuro-impaired gait. As evident from the present literature, although the vast majority of works reported are not clinically and extensively validated to derive a firm conclusion about the effectiveness of such techniques, wearable sensors present a promising impact toward dealing with neurodegenerative motor disorders.

Locomotion Mode Transition Prediction Based on Gait-Event Identification Using Wearable Sensors and Multilayer Perceptrons

People walk on different types of terrain daily; for instance, level-ground walking, ramp and stair ascent and descent, and stepping over obstacles are common activities in daily life. Movement patterns change as people move from one terrain to another. The prediction of transitions between locomotion modes is important for developing assistive devices, such as exoskeletons, as the optimal assistive strategies may differ for different locomotion modes. The prediction of locomotion mode transitions is often accompanied by gait-event detection that provides important information during locomotion about critical events, such as foot contact (FC) and toe off (TO). In this study, we introduce a method to integrate locomotion mode prediction and gait-event identification into one machine learning framework, comprised of two multilayer perceptrons (MLP). Input features to the framework were from fused data from wearable sensors—specifically, electromyography sensors and inertial measurement units. The first MLP successfully identified FC and TO, FC events were identified accurately, and a small number of misclassifications only occurred near TO events. A small time difference (2.5 ms and −5.3 ms for FC and TO, respectively) was found between predicted and true gait events. The second MLP correctly identified walking, ramp ascent, and ramp descent transitions with the best aggregate accuracy of 96.3%, 90.1%, and 90.6%, respectively, with sufficient prediction time prior to the critical events. The models in this study demonstrate high accuracy in predicting transitions between different locomotion modes in the same side’s mid- to late stance of the stride prior to the step into the new mode using data from EMG and IMU sensors. Our results may help assistive devices achieve smooth and seamless transitions in different locomotion modes for those with motor disorders.

Noninvasive Human-Prosthesis Interfaces for Locomotion Intent Recognition: A Review

The lower-limb robotic prostheses can provide assistance for amputees' daily activities by restoring the biomechanical functions of missing limb(s). To set proper control strategies and develop the corresponding controller for robotic prosthesis, a prosthesis user's intent must be acquired in time, which is still a major challenge and has attracted intensive attentions. This work focuses on the robotic prosthesis user's locomotion intent recognition based on the noninvasive sensing methods from the recognition task perspective (locomotion mode recognition, gait event detection, and continuous gait phase estimation) and reviews the state-of-the-art intent recognition techniques in a lower-limb prosthesis scope. The current research status, including recognition approach, progress, challenges, and future prospects in the human's intent recognition, has been reviewed. In particular for the recognition approach, the paper analyzes the recent studies and discusses the role of each element in locomotion intent recognition. This work summarizes the existing research results and problems and contributes a general framework for the intent recognition based on lower-limb prosthesis.

Machine-Learning-Based Prediction of Gait Events From EMG in Cerebral Palsy Children

Machine-learning techniques are suitably employed for gait-event prediction from only surface electromyographic (sEMG) signals in control subjects during walking. Nevertheless, a reference approach is not available in cerebral-palsy hemiplegic children, likely due to the large variability of foot-floor contacts. This study is designed to investigate a machine-learning-based approach, specifically developed to binary classify gait events and to predict heel-strike (HS) and toe-off (TO) timing from sEMG signals in hemiplegic-child walking. To this objective, sEMG signals are acquired from five hemiplegic-leg muscles in nearly 2500 strides from 20 hemiplegic children, acknowledged as Winters' group 1 and 2. sEMG signals, segmented in overlapping windows of 600 samples (pace = 5 samples), are used to train a multi-layer perceptron model. Intra-subject and inter-subject experimental settings are tested. The best-performing intra-subject approach is able to provide in the hemiplegic population a mean classification accuracy (±SD) of 0.97±0.01 and a suitable prediction of HS and TO events, in terms of average mean absolute error (MAE, 14.8±3.2 ms for HS and 17.6±4.2 ms for TO) and F1-score (0.95±0.03 for HS and 0.92±0.07 for TO). These results outperform previous sEMG-based attempts in cerebral-palsy populations and are comparable with outcomes achieved by reference approaches in control populations. In conclusion, the findings of the study prove the feasibility of neural networks in predicting the two main gait events using surface EMG signals, also in condition of high variability of the signal to predict as in hemiplegic cerebral palsy.

Derivative Based Gait Event Detection Algorithm Using Unfiltered Accelerometer Signals

Wearable sensors have been investigated for the purpose of gait analysis, namely gait event detection. Many types of algorithms have been developed specifically using inertial sensor data for detecting gait events. Though much attention has turned toward machine learning algorithms, most of these approaches suffer from large computational requirements and are not yet suitable for real-time applications such as in prostheses or for feedback control. Current rules-based algorithms for real-time use often require fusion of multiple sensor signals to achieve high accuracy, thus increasing complexity and decreasing usability of the instrument. We present our results of a novel, rules-based algorithm using a single accelerometer signal from the foot to reliably detect heel-strike and toe-off events. Using the derivative of the raw accelerometer signal and applying an optimizer and windowing approach, high performance was achieved with a sensitivity and specificity of 94.32% and 94.70% respectively, and a timing error of 6.52 ± 22.37 ms, including trials involving multiple speed transitions. This would enable development of a compact wearable system for robust gait analysis in real-world settings, providing key insights into gait quality with the capability for real-time system control.

An Adaptive Method for Gait Event Detection of Gait Rehabilitation Robots

Accurate gait event detection is necessary for control strategies of gait rehabilitation robots. However, due to personal diversity between individuals, it is a challenge for robots to detect a gait event at various stride frequencies. This paper proposes a novel method for gait event detection of a gait rehabilitation robot using a single inertial sensor mounted on the thigh. A self-adaptive threshold for detecting heel strike is obtained in real time via a linear regression model. Observable thresholds for toe off detection are constant at various stride frequencies. Experiments are conducted based on 20 healthy subjects and six hemiplegic patients wearing a gait rehabilitation robot and walking at various kinds of stride frequencies. The experimental results show that the proposed method can detect heel strike and toe off gait events within an average 2% gait cycle temporal errors and never miss two-gait event detection. Compared to the conventional thresholding method, this work presents a simple and robust application for gait event detection in healthy and hemiplegic subjects by one inertial sensor. The linear regression model can be applicable to different subjects walking at various stride frequencies.

A Review of Gait Phase Detection Algorithms for Lower Limb Prostheses

Fast and accurate gait phase detection is essential to achieve effective powered lower-limb prostheses and exoskeletons. As the versatility but also the complexity of these robotic devices increases, the research on how to make gait detection algorithms more performant and their sensing devices smaller and more wearable gains interest. A functional gait detection algorithm will improve the precision, stability, and safety of prostheses, and other rehabilitation devices. In the past years the state-of-the-art has advanced significantly in terms of sensors, signal processing, and gait detection algorithms. In this review, we investigate studies and developments in the field of gait event detection methods, more precisely applied to prosthetic devices. We compared advantages and limitations between all the proposed methods and extracted the relevant questions and recommendations about gait detection methods for future developments.

IMU-Based Locomotion Mode Identification for Transtibial Prostheses, Orthoses, and Exoskeletons

Active transtibial prostheses, orthoses, and exoskeletons hold the promise of improving the mobility of lower-limb impaired or amputated individuals. Locomotion mode identification (LMI) is essential for these devices precisely reproducing the required function in different terrains. In this study, a terrain geometry-based LMI algorithm is proposed. The environment should be built according to the inclination grade of the ground. For example, when the inclination angle is between 7 degrees and 15 degrees, the environment should be a ramp. If the inclination angle is around 30 degrees, the environment is preferred to be equipped with stairs. Given that, the locomotion mode/terrain can be classified by the inclination grade. Besides, human feet always move along the surface of terrain to minimize the energy expenditure for transporting lower limbs and get the required foot clearance. Hence, the foot trajectory estimated by an IMU was used to derive the inclination grade of the terrain that we traverse to identify the locomotion mode. In addition, a novel trigger condition (an elliptical boundary) is proposed to activate the decision-making of the LMI algorithm before the next foot strike thus leaving enough time for performing preparatory work in the swing phase. When the estimated foot trajectory goes across the elliptical boundary, the decision-making will be executed. Experimental results show that the average accuracy for three healthy subjects and three below-knee amputees is 98.5% in five locomotion modes: level-ground walking, up slope, down slope, stair descent, and stair ascent. Besides, all the locomotion modes can be identified before the next foot strike.

Gait event detection using inertial measurement units in people with transfemoral amputation: a comparative study

In recent years, inertial measurement units (IMUs) have been proposed as an alternative to force platforms and pressure sensors for gait events (i.e., initial and final contacts) detection. While multiple algorithms have been developed, the impact of gait event timing errors on temporal parameters and asymmetry has never been investigated in people with transfemoral amputation walking freely on level ground. In this study, five algorithms were comparatively assessed on gait data of seven people with transfemoral amputation, equipped with three IMUs mounted at the pelvis and both shanks, using pressure insoles for reference. Algorithms' performance was first quantified in terms of gait event detection rate (sensitivity, positive predictive value). Only two algorithms, based on shank mounted IMUs, achieved an acceptable detection rate (positive predictive value > 99%). For these two, accuracy of gait events timings, temporal parameters, and absolute symmetry index of stance-phase duration (SPD-ASI) were assessed. Whereas both algorithms achieved high accuracy for stride duration estimates (median errors: 0%, interquartile ranges < 1.75%), lower accuracy was found for other temporal parameters due to relatively high errors in the detection of final contact events. Furthermore, SPD-ASI derived from IMU-based algorithms proved to be significantly different to that obtained from insoles data. Graphical abstract Gait event detection with IMU in people with transfemoral amputation: initial contact (IC) and final contact (FC) events at the sound (s) and prosthetic (p) side are identified. Five algorithms were implemented using either shank-mounted or pelvis-mounted IMUs. Gait events were used to estimate temporal parameters (stride duration, stance phase duration [SPD], and double support time) and SPD asymmetry.

Design and Speed-Adaptive Control of a Powered Geared Five-Bar Prosthetic Knee Using BP Neural Network Gait Recognition

To improve the multi-speed adaptability of the powered prosthetic knee, this paper presented a speed-adaptive neural network control based on a powered geared five-bar (GFB) prosthetic knee. The GFB prosthetic knee is actuated via a cylindrical cam-based nonlinear series elastic actuator that can provide the desired actuation for level-ground walking, and its attitude measurement is realized by two inertial sensors and one load cell on the prosthetic knee. To improve the performance of the control system, the motor control and the attitude measurement of the GFB prosthetic knee are run in parallel. The BP neural network uses input data from only the GFB prosthetic knee, and is trained by natural and artificially modified various gait patterns of different able-bodied subjects. To realize the speed-adaptive control, the prosthetic knee speed and gait cycle percentage are identified by the Gaussian mixture model-based gait classifier. Specific knee motion control instructions are generated by matching the neural network predicted gait percentage with the ideal walking gait. Habitual and variable speed level-ground walking experiments are conducted via an able-bodied subject, and the experimental results show that the neural network control system can handle both self-selected walking and variable speed walking with high adaptability.

Detecting Toe-Off Events Utilizing a Vision-Based Method

Detecting gait events from video data accurately would be a challenging problem. However, most detection methods for gait events are currently based on wearable sensors, which need high cooperation from users and power consumption restriction. This study presents a novel algorithm for achieving accurate detection of toe-off events using a single 2D vision camera without the cooperation of participants. First, a set of novel feature, namely consecutive silhouettes difference maps (CSD-maps), is proposed to represent gait pattern. A CSD-map can encode several consecutive pedestrian silhouettes extracted from video frames into a map. And different number of consecutive pedestrian silhouettes will result in different types of CSD-maps, which can provide significant features for toe-off events detection. Convolutional neural network is then employed to reduce feature dimensions and classify toe-off events. Experiments on a public database demonstrate that the proposed method achieves good detection accuracy.

ED-FNN: A New Deep Learning Algorithm to Detect Percentage of the Gait Cycle for Powered Prostheses

Throughout the last decade, a whole new generation of powered transtibial prostheses and exoskeletons has been developed. However, these technologies are limited by a gait phase detection which controls the wearable device as a function of the activities of the wearer. Consequently, gait phase detection is considered to be of great importance, as achieving high detection accuracy will produce a more precise, stable, and safe rehabilitation device. In this paper, we propose a novel gait percent detection algorithm that can predict a full gait cycle discretised within a 1% interval. We called this algorithm an exponentially delayed fully connected neural network (ED-FNN). A dataset was obtained from seven healthy subjects that performed daily walking activities on the flat ground and a 15-degree slope. The signals were taken from only one inertial measurement unit (IMU) attached to the lower shank. The dataset was divided into training and validation datasets for every subject, and the mean square error (MSE) error between the model prediction and the real percentage of the gait was computed. An average MSE of 0.00522 was obtained for every subject in both training and validation sets, and an average MSE of 0.006 for the training set and 0.0116 for the validation set was obtained when combining all subjects' signals together. Although our experiments were conducted in an offline setting, due to the forecasting capabilities of the ED-FNN, our system provides an opportunity to eliminate detection delays for real-time applications.