Ground Reaction Force Signals Research Articles

An efficient tool for Parkinson's disease detection and severity grading based on time-frequency and fuzzy features of cumulative gait signals through improved LSTM networks

Parkinson's disease (PD) is a widespread neurodegenerative condition that affects many individuals annually. Early identification and monitoring of disease progression are crucial to effectively managing symptoms and preventing motor complications. This research proposes an automated PD diagnosis and severity-grading model based on time-frequency and fuzzy features using improved uni-directional and bi-directional long short-term memory networks with sensitive hyperparameters optimization. We utilize vertical ground reaction force signals collected from Physionet's publicly available dataset recorded during regular and dual-task clinical trials of walking measurements. Only the cumulative signal of both feet was then utilized and segmented into 30-s windows without further pre-processing. Subsequently, we extracted only four key time-frequency and fuzzy features from each segment, effectively capturing the signal's inherent uncertainty. Bayesian optimization is employed in both detection and grading approaches to fine-tune the two critical hyperparameters: the initial learning rate and the number of hidden units in the network. The detection phase yields an exceptional accuracy of 99.19%, surpassing state-of-the-art studies with the same dataset. In the grading phase, classification based on the unified PD rating scale values achieves an accuracy of 92.28%. The proposed study delves into the potential of cumulative gait signals as a powerful diagnostic tool for PD, aiming to extract precise and intricate information by implementing straightforward and minimal processing endeavors. This method demonstrates significant efficiency in terms of complexity, cost, and energy consumption by utilizing a single-dimensional signal, eliminating the need for pre-processing steps, and limiting the features used for training.

Antagonist Coactivation of Muscles of Ankle and Thigh in Post-Stroke vs. Healthy Subjects during Sit-to-Stand Task

This study aims to analyse the coactivation of antagonist muscles of the thigh and ankle during the sit-to-stand task in post-stroke subjects, specifically during forward and antigravity sub-phases. A group of 18 healthy subjects and another with 18 subjects with a history of stroke participated voluntarily in this study. Bilateral surface electromyography (EMGs) of the soleus, gastrocnemius medialis, tibialis anterior, rectus femoris and biceps femoris muscles were collected synchronously with ground reaction forces (GRF) during the sit-to-stand task. The magnitude of electromyographic (EMG) activity was analysed during forward translation and antigravity sub-phases which were determined through GRF signals. The coactivation was calculated to quantify the degree of antagonist coactivation according to the role of the muscles during the task. Statistically significant values were found between antagonist coactivation on both sub-phases of the sit-to-stand task when comparing healthy and post-stroke subjects (healthy with ipsilesional (IPSI); healthy with contralesional (CONTRA); and healthy with IPSI and with CONTRA limbs) in all muscle pairs analysed (p < 0.01), except on thigh muscles (p > 0.05), in the antigravity sub-phase. When comparing IPSI with CONTRA sides in post-stroke subjects, no statistically significant differences were found. Increased values of antagonist coactivation were observed in post-stroke subjects compared to healthy subjects (both IPSI and CONTRA limb) in the two sub-phases analysed. The forward sub-phase CONTRA limb showed higher antagonist coactivation compared to IPSI, while in the antigravity sub-phase, IPSI antagonist coactivation was higher than in the CONTRA. In conclusion, post-stroke subjects presented an antagonist coactivation more dysfunctional at the ankle joint muscles compared to the thigh segment. So, it seems that the distal segment could express more accurately the central nervous system dysfunction in post-stroke subjects, despite the need for further studies to achieve a better spatiotemporal understanding of the variability on coactivation levels.

The effect of severe obesity on three-dimensional ground reaction force signals during walking

BackgroundThe gait pattern of adults with class I obesity [30 ≤ body mass index < 35kg/m2] was characterized by altered three-dimensional ground reaction force signals compared to lean adults (18.5 ≤ body mass index < 25 kg/m2). However, results might not be generalizable to adults with severe obesity (class II and III; body mass index ≥ 35 kg/m2). Hence, the purpose of the present study was to investigate the differences in relative ground reaction force signals, i.e., normalized by body weight, between adults with severe obesity and lean adults using functional principal component analysis. MethodsThirteen lean and eighteen sedentary adults with severe obesity performed a 5-min walking trial (1.11 m/s) on an instrumented treadmill. The first five functional principal components of the relative force signals (mediolateral, anterior-posterior, and vertical directions) were obtained using functional principal component analysis. Functional principal component scores were compared between groups using an analysis of covariance with age as covariable. FindingsFunctional principal component analysis reported a statistically significant group effect for first functional principal component score for mediolateral (P = 0.004), and second and fifth functional principal component scores for anterior-posterior (P ≤ 0.02) force signals. Adults with severe obesity displayed a greater mediolateral force during most of the stance but similar magnitudes of the anterior-posterior and vertical forces compared to lean adults. InterpretationTherefore, increasing the obesity level accentuates differences in mediolateral force but promotes no specific changes in anterior-posterior force likely due to chronic loading adaptation.

A Performance Evaluation of Overground Gait Training With a Mobile Body Weight Support System Using Wearable Sensors

Overground gait training under body weight support (BWS) for patients who suffer from neurological injuries has been proven practical in recovering from walking ability. Conventionally, skilled therapists or additional robots are required to assist the patient’s body weight and pelvis movement, making the rehabilitation process physically and economically burdensome. We investigate if a BWS walker using only two actuators can support the user’s body weight and simultaneously protect/assist the transverse pelvis rotation, improving natural gait with minimal motion compensation. In this paper, a BWS strategy called transverse pelvis rotation support (TPRS) is proposed to enable the BWS system to generate cable tension in the forward direction, as a purpose to support transverse pelvis rotation in addition to our previously proposed static or variable BWS. Wearable sensory devices, including instrumented shoes and harness, were developed to provide real-time ground reaction force and pelvis rotation signals simultaneously. Ten non-disabled participants were unloaded with 0% ~ 15% BWS under four different controls. Vertical ground reaction force, transverse pelvis kinematics, and user experience were compared using proposed controls. One-Way repeated measures ANOVA analysis assessed if control strategies generally affect the performance. All proposed controls enable the walker to support part of the user’s body weight. SBWS-TPRS and VBWS-TPRS control enable users to achieve a significantly improved pelvic motion and prolonged single support phase than pure static BWS or variable BWS, although users perceive a higher workload under them. The proposed BWS controls show the potential to become a complementary method in gait rehabilitation.

Isolating the speed factor is crucial in gait analysis for Parkinson's disease.

Parkinson's disease (PD) is characterized by an alteration of the walking gait, frequently including a slower self-selected walking speed (SSWS). Although the reduction of walking speed is inherent to people with PD, such speed reduction also represents a potential confounding factor that might partly explain the observed gait differences between PD and control participants. In this study, each participant walked along a 25 m level corridor during which vertical ground reaction force signals were recorded using shoes equipped with eight pressure sensors. Vertical ground reaction force signals (using statistical parametric mapping) and temporal and kinetic variables as well as their related variability and asymmetry (using Student's t-test) were compared between PD (n = 54) and walking-speed-matched control subjects (n = 39). Statistical parametric mapping did not yield significant differences between PD and control groups for the vertical ground reaction force signal along the walking stance phase. Stride time and single support time (equivalent to swing time) were shorter and peak vertical ground reaction force was larger in PD patients compared to controls (p ≤ 0.05). However, the single support time was no longer different between people with PD and healthy subjects when expressed relatively to stride time (p = 0.07). While single support, double support, and stance times were significantly more variable and asymmetric for PD than for the control group (p ≤ 0.05), stride time was similar (p ≥ 0.07). These results indicate that at matched SSWS, PD patients adopt a higher cadence than control participants. Moreover, the temporal subdivision of the walking gait of people with PD is similar to healthy individuals but the coordination during the double support phase is different. Hence, this study indicates that isolating the speed factor is crucial in gait analysis for PD.

A Robust Frequency-Domain-Based Graph Adaptive Network for Parkinson's Disease Detection From Gait Data

Parkinson's disease (PD) is a neurodegenerative disease with a high incidence rate. Effective early diagnosis of PD is critical to prevent further deterioration of a patient's condition, where gait abnormalities are important factors for doctors to diagnose PD. Deep learning (DL)-based methods for PD detection using gait information recorded by non-invasive sensors have emerged to assist doctors in accurate and efficient disease diagnosis. However, most existing DL-based PD detection models neglect information in the frequency domain and do not adaptively model the correlation of signals among sensors. Moreover, different people have different gait patterns. Therefore, the generalization capabilities of PD detection models on diversities of individuals' gaits are essential. This work proposes a novel robust frequency-domain-based graph adaptive network (RFdGAD) for PD detection from gait information (i.e., vertical ground reaction force signals recorded by foot sensors). Specifically, the RFdGAD first learns the frequency-domain features of signals from each foot sensor by a frequency representation learning block. Then, the RFdGAD utilizes a graph adaptive network block taking frequency-domain features as input to adaptively learn and exploit the interconnection between different sensor signals for accurate PD detection. Moreover, the RFdGAD is trained by minimizing the proposed Jensen-Shannon divergence-based localized generalization error to improve the generalization performance of RFdGAD on unseen subjects. Experimental results show that the RFdGAD outperforms existing DL-based models for PD detection on three widely used datasets in terms of three metrics, including accuracy, F1-score, and geometric mean.

Modeling biological individuality using machine learning: A study on human gait

Human gait is a complex and unique biological process that can offer valuable insights into an individual’s health and well-being. In this work, we leverage a machine learning-based approach to model individual gait signatures and identify factors contributing to inter-individual variability in gait patterns. We provide a comprehensive analysis of gait individuality by (1) demonstrating the uniqueness of gait signatures in a large-scale dataset and (2) highlighting the gait characteristics that are most distinctive to each individual. We utilized the data from three publicly available datasets comprising 5368 bilateral ground reaction force recordings during level overground walking from 671 distinct healthy individuals. Our results show that individuals can be identified with a prediction accuracy of 99.3% by using the bilateral signals of all three ground reaction force components, with only 10 out of 1342 recordings in our test data being misclassified. This indicates that the combination of bilateral ground reaction force signals with all three components provides a more comprehensive and accurate representation of an individual’s gait signature. The highest accuracy was achieved by (linear) Support Vector Machines (99.3%), followed by Random Forests (98.7%), Convolutional Neural Networks (95.8%), and Decision Trees (82.8%). The proposed approach provides a powerful tool to better understand biological individuality and has potential applications in personalized healthcare, clinical diagnosis, and therapeutic interventions.

Machine learning-based classification of healthy and impaired gaits using 3D-GRF signals

Gait analysis is helpful for rehabilitation, clinical diagnoses, and sporting activities. Among the gathered signals, ground reaction forces (GRF) may be used for assisting doctors in recognizing and categorizing gait patterns using Machine-Learning methods. In this study, GaitRec and Gutenberg databases were used, where GaitRec contains 2645 gait disorder (GD) patients and 211 Healthy Controls (HCs), and the Gutenberg database has 350 HCs. The combined database has HCs and four GD classes: hip, knee, ankle, and calcaneus. GD is an abnormality in the hip, knee, or ankle joints, whereas Calcaneus gait is calcaneus fractures or ankle fusions. We pre-processed the GRF signals, applied different feature extraction techniques, removed the highly correlated features, and ranked the features using three feature selection algorithms. K-nearest neighbour model (KNN) showed the top performance in terms of accuracy in all experiments. Four different experimental schemes were pursued: (i) 6 binary classifications; (ii) 1 three-class classification; (iii) 2 four-class classifications; (iv) one five-class classification. We also compared the performance of vertical GRF with three-dimensional GRF. We found that using three-dimensional GRF increased the overall performance. Furthermore, it is found that time-domain and Wavelet features are among the most useful in identifying gait patterns. The findings show promising performance in automated gait disorder classification.

Using statistical parametric mapping to assess the association of duty factor and step frequency on running kinetic.

Duty factor (DF) and step frequency (SF) were previously defined as the key running pattern determinants. Hence, this study aimed to investigate the association of DF and SF on 1) the vertical and fore-aft ground reaction force signals using statistical parametric mapping; 2) the force related variables (peaks, loading rates, impulses); and 3) the spring-mass characteristics of the lower limb, assessed by computing the force-length relationship and leg stiffness, for treadmill runs at several endurance running speeds. One hundred and fifteen runners ran at 9, 11, and 13km/h. Force data (1000Hz) and whole-body three-dimensional kinematics (200Hz) were acquired by an instrumented treadmill and optoelectronic system, respectively. Both lower DF and SF led to larger vertical and fore-aft ground reaction force fluctuations, but to a lower extent for SF than for DF. Besides, the linearity of the force-length relationship during the leg compression decreased with increasing DF or with decreasing SF but did not change during the leg decompression. These findings showed that the lower the DF and the higher the SF, the more the runner relies on the optimization of the spring-mass model, whereas the higher the DF and the lower the SF, the more the runner promotes forward propulsion.

A stochastic approach for generating individual jumping loads considering different jumping force patterns

This study establishes a stochastic model for generating individual jumping force signals based on a database of individual jumping force records. A total of 1296 vertical jumping loads at a jumping frequency of 1.5–3.5 Hz were experimentally measured with high quality. To generate a jumping ground reaction forces (GRFs) curve, the proportions of different force patterns were first obtained based on the database and subsequently used to assign the numbers of individual jumping pulses for different force patterns. Then, a half-sine square function, three Gauss superposition functions, and two Gauss superposition functions were respectively used to generate ‘single-peaked’, ‘double-peaked’, and ‘merging’ shape jumping forces. To reflect the near-periodic nature of the jumping forces, the function parameters and coefficients were randomly generated from their normal distributions, and these parameters and coefficients were assigned to each jumping force. Finally, the generated individual jumping pulses were randomly arranged and connected to form a jumping GRF signal. A comparison of the jumping load–time history curves and Fourier amplitude spectra between the measured and generated results verified the reliability of the model. Compared with the conventional half-sine model, the stochastic approach for generating jumping loads in this study can better reflect the temporal characteristics of the real jumping GRF signals, and the model is convenient to be applied in the vibration serviceability assessment of civil engineering structures, such as stadium stands. • A stochastic model to generate individual jumping loads was established based on a database of 1296 vertical jumping force records. • Different jumping force shapes can be described by the model. • Comparison of jumping load time-history curves and the Fourier amplitude spectra between measured and generated results verified the model's reliability. • The coefficient of correction for the jumping impact factor parabolically varies with the increase of the jumping frequency from 1.5 to 3.5 Hz.

Parkinson’s disease diagnosis and stage prediction based on gait signal analysis using EMD and CNN–LSTM network

Parkinson’s disease (PD) diagnosis is a complex and challenging task which needs the assessment of various motor and non-motor symptoms. As gait impairment is one of the early and important symptoms of PD, in a clinical setting, physicians generally evaluate the gait abnormality based on visual observations along with other numerous manifestations to assess the severity of PD. As such kind of assessment majorly depends on the experience and expertise of the physicians, there is scope for bias in assessment, leading to misdiagnosis. In this context, to assist the physicians to diagnose PD effectively, this study aims to design and investigate a gait analysis based classifier model using a hybrid convolutional neural network-long short term memory (CNN–LSTM) network to predict the severity rating of PD. For evaluation, we utilize the openly available gait dataset from Physionet that consists of vertical ground reaction force (VGRF) signals from three different walking tests. Firstly, the prominent VGRF signals obtained using the variability analysis are decomposed using the empirical mode decomposition (EMD) technique to extract the significant intrinsic mode functions (IMFs) that contain the vital gait features. Secondly, through the power spectral analysis the dominant IMFs of the selected VGRF signals are extracted to train the CNN–LSTM classifier model. To address the data overfitting problem in the classifier model, the proposed approach employs L2 regularization along with dropout techniques. Moreover, to solve the stochastic cost function, CNN–LSTM network utilizes the Adam optimizer for its minimal memory requirement and tuning. Finally, the experiments conducted using the gait patterns from 93 PD subjects and 73 healthy controls substantiate that the proposed CNN–LSTM​ classifier model can achieve a maximum multi-class classification accuracy of 98.32% and offer superior performance compared to several other similar methods which have used gait pattern to diagnose PD.

Machine Learning-Based Diabetic Neuropathy and Previous Foot Ulceration Patients Detection Using Electromyography and Ground Reaction Forces during Gait.

Diabetic neuropathy (DN) is one of the prevalent forms of neuropathy that involves alterations in biomechanical changes in the human gait. Diabetic foot ulceration (DFU) is one of the pervasive types of complications that arise due to DN. In the literature, for the last 50 years, researchers have been trying to observe the biomechanical changes due to DN and DFU by studying muscle electromyography (EMG) and ground reaction forces (GRF). However, the literature is contradictory. In such a scenario, we propose using Machine learning techniques to identify DN and DFU patients by using EMG and GRF data. We collected a dataset from the literature which involves three patient groups: Control (n = 6), DN (n = 6), and previous history of DFU (n = 9) and collected three lower limb muscles EMG (tibialis anterior (TA), vastus lateralis (VL), gastrocnemius lateralis (GL)), and three GRF components (GRFx, GRFy, and GRFz). Raw EMG and GRF signals were preprocessed, and different feature extraction techniques were applied to extract the best features from the signals. The extracted feature list was ranked using four different feature ranking techniques, and highly correlated features were removed. In this study, we considered different combinations of muscles and GRF components to find the best performing feature list for the identification of DN and DFU. We trained eight different conventional ML models: Discriminant analysis classifier (DAC), Ensemble classification model (ECM), Kernel classification model (KCM), k-nearest neighbor model (KNN), Linear classification model (LCM), Naive Bayes classifier (NBC), Support vector machine classifier (SVM), and Binary decision classification tree (BDC), to find the best-performing algorithm and optimized that model. We trained the optimized the ML algorithm for different combinations of muscles and GRF component features, and the performance matrix was evaluated. Our study found the KNN algorithm performed well in identifying DN and DFU, and we optimized it before training. We found the best accuracy of 96.18% for EMG analysis using the top 22 features from the chi-square feature ranking technique for features from GL and VL muscles combined. In the GRF analysis, the model showed 98.68% accuracy using the top 7 features from the Feature selection using neighborhood component analysis for the feature combinations from the GRFx-GRFz signal. In conclusion, our study has shown a potential solution for ML application in DN and DFU patient identification using EMG and GRF parameters. With careful signal preprocessing with strategic feature extraction from the biomechanical parameters, optimization of the ML model can provide a potential solution in the diagnosis and stratification of DN and DFU patients from the EMG and GRF signals.

A Feasibility Study on Tribological Origins of Knee Acoustic Emissions.

Considering the knee as a fluid-lubricated system, articulating surfaces undergo different lubrication modes and generate joint acoustic emissions (JAEs). The goal of this study is to compare knee biomechanical signals against synchronously recorded joint sounds and assess the hypothesis that JAEs are attributed to tribological origins. JAE, electromyography, ground reaction force signals, and motion capture markers were synchronously recorded from ten healthy subjects while performing two-leg and one-leg squat exercises. The biomechanical signals were processed to calculate a tribological parameter, lubrication coefficient, and JAEs were divided into short windows and processed to extract 64-time-frequency features. The lubrication coefficients and JAE features of two-leg squats were used to label the windows and train a classifier that discriminates the knee lubrication modes only based on JAE features. The classifier was used to predict the label of one-leg squat JAE windows and it achieved a high test-accuracy of 84%. The Pearson correlation coefficient between the estimated friction coefficient and predicted JAE scores was 0.83 ± 0.08. Furthermore, the lubrication coefficient threshold, separating two lubrication modes, decreased by half from two-leg to one-leg squats. This result was consistent with tribological changes in the knee load as it was inversely doubled in one-leg squats. This study supports the potential use of JAEs as a quantitative biomarker to extract tribological information. Since arthritis and similar disease impact the roughness of the joint cartilage, the use of JAEs could have broad implications for studying joint frictions and monitoring joint structural changes with wearable devices.

Musculoskeletal Injury Recovery Assessment using Gait Analysis with Ground Reaction Force Sensor

Patients with musculoskeletal injuries are required to exhibit specific tests for clinicians to monitor the recovery progress during the rehabilitation period. An automated system tracking the progress of an injured patient is essential for emerging applications in the healthcare domain. In this study, we propose a long short-term memory cell-based auto-encoder model (LSTM-AE) for the recovery assessment using data collected during walking. A recently published ground reaction force (GRF) dataset of musculoskeletal impairments, GaitRec, is used for this study. The model is trained to learn the gait pattern of healthy individuals. The model is then tested with GRF signals collected during different sessions of rehabilitation from injured patients. The reconstruction losses generated using the GRF signals are assessed to indicate the recovery status of the patient. The results analysis suggests that the reconstruction loss of the LSTM-AE model gradually reduces towards the ending phase of recovery. Similar trend is observed for both legs in all categories of impairments indicating gradual improvement in patients.

Gait Phase Recognition Using Fuzzy Logic Regulation with Multisensor Data Fusion

The health challenges brought by aging population and chronic noncommunicable diseases are increasingly severe. Scientific physical exercise is of great significance to prevent the occurrence of chronic diseases and subhealth intervention and promote health. However, improper or excessive exercise can cause injury. Research shows that the sports injury rate of people who often exercise is as high as 85%. Aiming at the problem of low accuracy of single sensor gait analysis, a real-time gait detection algorithm based on piezoelectric film and motion sensor is proposed. On this basis, a gait phase recognition method based on fuzzy logic is proposed, which enhances the ability of gait space-time measurement. Experimental results show that the proposed gait modeling method based on ground reaction force (GRF) signal can effectively recognize and quantify various gait patterns. At the same time, the introduction of heterogeneous sensor data fusion technology can effectively make up for the accuracy defects of single sensor measurement and improve the estimation accuracy of gait space-time measurement.

Gutenberg Gait Database, a ground reaction force database of level overground walking in healthy individuals

The Gutenberg Gait Database comprises data of 350 healthy individuals recorded in our laboratory over the past seven years. The database contains ground reaction force (GRF) and center of pressure (COP) data of two consecutive steps measured - by two force plates embedded in the ground - during level overground walking at self-selected walking speed. The database includes participants of varying ages, from 11 to 64 years. For each participant, up to eight gait analysis sessions were recorded, with each session comprising at least eight gait trials. The database provides unprocessed (raw) and processed (ready-to-use) data, including three-dimensional GRF and two-dimensional COP signals during the stance phase. These data records offer new possibilities for future studies on human gait, e.g., the application as a reference set for the analysis of pathological gait patterns, or for automatic classification using machine learning. In the future, the database will be expanded continuously to obtain an even larger and well-balanced database with respect to age, sex, and other gait-specific factors.

Implementation of a Deep Learning Algorithm Based on Vertical Ground Reaction Force Time-Frequency Features for the Detection and Severity Classification of Parkinson's Disease.

Conventional approaches to diagnosing Parkinson’s disease (PD) and rating its severity level are based on medical specialists’ clinical assessment of symptoms, which are subjective and can be inaccurate. These techniques are not very reliable, particularly in the early stages of the disease. A novel detection and severity classification algorithm using deep learning approaches was developed in this research to classify the PD severity level based on vertical ground reaction force (vGRF) signals. Different variations in force patterns generated by the irregularity in vGRF signals due to the gait abnormalities of PD patients can indicate their severity. The main purpose of this research is to aid physicians in detecting early stages of PD, planning efficient treatment, and monitoring disease progression. The detection algorithm comprises preprocessing, feature transformation, and classification processes. In preprocessing, the vGRF signal is divided into 10, 15, and 30 s successive time windows. In the feature transformation process, the time domain vGRF signal in windows with varying time lengths is modified into a time–frequency spectrogram using a continuous wavelet transform (CWT). Then, principal component analysis (PCA) is used for feature enhancement. Finally, different types of convolutional neural networks (CNNs) are employed as deep learning classifiers for classification. The algorithm performance was evaluated using k-fold cross-validation (kfoldCV). The best average accuracy of the proposed detection algorithm in classifying the PD severity stage classification was 96.52% using ResNet-50 with vGRF data from the PhysioNet database. The proposed detection algorithm can effectively differentiate gait patterns based on time–frequency spectrograms of vGRF signals associated with different PD severity levels.

Identification of Neurodegenerative Diseases Based on Vertical Ground Reaction Force Classification Using Time-Frequency Spectrogram and Deep Learning Neural Network Features.

A novel identification algorithm using a deep learning approach was developed in this study to classify neurodegenerative diseases (NDDs) based on the vertical ground reaction force (vGRF) signal. The irregularity of NDD vGRF signals caused by gait abnormalities can indicate different force pattern variations compared to a healthy control (HC). The main purpose of this research is to help physicians in the early detection of NDDs, efficient treatment planning, and monitoring of disease progression. The detection algorithm comprises a preprocessing process, a feature transformation process, and a classification process. In the preprocessing process, the five-minute vertical ground reaction force signal was divided into 10, 30, and 60 s successive time windows. In the feature transformation process, the time–domain vGRF signal was modified into a time–frequency spectrogram using a continuous wavelet transform (CWT). Then, feature enhancement with principal component analysis (PCA) was utilized. Finally, a convolutional neural network, as a deep learning classifier, was employed in the classification process of the proposed detection algorithm and evaluated using leave-one-out cross-validation (LOOCV) and k-fold cross-validation (k-fold CV, k = 5). The proposed detection algorithm can effectively differentiate gait patterns based on a time–frequency spectrogram of a vGRF signal between HC subjects and patients with neurodegenerative diseases.

Preliminary Study on Real-Time Prediction of Gait Acceleration Intention From Volition-Associated EEG Patterns

Volitional control of prosthetic devices can potentially provide the user with a more natural movement experience. Currently, there is no feasible volitional triggering method to adapt the prosthetic device to user’s intention to accelerate during walking. Therefore, real-time prediction of human acceleration intention from the pre-acceleration electroencephalogram (EEG), and subsequent adaptation of the prosthetic device’s control parameters for seamless transition remains a daunting research area. In that aspect, this study investigates the neural changes responsive to human intention to accelerate during walking. Furthermore, this study also explores whether the acceleration intention can be predicted from real-time EEG to subsequently enable parametric adaptation for an external prosthetic device. EEG, Inertial Measurement Unit (IMU), and ground reaction force (GRF) signals were collected from one healthy subject during walking with self-paced speed changes. A set of classifiers were explored to classify between constant speed and acceleration. The classifiers showed promising classification performance well above chance level in offline, pseudo-online and real-time scenarios. An accuracy of 85.9±2.9% was achieved in offline scenario, while pseudo-online classification resulted in a true positive rate (TPR) of 81.9±7.4% with 7.7±0.8 false positives/ min and a detection latency of −844±572 ms. In real-time scenario, 9 out of 12 acceleration events were predicted successfully with only 3 false predictions at an average latency of −741 ms. Moreover, offline data analysis suggested suppression of mu and beta rhythms related to gait acceleration between 2 seconds before and 1.5 seconds after the onset of acceleration. A slow increase in negative amplitude was also observed in near DC frequencies of EEG data acquired from the sensorimotor cortex. The findings of this study portray inspiring results in retrieving gait acceleration-associated neural changes for facilitating natural control of assistive devices.

Gait Analysis by Causal Decomposition

Recent studies have investigated bilateral gaits based on the causality analysis of kinetic (or kinematic) signals recorded using both feet. However, these approaches have not considered the influence of their simultaneous causation, which might lead to inaccurate causality inference. Furthermore, the causal interaction of these signals has not been investigated within their frequency domain. Therefore, in this study we attempted to employ a causal-decomposition approach to analyze bilateral gait. The vertical ground reaction force (VGRF) signals of Parkinson's disease (PD) patients and healthy control (HC) individuals were taken as an example to illustrate this method. To achieve this, we used ensemble empirical mode decomposition to decompose the left and right VGRF signals into intrinsic mode functions (IMFs) from the high to low frequency bands. The causal interaction strength (CIS) between each pair of IMFs was then assessed through the use of their instantaneous phase dependency. The results show that the CISes between pairwise IMFs decomposed in the high frequency band of VGRF signals can not only markedly distinguish PD patients from HC individuals, but also found a significant correlation with disease progression, while other pairwise IMFs were not able to produce this. In sum, we found for the first time that the frequency specific causality of bilateral gait may reflect the health status and disease progression of individuals. This finding may help to understand the underlying mechanisms of walking and walking-related diseases, and offer broad applications in the fields of medicine and engineering.