Intra-subject approach for gait-event prediction by neural network interpretation of EMG signals

Abstract

Background Gait analysis is acknowledged as the main approach for quantitatively assessing the alteration of motor function in different contexts, such as in basic research and clinics. Technological development is making available smart and wearable sensors (inertial Abstract Background: Machine learning models were satisfactorily implemented for estimating gait events from surface electromyographic (sEMG) signals during walking. Most of them are based on inter-subject approaches for data preparation. Aim of the study is to propose an intra-subject approach for binary classifying gait phases and predicting gait events based on neural network interpretation of sEMG signals and to test the hypothesis that the intra-subject approach is able to achieve better performances compared to an inter-subject one. To this aim, sEMG signals were acquired from 10 leg muscles in about 10.000 strides from 23 healthy adults, during ground walking, and a multi-layer perceptron (MLP) architecture was implemented. Results: Classification/prediction accuracy was tested vs. the ground truth, represented by the foot-floor-contact signal provided by three foot-switches, through samples not used during training phase. Average classification accuracy of 96.1 ± 1.9% and mean absolute value (MAE) of 14.4 ± 4.7 ms and 23.7 ± 11.3 ms in predicting heel-strike (HS) and toe-off (TO) timing were provided. Performances of the proposed approach were tested by a direct comparison with performances provided by the inter-subject approach in the same population. Comparison results showed 1.4% improvement of mean classification accuracy and a significant (p < 0.05) decrease of MAE in predicting HS and TO timing (23% and 33% reduction, respectively). Conclusions: The study developed an accurate methodology for classification and prediction of gait events, based on neural network interpretation of intra-subject sEMG data, able to outperform more typical inter-subject approaches. The clinically useful contribution consists in predicting gait events from only EMG signals from a single subject, contributing to remove the need of further sensors for the direct measurement of temporal data.

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

“…A mean F1-score of 99.0% and a MAE of 21.6 ms were detected for the prediction of HS events and a mean F1-score of 98.4% and a MAE of 38.1 ms were identified for the prediction of TO events [18]. This approach has been numerically outperformed by a subsequent study of the same group of researchers based on intra-subjects experiments in the same population [19]. Average classification accuracy of 96.1±1.9% and mean MAE of 14.4±4.7 ms (associated to an F1-score of 99.3%) and 23.7±11.3 ms (associated to an F1-score of 98.5%) in predicting HS and TO timing were provided.…”

“…The neural network used in this work is a simple single-layer network with 10 units and the Levenberg Marquardt algorithm was used to train the network. These methods are based on hand-crafted features; otherwise, previous studies by authors of the present paper adopted a featureless approach, employing Multi-Layer Perceptrons to process the envelope of sEMG signals [18,19]. One further difference is the optimization algorithm: Levenberg Marquardt in [17], Adam in [18,19].…”

“…These methods are based on hand-crafted features; otherwise, previous studies by authors of the present paper adopted a featureless approach, employing Multi-Layer Perceptrons to process the envelope of sEMG signals [18,19]. One further difference is the optimization algorithm: Levenberg Marquardt in [17], Adam in [18,19]. Inter-subject evaluation on a population of twentythree healthy adults reported overall mean classification accuracy (stance vs. swing) of 93.4%.…”

“…Nevertheless, literature reports just a few studies [14][15][16][17][18][19], trying to face the problem of gait-event detection by means of machine-learning-based methods (see Section II, Related Works, for details). To our knowledge, a recent intra-subject approach, introduced by the present group of researchers, is C still reporting the best performance among the EMG-based ones proposed in literature, showing a mean absolute error (MAE) of 14.4±4.7 ms and 23.7±11.3 ms in predicting HS and TO timing, respectively [19]. These encouraging performances were achieved considering data acquired during able-bodied-subject walking, where the clear majority of the strides (around 90%) follows the typical foot-floor-contact sequence, known as HFPS [20]: heel contact (0-6% of gait cycle, H), flat foot contact (6-38%, F), push-off (38-60%, P), and swing (60-100%, S).…”

“…Thus, the goal of the present study is to investigate a machine-learning-based approach specifically developed to binary classify gait events and to predict HS and TO timing from sEMG signals in mildhemiplegic-children walking. Depending on the application field, two main experimental settings are proposed: the intrasubject approach and the inter-subject approach [19]. In the first setting, sEMG data collected during walking of a single subject are used to train a neural network (NN) to recognize different gaits from the same individual.…”