A Smart Terrain Identification Technique Based on Electromyography, Ground Reaction Force, and Machine Learning for Lower Limb Rehabilitation

Abstract

Automatic terrain classification in lower limb rehabilitation systems has gained worldwide attention. In this field, a simple system architecture and high classification accuracy are two desired attributes. In this article, a smart neuromuscular–mechanical fusion and machine learning-based terrain classification technique utilizing only two electromyography (EMG) sensors and two ground reaction force (GRF) sensors is reported for classifying three different terrains (downhill, level, and uphill). The EMG and GRF signals from ten healthy subjects were collected, preprocessed and segmented to obtain the EMG and GRF profiles in each stride, based on which twenty-one statistical features, including 9 GRF features and 12 EMG features, were extracted. A support vector machine (SVM) machine learning model is established and trained by the extracted EMG features, GRF features and the fusion of them, respectively. Several methods or statistical metrics were used to evaluate the goodness of the proposed technique, including a paired-t-test and Kruskal–Wallis test for correlation analysis of the selected features and ten-fold cross-validation accuracy, confusion matrix, sensitivity and specificity for the performance of the SVM model. The results show that the extracted features are highly correlated with the terrain changes and the fusion of the EMG and GRF features produces the highest accuracy of 96.8%. The presented technique allows simple system construction to achieve the precise detection of outcomes, potentially advancing the development of terrain classification techniques for rehabilitation.

EMG-Centered Multisensory Based Technologies for Pattern Recognition in Rehabilitation: State of the Art and Challenges

“…The first one is locomotion mode, which aims at the process that the brain actively coordinates the limbs to make different actions when the human body is facing different environments in the real world. Specifically, up and downslope, walking on flat ground, up and downstairs are the most widely studied in the relevant research [ 35 , 42 , 43 ]. This is not only because these five sports modes are the most common and widespread scenes in life, but also because these modes have the process of overcoming the influence of gravity and changing the center of gravity of the body, so auxiliary equipment is needed to adjust the lower limbs [ 44 , 45 ].…”

“…Typically, there are two ways for operators to contact the electrode with the skin. One is by using a silver-chloride gel to achieve wet contact while the other applies dry electrodes with microneedles to record EMG signals [ 43 ].…”

“…For example, as for prosthetic control, which is mainly designed for the amputees, experimenters commonly choose the proximal muscle groups like thigh muscle group [ 29 , 35 , 42 ] and the upper arms [ 59 ] because the distal ones are considered to be amputated. While, as for the studies oriented for non-amputees subjects, the distal muscle groups such as the shank muscle tissues and forearm and wrist muscles are often selected as targeted muscles due to their lower inter-subject variability than the proximal [ 43 , 60 , 61 ]. Gao Shuo [ 43 ] chose a pair of distal antagonistic muscles, tibialis anterior (TA) and soleus (SL) to conduct the terrain identification experiment and got satisfying results.…”

“…While, as for the studies oriented for non-amputees subjects, the distal muscle groups such as the shank muscle tissues and forearm and wrist muscles are often selected as targeted muscles due to their lower inter-subject variability than the proximal [ 43 , 60 , 61 ]. Gao Shuo [ 43 ] chose a pair of distal antagonistic muscles, tibialis anterior (TA) and soleus (SL) to conduct the terrain identification experiment and got satisfying results. Besides, muscle characteristic is also an important aspect to be taken into account.…”

“…Instead, they use the prior knowledge of biomechanics to select the specific response point at the experimental curve. For example, in [ 43 ], twenty-one biomechanical features such as the peak EMG interval between two selected muscles and the duration of the EMG activation time are carefully selected. Such an extraction technique obtained an accuracy of 96.8% and is proven to outperform the traditional feature extraction technique.…”