Feature dimensionality reduction for myoelectric pattern recognition: a comparison study of feature selection and feature projection methods

Evaluation of feature projection techniques in object grasp classification using electromyogram signals from different limb positions

A myoelectric prosthesis is manipulated using electromyogram (EMG) signals from the existing muscles for performing the activities of daily living. A feature vector that is formed by concatenating data from many EMG channels may result in a high dimensional space, which may cause prolonged computation time, redundancy, and irrelevant information. We evaluated feature projection techniques, namely principal component analysis (PCA), linear discriminant analysis (LDA), t-Distributed Stochastic Neighbor Embedding (t-SNE), and spectral regression extreme learning machine (SRELM), applied to object grasp classification. These represent feature projections that are combinations of either linear or nonlinear, and supervised or unsupervised types. All pairs of the four types of feature projection with seven types of classifiers were evaluated, with data from six EMG channels and an IMU sensors for nine upper limb positions in the transverse plane. The results showed that SRELM outperformed LDA with supervised feature projections, and t-SNE was superior to PCA with unsupervised feature projections. The classification errors from SRELM and t-SNE paired with the seven classifiers were from 1.50% to 2.65% and from 1.27% to 17.15%, respectively. A one-way ANOVA test revealed no statistically significant difference by classifier type when using the SRELM projection, which is a nonlinear supervised feature projection (p = 0.334). On the other hand, we have to carefully select an appropriate classifier for use with t-SNE, which is a nonlinear unsupervised feature projection. We achieved the lowest classification error 1.27% using t-SNE paired with a k-nearest neighbors classifier. For SRELM, the lowest 1.50% classification error was obtained when paired with a neural network classifier.

Verification-Based Design of a Robust EMG Wake Word

Surface electromyography (sEMG) signals are now commonly used in continuous myoelectric control of prostheses. More recently, researchers have considered EMG-based gesture recognition systems for human computer interaction research. These systems instead focus on recognizing discrete gestures (like a finger snap). The majority of works, however, have focused on improving multi-class performance, with little consideration for false activations from "other" classes. Consequently, they lack the robustness needed for real-world applications which generally require a single motion class such as a mouse click or a wake word. Furthermore, many works have borrowed the windowed classification schemes from continuous control, and thus fail to leverage the temporal structure of the gesture. In this paper, we propose a verification-based approach to creating a robust EMG wake word using one-class classifiers (Support Vector Data Description, One Class-Support Vector Machine, Dynamic Time Warping (DTW) & Hidden Markov Models). The area under the ROC curve (AUC) is used as a feature optimization objective as it provides a better representation of the verification performance. Equal error rate (EER) and AUC are then used as evaluation metrics. The results are computed using both window-based and temporal classifiers on a dataset consisting of five different gestures, with a best EER of 0.04 and AUC of 0.98, recorded using a DTW scheme. These results demonstrate a design framework that may benefit the development of more robust solutions for EMG-based wake words or input commands for a variety of interactive applications.

Improvement of EMG Pattern Recognition Model Performance in Repeated Uses by Combining Feature Selection and Incremental Transfer Learning

Electromyography (EMG) pattern recognition is one of the widely used methods to control the rehabilitation robots and prostheses. However, the changes in the distribution of EMG data due to electrodes shifting results in classification decline, which hinders its clinical application in repeated uses. Adaptive learning can solve this problem but takes additional time. To address this, an efficient scheme is developed by comparing the performance of 12 combinations of three feature selection methods [no feature selection (NFS), sequential forward search (SFS), and particle swarm optimization (PSO)] and four classification methods [non-adaptive support vector machine (N-SVM), incremental SVM (I-SVM), SVM based on TrAdaBoost (T-SVM), and I-SVM based on TrAdaBoost (TI-SVM)] in the classification of EMG data of 12 subjects for 5 consecutive days. Our results showed that TI-SVM achieved the highest classification accuracy among the classification methods (p < 0.05). The SFS method achieved the same classification accuracy as that of the scheme trained with the feature vectors selected by the NFS method (p = 0.999) while achieving a lower training time than that of TI-SVM combined with the NFS method (p = 0.043). Although the PSO method outperformed the NFS and SFS methods by achieving reduced training and response times (p < 0.05), the PSO method achieved a considerably lower classification accuracy than that of the scheme trained with the feature vectors selected by the NFS (p = 0.001) or SFS (p = 0.001) method. Furthermore, TI-SVM combined with the SFS method outperformed the CNN method with fine-tuning in classification accuracy on a small data set (p = 0.001). The results indicate that TI-SVM combined with the SFS method is suitable for improving the performance of EMG pattern recognition in repeated uses.

Multiday Evaluation of Techniques for EMG-Based Classification of Hand Motions

Currently, most of the adopted myoelectric schemes for upper limb prostheses do not provide users with intuitive control. Higher accuracies have been reported using different classification algorithms but investigation on the reliability over time for these methods is very limited. In this study, we compared for the first time the longitudinal performance of selected state-of-the-art techniques for electromyography (EMG) based classification of hand motions. Experiments were conducted on ten able-bodied and six transradial amputees for seven continuous days. Linear discriminant analysis (LDA), artificial neural network (ANN), support vector machine (SVM), K-nearest neighbor (KNN), and decision trees (TREE) were compared. Comparative analysis showed that the ANN attained highest classification accuracy followed by LDA. Three-way repeated ANOVA test showed a significant difference (P < 0.001) between EMG types (surface, intramuscular, and combined), days (1-7), classifiers, and their interactions. Performance on the last day was significantly better (P < 0.05) than the first day for all classifiers and EMG types. Within-day, classification error (WCE) across all subject and days in ANN was: surface (9.12 ± 7.38%), intramuscular (11.86 ± 7.84%), and combined (6.11 ± 7.46%). The between-day analysis in a leave-one-day-out fashion showed that the ANN was the optimal classifier (surface (21.88 ± 4.14%), intramuscular (29.33 ± 2.58%), and combined (14.37 ± 3.10%). Results indicate that within day performances of classifiers may be similar but over time, it may lead to a substantially different outcome. Furthermore, training ANN on multiple days might allow capturing time-dependent variability in the EMG signals and thus minimizing the necessity for daily system recalibration.

A proportional control scheme for high density force myography

OBJECTIVE

Force myography (FMG) has been shown to be a potentially higher accuracy alternative to electromyography for pattern recognition based prosthetic control. Classification accuracy, however, is just one factor that affects the usability of a control system. Others, like the ability to start and stop, to coordinate dynamic movements, and to control the velocity of the device through some proportional control scheme can be of equal importance. To impart effective fine control using FMG-based pattern recognition, it is important that a method of controlling the velocity of each motion be developed.

METHODS

In this work force myography data were collected from 14 able bodied participants and one amputee participant as they performed a set of wrist and hand motions. The offline proportional control performance of a standard mean signal amplitude approach and a proposed regression-based alternative was compared. The impact of providing feedback during training, as well as the use of constrained or unconstrained hand and wrist contractions, were also evaluated.

RESULTS

It is shown that the commonly used mean of rectified channel amplitudes approach commonly employed with electromyography does not translate to force myography. The proposed class-based regression proportional control approach is shown significantly outperform this standard approach (ρ  <  0.001), yielding a R² correlation coefficients of 0.837 and 0.830 for constrained and unconstrained forearm contractions, respectively for able bodied participants. No significant difference (ρ  =  0.693) was found in R² performance when feedback was provided during training or not. The amputee subject achieved a classification accuracy of 83.4%  ±  3.47% demonstrating the ability to distinguish contractions well with FMG. In proportional control the amputee participant achieved an R² of of 0.375 for regression based proportional control during unconstrained contractions. This is lower than the unconstrained case for able-bodied subjects for this particular amputee, possibly due to difficultly in visualizing contraction level modulation without feedback. This may be remedied in the use of a prosthetic limb that would provide real-time feedback in the form of device speed.

CONCLUSION

A novel class-specific regression-based approach is proposed for multi-class control is described and shown to provide an effective means of providing FMG-based proportional control.

Feature-Level Fusion of Surface Electromyography for Activity Monitoring

Surface electromyography (sEMG) signals are commonly used in activity monitoring and rehabilitation applications as they reflect effectively the motor intentions of users. However, real-time sEMG signals are non-stationary and vary to a large extent within the time frame of signals. Although previous studies have focused on the issues, their results have not been satisfactory. Therefore, we present a new method of conducting feature-level fusion to obtain a new feature space for sEMG signals. Eight activities of daily life (ADLs), including falls, were performed to obtain raw data from EMG signals from the lower limb. A feature set combining the time domain, time–frequency domain, and entropy domain was applied to the raw data to establish an initial feature space. A new projection method, the weighting genetic algorithm for GCCA (WGA-GCCA), was introduced to obtain the final feature space. Different tests were carried out to evaluate the performance of the new feature space. The new feature space created with the WGA-GCCA effectively reduced the dimensions and selected the best feature vectors dynamically while improving monotonicity. The Davies–Bouldin index (DBI) based on fuzzy c-means algorithms of the space obtained the lowest value compared with several fusion methods. It also achieved the highest accuracy when applied to support vector machine classifier.