sEMG-angle estimation using feature engineering techniques for least square support vector machine

Abstract

In the practical implementation of control of electromyography (sEMG) driven devices, algorithms should recognize the human's motion from sEMG with fast speed and high accuracy. This study proposes two feature engineering (FE) techniques, namely, feature-vector resampling and time-lag techniques, to improve the accuracy and speed of least square support vector machine (LSSVM) for wrist palmar angle estimation from sEMG feature. The root mean square error and correlation coefficients of LSSVM with FE are 9.50 ± 2.32 degree and 0.971 ± 0.018 respectively. The average training time and average execution time of LSSVM with FE in processing 12600 sEMG points are 0.016 s and 0.053 s respectively. To evaluate the proposed algorithm, its estimation results are compared with those of three other methods, namely, LSSVM, radial basis function (RBF) neural network, and RBF with FE. Experimental results verify that introduction of time-lag into feature vector can greatly improve the estimation accuracy of both RBF and LSSVM; meanwhile the application of feature-vector resampling technique can significantly increase the training and execution speed of RBF neural network and LSSVM. Among different algorithms applied in this study, LSSVM with FE techniques performed best in terms of training and execution speed, as well as estimation accuracy.

Keywords: Least square support vector machine; angle estimation; electromyograph; feature engineering.

Figure 1.

The experimental platform.

Figure 2.

FE techniques. The red rectangle represents the sliding window.

Figure 3.

Basic structure of LSSVM with FE techniques. The blue lines represent the sEMG and angle data flow in the training stage. The red lines indicate the sEMG data flow in the validation stage.

Figure 4.

Color maps of the estimated results of C#1 when different parameters of feature vector resampling were applied. M is the number of resampling points and B is the interval of sampling. (a) Color map of the RMSE value of C#1. (b) Color map of the CC value of C#1.

Figure 5.

Estimated results of A#4 when the parameters of feature resampling is set according to Table 1. (a) Estimated results of A#4 when the parameters of feature resampling are set as group 1. (b) Zoomed portion of (a). (c) Estimated results of A#4 when the parameters of feature resampling are set as group 2. (d) Zoomed portion of (c).

Figure 6.

Structural diagram of sEMG-angle estimation process.

Figure 7.

Estimated results of E#3 when different parameters of time lag introduction are applied. The N and P are the window size and time lag. (a) Three dimensional graphs of RMSE. (b) Three dimensional graphs of CC. (c) Color map of RMSE. (d) Color map of CC.

Figure 8.

Estimated RMSE of each subject using different algorithms.

Figure 9.

Estimated CC of each subject using different algorithms.

Figure 10.

Estimation results of B#5. FE is the abbreviation of feature engineering (a) Comparison results between LSSVM with FE and RBF with FE. (b) Comparison results between RBF and RBF with FE. (c) Comparison results between LSSVM with FE and LSSVM. (d) Estimation results of LSSVM with FE. (e) Comparison results between LSSVM with FE and other models. (f) Zoomed portion of (e).