Electromyography data for non-invasive naturally-controlled robotic hand prostheses

Abstract

Recent advances in rehabilitation robotics suggest that it may be possible for hand-amputated subjects to recover at least a significant part of the lost hand functionality. The control of robotic prosthetic hands using non-invasive techniques is still a challenge in real life: myoelectric prostheses give limited control capabilities, the control is often unnatural and must be learned through long training times. Meanwhile, scientific literature results are promising but they are still far from fulfilling real-life needs. This work aims to close this gap by allowing worldwide research groups to develop and test movement recognition and force control algorithms on a benchmark scientific database. The database is targeted at studying the relationship between surface electromyography, hand kinematics and hand forces, with the final goal of developing non-invasive, naturally controlled, robotic hand prostheses. The validation section verifies that the data are similar to data acquired in real-life conditions, and that recognition of different hand tasks by applying state-of-the-art signal features and machine-learning algorithms is possible.

Figure 1. Acquisition procedure scheme for exercises A, B and C.

The subjects are asked to mimic movies of movement shown on the screen of the laptop. The sEMG signal is recorded through up to 12 electrodes and can be used to test methods to control robotic hand prostheses naturally (the electrode on the flexor digitorum superficialis is not represented due to perspective reasons).

Figure 2. Movements and force patterns divided by exercise.

Exercise A (light blue): 12 basic movements of the fingers; Exercise B (red): 8 isometric and isotonic hand configurations and 9 basic movements of the wrist; Exercise C (green): 23 grasping and functional movements (everyday objects are presented to the subject for grasping, in order to mimic daily-life actions); Exercise D (purple): 9 force patterns; Rest position (white).

Figure 3. Experimental conditions effect on muscular activity for different sub-databases.

Different rows represent different experimental conditions: movement repetition (1st row; subplots a–c); movement (2nd row; subplots d–f); subject (3rd row; subplots g–i). Different columns represent different sub-databases: database 1 (1st column; subplots a,d,g); database 2 (2nd column; subplots b,e,h); database 3 (3rd column; subplots c,f,i). The horizontal central mark in the boxes is the median; the edges of the boxes are the 25th and 75th percentiles; the whiskers extend to approximately 2.7 times the standard deviation.

Figure 4. Experimental conditions effect on hand kinematics for different sub-databases.

Different rows represent different experimental conditions: movement repetition (1st row; subplots a,b); movement (2nd row; subplots c,d); subject (3rd row; subplots e,f). Different columns represent different sub-databases: database 1 (1st column; subplots a,c,e); database 2 (2nd column; subplots b,d,f). The horizontal central mark in the boxes is the median; the edges of the boxes are the 25th and 75th percentiles; the whiskers extend to approximately 2.7 times the standard deviation.

Figure 5. Experimental conditions effect on acceleration sensors for different sub-databases.

Different rows represent different experimental conditions: movement repetition (1st row; subplots a,b); movement (2nd row; subplots c,d); subject (3rd row; subplots e,f). Different columns represent different sub-databases: database 2 (1st column; subplots a,c,e); database 3 (2nd column; subplots b,d,f). The horizontal central mark in the boxes is the median; the edges of the boxes are the 25th and 75th percentiles; the whiskers extend to approximately 2.7 times the standard deviation.

Figure 6. Movement classification results for different sub-databases, classifiers and features.

Different histograms represent different databases: (a) database 1; (b) database 2; (c) database 3. Each group of columns represents a specific classifier (k-nn, k-nearest neighbors; SVM, Support Vector Machine; random forests; LDA, Linear Discriminant Analysis). Different colours represent different features (RMS, Root Mean Square; TD, time domain statistics; HIST, Histogram; mDWT, marginal Discrete Wavelet Transform, normalized combination of all features). The height of each column represents the average accuracy, while the error bar represents the standard deviation.