Repeatability of grasp recognition for robotic hand prosthesis control based on sEMG data

Abstract

Control methods based on sEMG obtained promising results for hand prosthetics. Control system robustness is still often inadequate and does not allow the amputees to perform a large number of movements useful for everyday life. Only few studies analyzed the repeatability of sEMG classification of hand grasps. The main goals of this paper are to explore repeatability in sEMG data and to release a repeatability database with the recorded experiments. The data are recorded from 10 intact subjects repeating 7 grasps 12 times, twice a day for 5 days. The data are publicly available on the Ninapro web page. The analysis for the repeatability is based on the comparison of movement classification accuracy in several data acquisitions and for different subjects. The analysis is performed using mean absolute value and waveform length features and a Random Forest classifier. The accuracy obtained by training and testing on acquisitions at different times is on average 27.03% lower than training and testing on the same acquisition. The results obtained by training and testing on different acquisitions suggest that previous acquisitions can be used to train the classification algorithms. The inter-subject variability is remarkable, suggesting that specific characteristics of the subjects can affect repeatibility and sEMG classification accuracy. In conclusion, the results of this paper can contribute to develop more robust control systems for hand prostheses, while the presented data allows researchers to test repeatability in further analyses.

Exploring Arm Posture and Temporal Variability in Myoelectric Hand Gesture Recognition

“…We use a low-pass filter as the pre-processing stage and we feed such filtered signals as input for the recognition algorithms without further feature extraction. Such choice allows us to provide a common processing pipeline for all the considered algorithms and was based on preliminary analysis and literature results showing good performance of similar features such as the mean absolute value [23], [24].…”

“…Significant contribution in this direction was provided by the Ninapro database [23], which collects acquisitions totaling 67 subjects performing up to 52 hand movements. It focuses on gesture and subject variability collecting single day sessions, but the authors recently published an extension with data from 10 subjects performing 7 gestures over 5 days, with 2 sessions per day [24]. The recognition of afternoon sessions when training on morning data shows a decrease in accuracy of 27%.…”

“…Available studies use various methods to partition training and test data, even in single-session analysis. Alternatives include 5-fold cross validation [21] or using some of the first contractions [17], [20] or using contractions spaced throughout the session, simulating a more equally distributed division [23], [24]. In order to evaluate their effects, we performed the following train/test splits on each session's data: Training 20% (5-fold linear split alternating 1 fold for training and the rest for test), Training 50% (2-fold linear split, alternating them for training/test), Training 80% (5-fold linear split alternating 4 folds for training and the remaining for test), Training 50%D (10-fold linear split, alternating the odd and the even ones for training/test).…”

Implementing Tactile and Proximity Sensing for Crack Detection

“…On the other hand, they are sensitive to noise. This feature has been previously implemented in surface Electromyography (sEMG) and demonstrated high performance [25], [26]. A Random Forest Classifier was used to determine both the surface pattern of examined material and the size of the detected cracks.…”

Dimensionality of forearm muscle activation in transradial amputees

“…These approaches detect muscle contraction patterns as discrete classes to drive function modules [5]. A variety of classi cation algorithms have been implemented and tested to discriminate muscle contraction patterns, such as support vector machines [6], random forest classi ers [7], linear discrimination analysis [8], and convolution neural networks [9]. Earlier pattern recognition approaches operate at xed speed for each function module, whereas recent development has shown that augmenting the speed with the extraction of proportional 'class activation' information in addition to class labels could improve overall performance [10], [11].…”