sEMG-Based Robust Recognition of Grasping Postures with a Machine Learning Approach for Low-Cost Hand Control

The design and control of artificial hands remains a challenge in engineering. Popular prostheses are bio-mechanically simple with restricted manipulation capabilities, as advanced devices are pricy or abandoned due to their difficult communication with the hand. For social robots, the interpretation of human intention is key for their integration in daily life. This can be achieved with machine learning (ML) algorithms, which are barely used for grasping posture recognition. This work proposes an ML approach to recognize nine hand postures, representing 90% of the activities of daily living in real time using an sEMG human-robot interface (HRI). Data from 20 subjects wearing a Myo armband (8 sEMG signals) were gathered from the NinaPro DS5 and from experimental tests with the YCB Object Set, and they were used jointly in the development of a simple multi-layer perceptron in MATLAB, with a global percentage success of 73% using only two features. GPU-based implementations were run to select the best architecture, with generalization capabilities, robustness-versus-electrode shift, low memory expense, and real-time performance. This architecture enables the implementation of grasping posture recognition in low-cost devices, aimed at the development of affordable functional prostheses and HRI for social robots.

Relevance of grasp types to assess functionality for personal autonomy

STUDY DESIGN

Cross-sectional research design.

INTRODUCTION

Current assessment of hand function is not focused on evaluating the real abilities required for autonomy.

PURPOSE OF THE STUDY

To quantify the relevance of grasp types for autonomy to guide hand recovery and its assessment.

METHODS

Representative tasks of the International Classification of Functioning, Disability and Health activities in which the hands are directly involved were recorded. The videos were analyzed to identify the grasps used with each hand, and their relevance for autonomy was determined by weighting time with the frequency of appearance of each activity in disability and dependency scales. Relevance is provided globally and distinguished by hand (right-left) and bimanual function. Significant differences in relevance are also checked.

RESULTS

The most relevant grasps are pad-to-pad pinch (31.9%), lumbrical (15.4%), cylindrical (12%), and special pinch (7.3%) together with the nonprehensile (18.6%) use of the hand. Lumbrical grasp has higher relevance for the left hand (19.9% vs 12%) while cylindrical grasp for the right hand (15.3% vs 7.7%). Relevancies are also different depending on bimanual function.

DISCUSSION

Different relative importance was obtained when considering dependency vs disability scales. Pad-to-pad pinch and nonprehensile grasp are the most relevant grasps for both hands, whereas lumbrical grasp is more relevant for the left hand and cylindrical grasp for the right one. The most significant difference in bimanual function refers to pad-to-pad pinch (more relevant for unimanual actions of the left hand and bimanual actions of the right).

CONCLUSIONS

The relative importance of each grasp type for autonomy and the differences observed between hand and bimanual action should be used in medical and physical decision-making.

LEVEL OF EVIDENCE

N/A.

Using kinematic reduction for studying grasping postures. An application to power and precision grasp of cylinders

The kinematic analysis of human grasping is challenging because of the high number of degrees of freedom involved. The use of principal component and factorial analyses is proposed in the present study to reduce the hand kinematics dimensionality in the analysis of posture for ergonomic purposes, allowing for a comprehensive study without losing accuracy while also enabling velocity and acceleration analyses to be performed. A laboratory study was designed to analyse the effect of weight and diameter in the grasping posture for cylinders. This study measured the hand posture from six subjects when transporting cylinders of different weights and diameters with precision and power grasps. The hand posture was measured using a Vicon(®) motion-tracking system, and the principal component analysis was applied to reduce the kinematics dimensionality. Different ANOVAs were performed on the reduced kinematic variables to check the effect of weight and diameter of the cylinders, as well as that of the subject. The results show that the original twenty-three degrees of freedom of the hand were reduced to five, which were identified as digit arching, closeness, palmar arching, finger adduction and thumb opposition. Both cylinder diameter and weight significantly affected the precision grasping posture: diameter affects closeness, palmar arching and opposition, while weight affects digit arching, palmar arching and closeness. The power-grasping posture was mainly affected by the cylinder diameter, through digit arching, closeness and opposition. The grasping posture was largely affected by the subject factor and this effect couldn't be attributed only to hand size. In conclusion, this kinematic reduction allowed identifying the effect of the diameter and weight of the cylinders in a comprehensive way, being diameter more important than weight.