Dimensionality reduction for classification of object weight from electromyography

Electromyography (EMG) is a simple, non-invasive, and cost-effective technology for measuring muscle activity. However, multi-muscle EMG is also a noisy, complex, and high-dimensional signal. It has nevertheless been widely used in a host of human-machine-interface applications (electrical wheelchairs, virtual computer mice, prosthesis, robotic fingers, etc.) and, in particular, to measure the reach-and-grasp motions of the human hand. Here, we developed an automated pipeline to predict object weight in a reach-grasp-lift task from an open dataset, relying only on EMG data. In doing so, we shifted the focus from manual feature-engineering to automated feature-extraction by using pre-processed EMG signals and thus letting the algorithms select the features. We further compared intrinsic EMG features, derived from several dimensionality-reduction methods, and then ran several classification algorithms on these low-dimensional representations. We found that the Laplacian Eigenmap algorithm generally outperformed other dimensionality-reduction methods. What is more, optimal classification accuracy was achieved using a combination of Laplacian Eigenmaps (simple-minded) and k-Nearest Neighbors (88% F1 score for 3-way classification). Our results, using EMG alone, are comparable to other researchers’, who used EMG and EEG together, in the literature. A running-window analysis further suggests that our method captures information in the EMG signal quickly and remains stable throughout the time that subjects grasp and move the object.

Planning to repose long and heavy objects considering a combination of regrasp and constrained drooping

Purpose This paper aims to present a hierarchical motion planner for planning the manipulation motion to repose long and heavy objects considering external support surfaces. Design/methodology/approach The planner includes a task-level layer and a motion-level layer. This paper formulates the manipulation planning problem at the task level by considering grasp poses as nodes and object poses for edges. This paper considers regrasping and constrained in-hand slip (drooping) during building graphs and find mixed regrasping and drooping sequences by searching the graph. The generated sequences autonomously divide the object weight between the arm and the support surface and avoid configuration obstacles. Cartesian planning is used at the robot motion level to generate motions between adjacent critical grasp poses of the sequence found by the task-level layer. Findings Various experiments are carried out to examine the performance of the proposed planner. The results show improved capability of robot arms to manipulate long and heavy objects using the proposed planner. Originality/value The authors’ contribution is that they initially develop a graph-based planning system that reasons both in-hand and regrasp manipulation motion considering external supports. On one hand, the planner integrates regrasping and drooping to realize in-hand manipulation with external support. On the other hand, it switches states by releasing and regrasping objects when the object is in stably placed. The search graphs' nodes could be retrieved from remote cloud servers that provide a large amount of pre-annotated data to implement cyber intelligence.

Decoding Object Weight from Electromyography during Human Grasping

Human urges, desires, and intentions manifest themselves in voluntary action. The final stages of such voluntary action are the muscle contractions that bring it about. Electromyography (EMG) signals measure such muscle contractions. Decoding action contents from EMG require advanced methods for detection, decomposition, processing, and classification and remains a challenge in neuroscience. This study presents a new, time-domain method of classifying EMG for grasping different types of objects. Our proposed method can classify objects with different weights with an accuracy of up to 90%. This progress in neuroscience affects other fields like physiology, brain-computer interfaces, robotics, and so on.

A Scooping-Binding Robotic Gripper for Handling Various Food Products

Food products are usually difficult to handle for robots because of their large variations in shape, size, softness, and surface conditions. It is ideal to use one robotic gripper to handle as many food products as possible. In this study, a scooping-binding robotic gripper is proposed to achieve this goal. The gripper was constructed using a pneumatic parallel actuator and two identical scooping-binding mechanisms. The mechanism consists of a thin scooping plate and multiple rubber strings for binding. When grasping an object, the mechanisms actively makes contact with the environment for scooping, and the object weight is mainly supported by the scooping plate. The binding strings are responsible for stabilizing the grasping by wrapping around the object. Therefore, the gripper can perform high-speed pick-and-place operations. Contact analysis was conducted using a simple beam model and a finite element model that were experimentally validated. Tension property of the binding string was characterized and an analytical model was established to predict binding force based on object geometry and binding displacement. Finally, handling tests on 20 food items, including products with thin profiles and slippery surfaces, were performed. The scooping-binding gripper succeeded in handling all items with a takt time of approximately 4 s. The gripper showed potential for actual applications in the food industry.

Lift observation conveys object weight distribution but partly enhances predictive lift planning

Observation of object lifting allows updating of internal object representations for object weight, in turn enabling accurate scaling of fingertip forces when lifting the same object. Here, we investigated whether lift observation also enables updating of internal representations for an object's weight distribution. We asked participants to lift an inverted T-shaped manipulandum, of which the weight distribution could be changed, in turns with an actor. Participants were required to minimize object roll (i.e. 'lift performance') during lifting and were allowed to place their fingertips at self-chosen locations. The center of mass changed unpredictably every third to sixth trial performed by the actor and participants were informed that they would always lift the same weight distribution as the actor. Participants observed either erroneous (i.e. object rolling towards its heavy side) or skilled (i.e. minimized object roll) lifts. Lifting performance after observation was compared to lifts without prior observation and to lifts after active lifting, which provided haptic feedback about the weight distribution. Our results show that observing both skilled and erroneous lifts convey an object's weight distribution similar to active lifting, resulting in altered digit positioning strategies. However, minimizing object roll on novel weight distributions was only improved after observing error lifts and not after observing skilled lifts. In sum, these findings suggest that although observing motor errors and skilled motor performance enables updating of digit positioning strategy, only observing error lifts enables changes in predictive motor control when lifting objects with unexpected weight distributions.

Assessment of Upper Limb Movement Impairments after Stroke Using Wearable Inertial Sensing

Precise and objective assessments of upper limb movement quality after strokes in functional task conditions are an important prerequisite to improve understanding of the pathophysiology of movement deficits and to prove the effectiveness of interventions. Herein, a wearable inertial sensing system was used to capture movements from the fingers to the trunk in 10 chronic stroke subjects when performing reach-to-grasp activities with the affected and non-affected upper limb. It was investigated whether the factors, tested arm, object weight, and target height, affect the expressions of range of motion in trunk compensation and flexion-extension of the elbow, wrist, and finger during object displacement. The relationship between these metrics and clinically measured impairment was explored. Nine subjects were included in the analysis, as one had to be excluded due to defective data. The tested arm and target height showed strong effects on all metrics, while an increased object weight showed effects on trunk compensation. High inter- and intrasubject variability was found in all metrics without clear relationships to clinical measures. Relating all metrics to each other resulted in significant negative correlations between trunk compensation and elbow flexion-extension in the affected arm. The findings support the clinical usability of sensor-based motion analysis.

Lift observation conveys object weight distribution but partly enhances predictive lift planning

Observation of object lifting allows updating of internal object representations for object weight, in turn enabling accurate scaling of fingertip forces when lifting the same object. Here, we investigated whether lift observation also enables updating of internal representations for an object’s weight distribution. We asked participants to lift an inverted T-shaped manipulandum, of which the weight distribution could be changed, in turns with an actor. Participants were required to minimize object roll (i.e. ‘lift performance’) during lifting and were allowed to place their fingertips at self-chosen locations. The center of mass changed unpredictably every third to sixth trial performed by the actor and participants were informed that they would always lift the same weight distribution as the actor. Participants observed either erroneous (i.e. object rolling towards its heavy side) or skilled (i.e. minimized object roll) lifts. Lifting performance after observation was compared to lifts without prior observation and to lifts after active lifting, which provided haptic feedback about the weight distribution. Our results show that observing both skilled and erroneous lifts convey an object’s weight distribution similar to active lifting, resulting in altered digit positioning strategies. However, minimizing object roll on novel weight distributions was only improved after observing error lifts and not after observing skilled lifts. In sum, these findings suggest that although observing motor errors and skilled motor performance enables updating of digit positioning strategy, only observing error lifts enables changes in predictive motor control when lifting objects with unexpected weight distributions.New and noteworthyIndividuals are able to extract an object’s size and weight by observing interactions with objects and subsequently integrate this information in their own motor repertoire. Here, we show that this ability extrapolates to weight distributions. Specifically, we highlighted that individuals can perceive an object’s weight distribution during lift observation but can only partially embody this information when planning their own actions.

Towards Autonomous Estimation of Lightweight Object's Mass by a Humanoid Robot during a Precision Grip with Soft Tactile Sensors

The estimation of the intrinsic properties of an unknown ob- ject is a very challenging problem, mainly due the limitations on the tactile technology. In this article we present a method to estimate an ob- ject's weight during a precision grip made by a humanoid robot. Tactile sensors on the ngertips provide information on the 3D force vector dur- ing a movement of grasping and lifting a cup lled with dierent masses (30-100g). Using the force measurements across time, we were able to successfully calculate the object weight for 8 dierent masses in two sce- narios: (i) Manually segmented force measurments and (ii) automatically segmented force measurments. Regarding the manually segmented data, we are able to have repeatable measurement and low deviations from the real value, especially for higher object masses. Regarding the automat- ically segmented data, we are able to identify the various phases of the grasping experiment and use the segmented phases to compute the mass automatically.