The hand is a vital, multifunctional organ for humans that plays an important role in physical activities and in the development of fine and gross motor skills. Any disease disrupting the hand function will lead to disability and, consequently, worsen the quality of human life. The need to develop hand rehabilitation strategies became a vital issue with an increase in the number of patients (and old people in particular) who suffer from impairments in their hand motor function after strokes and injuries. The strategies for hand rehabilitation depend on maintaining the muscles' ability and preventing their atrophy by doing repetitive hand exercises . As traditional, human-assisted therapy is challenging and expensive, many hand rehabilitation devices or robot-assisted therapies were recently introduced and achieved excellent results by enhancing the motor function of the muscles. Those devices offer a new kind of physiotherapy; in this way, patients practice moving their muscles by following or withstanding the robot's force. Recently, there have been significant developments with the advances in robotic technology, such as the exoskeleton and bioengineering, that have become a significant addition to physiotherapy methods.

As a result of the variety of hand rehabilitation protocols, different robots have been designed to run alternative therapies. The rehabilitation clinics have different types of stationary hand rehabilitation devices which are set up in the clinics and have acquired good results, especially with the computer interface assistance that increases the patients' motivation . The availability of these devices is still insufficient for the majority of patients, due to the long time required for motor recovery training and the expensive fees for using the equipment. Also, these devices do not help them directly in their daily life activities.

Portable devices introduce an alternative to avoid some of these issues as the patients can train themselves at home. The bulky exoskeleton device is the most common type of portable assistive system which supplies highly accurate motion. It is a mechanism of rigid links fixed onto the palm side of the hand and controls the hand–finger motion . Despite its excellent results, the bulky size and complicated structure present a burden for the patient and limit its workspace area. Also, the rigid structures of some of these devices are impeding the therapeutic potential of robotics by reducing their biomimetic qualities, and the motion in non-actuated directions, such as finger abduction, could include having rigid axes of rotation that become misaligned with the finger's anatomic axis during motion .

In contrast, soft rehabilitation devices, which are simple and lightweight, solve this problem by using the natural exoskeleton of the hand and wearable gloves to control hand motion. These devices have been fabricated from easily deformable materials such as fluids, gels, and soft polymers that have better biomimetic qualities, due to their increased compliance and inconstancy, while complying with the contours of the human body. The lack of rigid components removes constraints on non-actuated degrees of freedom and reduces joint alignment issues, which could prevent joint damage. The main problem of this type of device is its actuation system, as most of them are actuated by the pneumatic system, which has a large complex volume and low accuracy . Recently, a soft glove for hand rehabilitation with a cable-driven mechanism was introduced . Another trial was made by , who proposed a cable-driven structure of the hand exoskeleton system for virtual reality purposes. Cherian et al. (2018) introduced the Exo-Glove, the most popular one, which is suitable for the extension and flexion of the finger. However, tendon-driven systems have a major limitation in that they induce significant joint reaction forces. In and Cho (2013) indicated that, since the tendons are attached further away with a larger moment arm compared to the actual human tendons, these forces should not be a problem unless the spasticity of the patient's fingers is very high.

Several types of actuators are used in hand rehabilitation systems. The electric motors are the standard type due to their availability, reliability, and easy torque control with high precision, but the rigid structure may affect safety. Although the pneumatic actuator has fewer requirements for maintenance and might be stopped under a load without causing damages, the difficulty in controlling it and the air storage bulk size limits its use. Also, the hydraulic actuator has the same problem as the size of its parts; however, it may supply higher actuated torque and can be controlled with high precision, as reported by Yue et al. (2017). There are some trials for using smart actuators in hand rehabilitation systems to decrease the disadvantage of the previous actuators; for example, Tang et al. (2013) designed an exoskeleton system to assist hand rehabilitation exercises based on a shape memory alloy actuator. While it offered a light and straightforward structure, its temperature-dependent nature makes precise control difficult.

Using dielectric elastomer materials (DEMs) as a smart actuator for the system might be a remedy for the common defects of actuation methods. The dielectric elastomer material has the closest material properties for natural muscles, particularly in the stress and strain levels. As mentioned by Pelrine et al. (2002) and Madden (2008), its strain is between 10 % and 100 %, which is similar to or higher than in muscle. Also, the stress is 0.1–8 MPa. Work density in the muscle may reach 40 kJm-3, compared to between 10 and 150 kJm-3 in silicone and VHB-based elastomers, respectively. The continuous power output of dielectric elastomers also matches or exceeds that of our skeletal muscle, with human muscle producing about 50 Wkg-1 compared to about 400 Wkg-1 in the elastomers. In addition, it does not have a temperature-dependent response like the shape memory alloys so that it can work in different environments.

As the design of the hand rehabilitation robotics is different from the traditional design of robots because it involves humans, the safety and availability of the device are essential requirements for the rehabilitation system. This research aims to design a bio-mimic, soft hand rehabilitation system that must be portable, simple, and cheap. A cable-driven mechanism is proposed to work instead of the rigid links and be actuated by the dielectric elastomer actuator. This design mimics the natural procedure of finger motion as the cable-driven and the dielectric elastomer actuator (DEA) represent, instead, the natural tendon and muscle, respectively. While there are a variety of hand rehabilitation devices for the patients whose therapy needs are different to others, the target patients for this rehabilitation system are those who have a solitary impairment of their hand extension, as in peripheral radial nerve compression or injury. Nicholls and Furness (2019) indicated that most patients respond well to conservative therapy. They are considering relative rest from offending muscle activity, such as limiting repetitive pronation, supination, wrist flexion, and ulnar deviation. If symptoms are not resolved with the cessation of activity and rest, then splinting is considered.

The rest of the paper is organized as follows. The kinematic and dynamic analysis of the natural hand is given in Sect. 2. The proposed design of the cable system is presented in Sect. 3. The development and fabrication of the DEA are presented in Sect. 4, while Sect. 5 presents the performance characterization of the DEA. The conclusion and future work are presented in Sects. 6 and 7, respectively.

The natural hand is a complex combination of bones, muscles, and ligaments which determine the direction and range of motion. The natural hand has several kinds of motion such as flexion and extension and adduction and abduction for each finger and opposing movements between the thumb and other fingers, which present 27 DOF, as mentioned in Agur and Lee (1999). This large degree of freedom (DOF) makes the rehabilitation of all simultaneous motion challenging. For this research, we focus on the patients who have problems in their hand extension motion, so the other types of motion are being neglected, and each finger will be represented as a planar 3R mechanism. The patients will control their hand's flexion motion while the wiring systems will control the hand extension, as shown in Fig. 1. The fingertip pose Pi for each finger can be defined as a function of joint coordinates and described by the following equations: 1Pi=PxiPyiφi=l1iCθ1i+l2iCθ12i+l3iCθ123il1iSθ1i+l2iSθ12i+l3iSθ123iθ1i+θ12i+θ123i,

where i is the number of fingers (i=1,2,3,4) excluding the thumb, Px and Py are the position of fingertip position in x and y direction, l1i, l2i, l3i are the proximal, middle, and distal phalanges lengths, respectively, and θ1, θ2, and θ3 are the angular displacement around the z axis for the three revolute joints. Table 1 shows the average dimension of the human hand finger which was reported by Serbest et al. (2018). According to the Lagrangian form, the equation of motion of each finger can be defined as follows: 2M(θ)θ¨+C(θ,θ˙)θ˙+g(θ)=τ, where M(θ), C(θ,θ˙), and g are the joint inertia matrix, the matrix of the centrifugal and Coriolis torques, and the gravity vector, respectively. τ is the required joint torque to move the finger, and it can be described as the difference between τa, the active joint torque (from active muscle or the external actuator), and τP, the passive resistance torque (from the natural elasticity and damping of human joints), as follows: 3τ=τa-τP.

By using biomechanical models, as reported in Kuo and Deshpande (2012) and Agarwal et al. (2013), a passive joint torque for the various joints of the finger was calculated. To estimate the actuator's power requirements for the fingers' extension, a rigid dynamic model of the natural hand is developed by using ANSYS workbench software. In this model, rigid bodies to model the hand parts, such as the lower arm, the palm, and the phalanges, have been assumed. The links between the phalanges are described by rotational joints having fixed positions and fixed revolute axes. The thumb movements and both the abduction and adduction movements of the fingers are neglected. The wrist is fixed. The joint orientation was the input, and the joint torque τ was the output. The required joint torques from the actuator to extend the fingers joints from the relaxation state in (N mm) are presented in Table 2. Figure 2 shows the active metacarpal phalangeal (MCP) joint torque for the index finger to extend the finger from the relaxed position (θ=35∘) to (θ=0∘).

The proposed hand rehabilitation mechanism is physically connected to the natural hand exoskeleton, so it has a direct correspondence with the human joints and limbs. The device has been designed to exercise the extension motion for each finger, excluding the thumb. Using the cable-driven system instead of rigid links decreases the number of DOF from three to one as the linear motion of the cable wire moves the three revolute joints of the finger instantly. The system structure has been designed independently for each finger phalanx, as shown in Fig. 3a. A pair of small plastic links with pulleys are used to define a specific cable path around each finger joint. These links save the cable wire tension and prevent injuries due to the contact between the finger skin and the wire. These pairs of links are related to other fixed links built above the finger to support the system and are fabricated from Teflon to decrease the friction losses. To make the system more compact, the actuator will be on the hand's dorsal side. The kinematic model of the cable-driven system (CDS) was analysed to refer to the relationships between the cable lengths and the motion of the finger.

To supply the required tension force for the cable-driven mechanism, a dielectric elastomer (DE) actuator has been designed and fabricated. The spring-rolled DE actuator is a linear actuator that extracts under the application of a high voltage and then returns to its original length after releasing the voltage. So, for this application, the actuator motion is inversely proportional to the hand finger motion, as the actuator will be activated for fingers' flexion and deactivated for fingers' extension.

A DE actuator is basically a compliant capacitor, where a soft, thin (µm) elastomer film is sandwiched between two compliant electrodes. When applying an electric voltage between the electrodes, plus charges on one electrode attract minus charges on the other. Then, electrostatic forces are the result and cause polymer stretching. The electrostatic pressure Pe depends on the applied voltage V and the thickness of the DE film t, according to the following relation: 18Pe=ε0εr×Vt2, where ε0 is the free space permittivity (8.85×10–12 AsVm-1), and εr is the relative dielectric constant of the elastomer. However, the DEAs require a high voltage to work, and the supplied current is very low and with a magnitude in micro-amperes, so there is no fear of danger in using the high voltage as reported by Pourazadi et al. (2017).

Kofod (2008) indicated that the pre-strain of the DE material has a good impact on the actuator performance and enhances the breakdown strength. Due to the need to balance the pre-stretch of DE material and sustain the reaction force, Kovacs et al. (2008) proposed different methods to maintain the required pre-stretch. The spring roll actuator represents a common solution for this problem and achieves a large activation force with DE material. Since the force resulting from one layer of rolled DE material is not enough, several layers should be stacked together and warped around an elastic compressed spring core to increase the resulting force. The balance of the forces' reactions between the compressed spring and the pre-strained DE material is extremely vital in preventing the failure of the actuator.

In order to determine the actuator performance, uniaxial tension tests in the deactivated stage and isotonic and isometric tests in the activated stage have been carried out. The isotonic test gives the relative change in the actuator's length vs. the voltage under a constant load. The isometric test measures the resulting force vs. the voltage, while the actuator's length remains constant.

The design of a soft hand rehabilitation system, based on a cable-driven mechanism, to exercise the fingers' extension movements was proposed. A kinematic and dynamic analysis of the human hand was obtained to determine the required torque to extend the fingers. The cable-driven design was analysed, and a prototype of the system was fabricated and assembled. The dielectric elastomer actuator was introduced as a smart actuator for the system. Theoretical analysis of the actuator was done to choose its suitable parameter. Several DE actuators were fabricated manually. Uniaxial, isotonic, and isometric tests for DE actuator were done to characterize the actuator's performance. A spring roll dielectric elastomer actuator with maximum axial displacement (11.3 mm) and maximum axial force (4.35 N) was produced. The number of the required DE actuators for the hand rehabilitation system is four, as each actuator can only move one finger.

Future work will focus on evaluating the performance of the cable-driven system on several patients. The evaluation will include its efficacy, safety, and comfort.