Controlling a Nursing Robot Based on sEMG Signal

Biopotential signals such as surface electromyography (sEMG) are common signals that can be used to control machines such as medical and nursing robots. The main objective of this paper is to develop a system to control a nursing robot using a sEMG signal. The robot was designed based on the Asian adult hand using CATIA and manufactured using a 3D printing machine. This robot was designated as a care robot to replace and assist physiotherapists in their work, especially in repetitive tasks. The sEMG signal was used as input to control the movement of the care robot. The raw signal from the sEMG was filtered to avoid the noise signal and irrelevant signals were to be rejected. In this project, the MyoWare muscle sensor attached to electrodes on the upper limbs was used to capture the EMG signal. The different readings of the signal are used as an algorithm for the movement of the care robot. As an actuator, the servo motor was operated in two positions: Flexion and extension in 0 degrees (minimum angle) to 100 degrees (maximum angle). The result shows that the robot can only move up and down in one direction with the right hand. This method is similar to the work that therapists do for their patients, especially in rehabilitation.


Introduction
Recently, research into biopotential has been increasing relatively rapidly, especially in the medical and sports fields.Biopotential is an electrical signal in the form of a voltage generated by the volume conduction of electricity in the living body [1].It is generated by the electrochemical activity of excitable cells located in the nervous, muscular, and glandular systems of the individual body [2].Biopotential in the human body is measured by the electrocardiogram (ECG), electroencephalogram (EEG), electromyogram (EMG) and electrooculogram (EOG).The characterization of biopotentials is based on the activity of the respective organs: heart, brain, muscles, and eyes [3].
The ECG is electrical activity of the heart muscle when that muscle contracts in response to electrical depolarization of the heart muscle.Meanwhile, the EEG is muscle activity when the difference in electrical potential occurs between points on the scalp.These signals are generated by the neurons in the brain.The contraction and relaxation of muscles results in an EMG signal.It is the simplest motor unit, connecting a single muscle fiber to a single motor neuron.Therefore, it generates the forces and movements in the body [4].EOG signals are generated by the activity of eye movement and position.The potential created between the cornea (positive) and the retina (negative) will measure the EOG signal and can then be used for all devices [5].
From previous studies, the biopotential signal has a low frequency, i.e. less than 1 kHz [6].They also have a very small amplitude in the range of 10 μV to 100 μV when measured with surface electrodes.Therefore, biopotential signals, including EEG, ECG, EMG and EOG, can be monitored using a few signals acquisition, processing, and classification techniques.Table 1 shows the characteristics of all signals in terms of frequency, bandwidth, and amplitude, as well as their applications.
Table 1.Bio-potential, characteristic and its application [7] Type of biopotentials Amplitude (mV) In recent years, the hardware implementation of biopotentials, especially EMG, as a signal to control equipment is in high demand in many industries.In medical research, EMG is mainly used in gait and posture analysis and surgery [8].For example, it can help medical staff diagnose diseases related to the spine.Some EMG signals can also be used for rehabilitation purposes.It can be used for active exercise therapy to assist human movements for rehabilitation training [9].However, in modern times, research in the field of sports science related to EMG is increasing rapidly.Most research in sport focuses on biomechanics, movement analysis and sports rehabilitation.EMG research can also be conducted in the field of ergonomics with regard to risk prevention, needs analysis and ergonomic design.
In this paper, the hardware of the care robot is combined with the EMG signal to control the movement of the robot.The care robot was attached to the human arm, and then the electrode sensor was attached to the upper limb for data acquisition.The EMG signal can be used to activate the actuator of the care robot in two positions, either in extension or in flexion.Based on the results, the applicability of the algorithm can be used to control the movement of the care robot and as an aid for disabled people, including rehabilitation training and for strength enhancement in healthy people.

Related Work
The EMG signal is the activity of muscle movement that produces an electrical signal, which is then displayed on the graph.There are two techniques for recording electromyography signals, namely surface and needle electrodes [10].Therefore, the concept of sEMG with surface electrodes is used.The surface electrode method is more convenient and easier to use compared to the needle electrode.In terms of accuracy of data acquisition, the needle electrode is certainly better than the surface electrode, but it can be ignored for control purposes [11].In addition, surface EMG also provides a non-invasive technique, i.e. this treatment is not harmful to the skin for measuring and detecting the EMG signal [12].
For data collection, the EMG signal can be amplified by an EMG circuit that includes an instrumentation amplifier, a high-pass filter, a low-pass filter, and a precision rectifier.The amplification method is important to reduce noise so that the EMG signal can be better and more effectively understood and used for any purpose.Figure 1 shows the concept of EMG data acquisition for robot control.The raw EMG data is amplified using low-pass and high-pass filters to reduce the noise [13].

Figure 1. The way of amplified EMG signal for controlling the robot device
In the previous study, the basic EMG circuit was connected to the three surface electrodes for recording muscle activity.In this project, these electrodes are placed in pairs on the biceps, one in the middle of the muscle, one at the end of the muscle and the last one as a reference on the triceps, which is close to the muscle group.The sEMG signals are recorded in this way using silver-silver chloride (Ag-AgCl) electrodes.This type of electrode is well suited for recording EMG data, even though this electrode is expensive.The placement of the electrode can be seen in Figure 2.Meanwhile, Table 2 shows the summary of the placement of electrodes.

Method of Development the Nursing Robot
In general, the method of the nursing robot system focuses on three phases: sEMG, design and control of the nursing robot.In phase 1, the project started with exploring the concept of the sEMG.The general knowledge about the sEMG and how to capture it was explored by the people previously studied.In this part, the data analzye were collected for classification to control the nursing robot.Meanwhile, in phase 2, the nursing robot was designed using CATIA software based on the size of human upper limbs.Analyse this robot for finite element analysis were also used in finding the appropriate stress and force.Once the design was completed, the robot was manufactured using a 3D printer.The final phase focused on developing the robot's motion control, including the algorithm and circuit combination implemented in the Fritzing software.Finally, the implementation between hardware and software was carried out to test the functionality.Figure 3 shows the flow of the process to produce the grooming robot system.

sEMG data collection and acquisition
Before the sEMG data is collected by the MyoWare sensor, the pattern of its reading on the muscle should be known.At this stage, two movements were considered, flexion and extension.Flexion is a bending movement in which the angle between two parts decreases; the counterpart of the flexor is the extensor.Extension is a stretching movement in which the angle between the forearm and the upper arm increases.The illustration for both movements can be seen in Figure 4.During the flexion movement, the muscle fibers contract until the forearm has reached the maximum range of motion, while during the extension movement, the muscle fibers relax until the forearm has reached the fully extended range of motion (ROM).

sEMG signal pattern recognition
First, the sEMG signal was acquired from the Myoware sensor.The raw EMG signal was processed by filtering of feature extraction.From these signals, two states are classified, either flexion or extension movement.The signal can be read by the controller to control the robotic device.Figure 6 shows the sEMG in tension, fluttering in three states: at rest, bending motion and stretching motion.When the hand starts to bend upwards from the rest position, i.e. during the bending movement, the sensor starts to rise to reach the maximum data of the sEMG.During the other period, the sEMG starts to sink and enters the resting state.

sEMG signal pattern recognition between male and female
In this project, the experiment was conducted by two subjects: -a male with age 28 years old and a female with age 23 years old.The difference between the sexes should provide information about the extent to which one sex can cause different values of the sEMG signal.For this experimental set-up, each subject had to perform 3 complete movements: Flexion -extension -flexion -extension -flexion -extension.Data from the sEMG was collected via an Arduino serial portal.This data was then recorded as shown in Figure .7 and Figure 8. Figure 7 shows five readings describing the pattern of the sEMG in three movement sequences for men.In the first two patterns, the sEMG signal has the highest peak compared to the other three patterns.This is because each subject must be fit or in tune for this movement.The good pattern of the movement in Figure 7 c) and d).Meanwhile, Figure 8 shows the pattern of sEMG for women.The pattern of sEMG is slightly the same for both sexes.Only in Figure 8 a) and e) is there a sudden spike.The sudden increase in readings was caused by the subject touching the electrode while the system was being tested.Although there were some measurement disturbances while performing the experiment, they can be neglected because they were minor and did not affect the overall system.4 are the values for five measured values based on the sEMG in Figure 7 and Figure 8.In this table, the values are considered as minimum and maximum values.The mean value was calculated by adding the maximum and minimum sEMG values and then dividing by 2. The mean value served as the reference point.For the five patterns, the measured values were in agreement, so the average value could be calculated using equation (1).Based on the recorded results, the highest value of the average sEMG mean of the subjects was 3.16 mV and the lowest was 1.77 mV.

Average value =
(1st Reading + 2nd Reading+3 rd Reading + 4th Reading + 5th Reading) 5 (1) The sEMG signal can also be analyzed by gender.From the result, it can be seen that the male value is high compared to the female value.From Figure 9, it can be seen that the sEMG signal is twice as high in males as in females.This is because the muscle strength of the male is higher than that of the female.The minimum value is also slightly twice as high as that of females.

Figure 9. The difference average value of sEMG for both male and female
To control the servo motor, the sEMG value had to be higher than the reference point to increase the servo angle.The servo angle was not rotated when the sEMG value was lower than the reference point.After the servo angle reached its maximum angle, the stretching movement was activated.This movement required the user to slowly extend his arm.When the sEMG value was lower than the reference point, the servo angle decreased to the minimum servo angle, and the system repeated the process of the two movements alternately until the user turned off the robot.

Conclusions
The signal from the sEMG can be detected by the active muscle of the human.Therefore, it can be concluded from this project that the sEMG signal from the muscle can be activated to control any device such as a nursing robot.The sEMG signal can be classified according to its purpose and function.The sEMG is suitable for both sexes with a wide range of sEMG values.Finally, the implementation of both hardware and software components worked as desired.The nursing robot is able to assist the patient in rehabilitation.
However, in future research, the robot can be improved by adding the Internet of Things (IoT) element.With this IoT, the user can control the movement of the robot independently.In terms of speed, actuators with a large torque such as the DC motor can be replaced to make the system smoother and cheaper.Last but not least, high technology in medical devices can help people such as the elderly and disabled live more comfortable and better lives.

Figure 2 .
Figure 2. The placement of electrode

Figure 3 .
Figure 3.The process flow of nursing robot system

Figure 4 .
Figure 4.The flexion movement (left) and the extension movement (right) For this project there are three main parts: -input, control, and output, as shown in Figure 5.The nursing robot was attached to the user's right hand.Then the three electrodes were attached to the user's bicep, which contains the MyoWare sensor.The MyoWare sensor is an electromyographic (EMG) sensor powered by Advancer Technologies using Arduino-powered.It measures the filtered and rectified electrical activity of the muscles with an output of 0V.The generated EMG signals are

Figure 5 .
Figure 5. Implementation hardware and software for nursing robot

Figure 6 .
Figure 6.The signal in two state position: flexion movement and extension movement

Table 3 .
The average value of sEMG signal for male

Table 4 .
The average value of sEMG signal for female

Table 3 and
Table