EMG Controlled Robotic Hand

Project Title

EMG Controlled Robotic Hand

Project Area of Specialization Electrical/Electronic EngineeringProject Summary

Historically, the ability to control a system with hand gestures has been limited. Gesture control often required bulky equipment or relied on image processing to track user motion within the viewing range of a camera. This project addressed this limitation by developing a lightweight system that is controlled only by hand gestures detected via electromyography. Hand gestures are detected by analysing the EMG signals produced by muscle activity in a user’s arm. The EMG signals from the arm are captured by the Myo Gesture Control Armband. We developed an algorithm to process the EMG data to quickly and accurately recognize three unique hand gestures. It is these hand gestures that are used to control the camera system. In addition to getting this system running, we explore advanced methods of pattern recognition, including neural networks and support vector machines. We achieve fairly high accuracy using a pattern recognition neural network, and are confident that given more time and effort, this method can be a viable form of gesture detection. From the start of this project, our goal is to collect and analyse raw EMG data with the intention of putting it to use in a control system. Our goal would be accomplished by implementing an algorithm to detect gestures and control a camera system. Additionally, we have laid the groundwork for future projects in the EMG based HMI field by starting the research into more accurate methods of gesture detection. The device was developed with focus on support of the rehabilitation process after hand injuries or strokes. As the device is designed for the later use on patients, which have limited hand mobility, fast undesired movements have to be averted. Safety precautions in the hardware and software design of the system must be taken to ensure this. The construction allows controlling the motion of finger joints. However, due to friction in gears and mechanical construction it is not possible to move finger joints within the construction without help of actuators. Therefore force sensors are integrated into the construction to measure force exchanged between human and exoskeleton. These allow the human to control the movements of the hand exoskeleton which is useful to teach new trajectories, for muscle training, or for diagnostic purposes. The control method using electromyography (EMG) sensor presented in this paper uses the EMG sensor values to generate a trajectory, which is executed by a position control loop based on sliding mode control

Project Objectives

The current market for gesture-based control of systems rely solely on the use of cameras to detect user movements. These systems require heavy processing and restrict the user to gesture only in the field of view of the cameras. To address these issues, this project created an EMG-based controlled system with the following goals.

A. Acquire EMG data from a user

The EMG data must be collected wirelessly so as to not restrict the user. The wireless communication needs to be reliable and quick to connect. Additionally, the data must be sampled at a rate high enough for real-time operation.

B. Detect different user hand gestures in real time

This system uses three different hand gestures to control it: a fist, wave inward, and wave outward. It has only been tested on the right hand, though it should be possible to use any hand. The system needs a calibration mode to allow for anybody to use it. The calibration should be quick and allow for fast and accurate gesture recognition. Users must receive feedback about the state of the system through the console.

C.Implement gesture detection to control a system

The system is comprised of two cameras, each attached to its own servo motor. The hand gestures allow the user to adjust the position of the motors, as well as the camera feed that is displayed on an external monitor. The motors rotate 180°, 90° in each direction from the initial position. The cameras operate at 30 frames per second and 720p resolution.

• The very main goal of project is to develop a non-invasive bionic hand model which should capable of grabbing almost any type of object and can use almost any hand gesture movement.
• To develop a mobile application for using hand in an effective way to complete a list of functionalities useful for activities of daily living through Bluetooth.
• Haptic feedback through vibration for alerts able to provide some kind of bio feedback to the users by means of different interfaces.
•  To allow a user’s to manipulate their grip according to object efficiently. Our project motivation is to develop a bionic hand which is easily functional for the users without any use of surgery. Bionic hand attachment to patients is rather a painful procedure in both invasive and non-invasive method. It takes time to adjust with your body so in first few months users has to do a lot muscle contraction in grabbing an object for that purpose we are using FSR on the tip of bionic hand finger which help and easy the grasping of object so patients does not have contract their muscles to hard.

Project Implementation Method

We motivated how the robust and anticipatory nature of our method has potential to improve the fluency for human-exoskeleton interaction in the augmentative case. The effects of the different controllers on human-exoskeleton fluency measures using motion capture and survey data collected during this experiment is ongoing work. The desire for improved fluency in the assistive case, where the exoskeleton aids in the movement of a non-healthy individual but is not doing so in a rehabilitative context is also of interest. For assistive exoskeletons, the robustness of our method to sensor placement may mean greater ease of use for at-home use cases. However, additional complexities arise when considering the assistive and rehabilitative use case. Progressive robot-assisted therapy, such as those using the wearable MyoPro exoskeleton or the MIT Manus manipulandum typically use muscle-specific sensors and threshold-based activation to encourage the association between certain muscle activations and a kinematic response in an assist-as-needed framework. There are several potential issues with using ILEXOS, and LfD in general, for a rehabilitative exoskeleton. First, it may be difficult or impossible for patients to train the system because they are unable to perform the requisite motions. This may be addressed by allowing subject intentions to be explicitly communicated via a switch, rather than inferred via pressure, while the patient attempts the action. Supervised labeling of intention may be most applicable in cases where the underlying sEMG signal is a weaker version of the normal signal, such as in multiple sclerosis. In contrast, pathologies that produce other abnormal sEMG activations, such as stroke, may result in significantly lower classification accuracy. The second concern relates to using an LfD mapping within therapy. While progressive therapy may be possible with the ILEXOS LfD framework by gradually increasing the confidence threshold needed to activate a particular action, the use of a personalized classifier may be problematic. The LfD mapping that makes ILEXOS easier for a healthy person to use may hinder rehabilitation as the users are encouraged to continue expressing the correctly classified sEMG signals that are nonetheless inappropriate for movement without the exoskeleton. This concern may be slightly alleviated by placing sEMG sensors on specific muscle

bodies, but the concern remains that a well-mapped LfD system may be actuating with sEMG signals that are not conducive to rehabilitation. A model of appropriate changes in sEMG over the course of rehabilitation, or re-learning sEMG models in staged intervals may be necessary for a rehabilitation context. Further work is required to better understand the behavior of this kind of learning exoskeleton with patients undergoing rehabilitation, and how it may affect the rehabilitation process

Benefits of the Project

• The main goal of rehabilitation is the reintegration of affected people back into normal life in an optimal way
• Comparing the amount of time it takes an autonomous robot to map a world to the time it takes a robot supervised by biosignals
• Further research and experiments in this area are viable and interesting from both a robotics and neuroscience perspective.
• The EMG signal is easily affected by factors such as muscle fatigue and sweating, the recognition results are unstable, and the operator needs to maintain a posture for a long time to continue to exert force, which is poor in experience
• It realizes the real-time interaction between human and robotic arm, which proves the feasibility of using surface electromyography signals to control the robotic arm and control external equipment for other physiological electrical signals in the future

Technical Details of Final Deliverable

• Filter: A filter that is commonly used to reduce this noise is called a low pass filter. It consists of two electronic components, a resistor and a capacitor. Based on the values of these components, each low pass filter has a cut off or corner frequency where the strength of the signals that are higher than that frequency begins to diminish. These low pass and high pass filter can also be paired together to create a band pass filter where only a range of frequencies can pass through. A practical, physical way of reducing noise in a circuit is to ensure that all wires and leads are as short as possible; the less wire and moving electrons to interact with, the smaller noise will be. Even once noise has been mostly taken care of, sometimes the signal is still too small.
• Amplification:  An instrumentation amp takes in two distinct signals and amplifies only the difference between them. These amplifiers can also be seen as providing some of their own filtering; noise that the two signals have in common is not amplified and not passed through the instrumentation amp. The AD623 is an integrated, single- or dual-supply instrumentation amplifier that delivers rail-to-rail output swing using supply voltages from 2.7 V to 12 V. ... The AD623 has a wide input common-mode range and amplifies signals with common-mode voltages as low as 150 mV below ground

• EMG sensor: Electromyography(EMG) sensors work by detecting electrical impulses sent out by muscles when they contract. Surface EMG(sEMG) sensors have electrodes that stick on top of the skin, and detect these impulses through the skin. This impulse is detected as a voltage. Because voltage must be the difference between two points by definition, there must be several electrodes used in an EMG sensor. One of these electrodes is referred to as the ground node, which is connected to ground. There are also several sensing electrodes placed on other parts of the muscle to detect signals. The sensor detects a signal when the voltage between two points change.

• Servo motor: A servo motor is a device that has a motor that spins an axle and can wind and unwind a cord from that spindle.

• Tendons:  Most gloves have used actuators for each finger and routed their artificial tendons in a specific way, though they only controlled two fingers and the thumb, this glove used thimble-like structures over the top of the fingers that resulted in no slipping of the tendons and comfort for the user. Additionally, they used two tendons for each finger and thumb to mimic real tendons in the hand and transmit normal forces to the body.

• Control system:

  A control system is here defined as the program that takes in sensor data, makes decisions based on that data, and then controls the movement of a glove. There have been several ways that researchers have controlled their various devices. Most gloves use the simplest of control systems: pre-programmed directives

Final Deliverable of the Project Hardware SystemCore Industry TelecommunicationOther IndustriesCore Technology Wearables and ImplantablesOther TechnologiesSustainable Development Goals Good Health and Well-Being for PeopleRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	29229
Resistors	Equipment	15	15	225
Capacitors	Equipment	10	210	2100
Battery (3.7v)	Equipment	4	590	2360
Battery (1.85v)	Equipment	4	426	1704
AD623 Amplifier	Equipment	2	2400	4800
Servo Motors	Equipment	7	340	2380
Arduino Mega 256	Equipment	2	2700	5400
Male to male connecting Wires	Equipment	2	250	500
Male to female connecting Wires	Equipment	2	250	500
Surface Electrode	Equipment	12	125	1500
TLC227x	Equipment	2	1980	3960
Soldering Iron	Miscellaneous	1	1500	1500
PCB drill machine kit	Miscellaneous	1	2300	2300