**2. Prior research related to EMG applications**

In the past, some researchers have reported work related to shape memory alloy (SMA) and other similar actuators to develop the bio-mimetic fingers but IPMC artificial muscle based finger related work is limited. Pfeiffer et al. (1999) have designed artificial limbs and robot prostheses that are lightweight, compact and dexterous. This mimics the human anatomy and maintains a high lifting capability. EMG control is used for SMA actuated fingers in robot prostheses. DeLaurentis & Mavroidis (2002) have discussed the design of a 12 degreeof-freedom (DOF) SMA actuated artificial hand where the SMA wires are embedded intrinsically within the hand structure. Cocaud & Jnifene (2003) have investigated the use of artificial muscles as SMA actuators for robot manipulators. A solution is established in order to determine the optimal position of a muscle in various musculoskeletal configurations. Herrera et al. (2004) have also designed and constructed a prosthesis where linear actuators are used for designing the mechanical system and EMG sensors are introduced for designing the electrical control system. Bundhoo et al. (2005, 2008) have reported the design of artificially actuated finger by SMA towards development of bio-mimetic prosthetic hands. Different finger joints are actuated through SMA wires via EMG and the relationship between elongation/contraction of the SMA wires & the finger joints have been obtained. O'Toole & McGrath (2007) have also focused on mechanical design of a 12 DOF SMA actuated artificial hand. The SMA material is used for combination of high strength polymers such as polytetrafluoroethylene (PTFE), polyether ether ketone (PEEK) and low density metals such as titanium. Lau (2009) have carried out research work on a design and development of an intelligent prosthetic hand based on hybrid actuation through DC motor & SMA wires. These are controlled by myoelectric signal. Two novel features are introduced in the new prosthetic hand. Firstly, its hybrid actuation mechanism has the advantage of increasing the active degrees of freedom and secondly, using only two myoelectric sensors, has the potential for controlling more than three patterns of fingers movements. Pittaccio & Viscuso (2011) have developed a SMA wire device for the rehabilitation of the ankle joint where active orthosis powered by two rotary actuators like, NiTi wire are used to obtain ankle dorsiflexion and EMG signal is used to control the orthosis and trigger activation from muscle. Stirling et al. (2011) have shown the potential of SMA wire for an active, soft orthotic in the knee where NiTi based SMA wires is also used. A prototype is tested on a suspended, robotic leg to simulate the swing phase of a typical gait. Thayer & Priya (2011) have designed a biomimetic dexterous humanoid hand where the dexterity of the DART hand have been measured by quantifying functionality and typing speed on a standard keyboard. The hand consists of 16 servo motors dedicated to finger motion and three motors for wrist motion where some of joints are activated through SMA wires.

Computational Intelligence in Electromyography Analysis – 364 A Perspective on Current Applications and Future Challenges

and rehabilitation technology.

object. During holding the object, stable EMG signal is required. For this purpose, stable EMG signal is sent through proportional–integral–derivative (PID) controller to the system. Experimentally, it is found that IPMC based artificial muscle finger achieves similar movement like human index finger. This IPMC based artificial muscle finger attains deflection upto 12 mm. By developing a prototype of IPMC artificial muscle finger based micro gripper, it is demonstrated that EMG driven system like IPMC artificial muscle finger based micro gripper can be applicable in handling of light weight components. The major advantages of such system are that IPMC based artificial muscle finger tip shows the compliant behavior and consumes less energy for actuation. Therefore, EMG driven system shows enough potential to substitute for conventional mechanism in micro manipulation

This chapter is organized as follows: Section 2 describes the prior research related to EMG applications in robotics and bio-mimetics. Section 3.1 explains the basic design of IPMC artificial muscle finger based micro gripper which is driven by EMG signal. The basic tendon of index finger is studied in section 3.2 where muscles are identified for actuation of IPMC based artificial muscle finger. In section 3.3, a model for controlling the EMG signal is highlighted. Different types of control system are implemented for achieving stable data from EMG signal via index finger which is sent to IPMC based artificial muscle finger. Section 4 discusses experimental testing setup for activation of IPMC based artificial muscle finger by human finger through EMG. In section 5, the results are discussed and the

conclusions are drawn in section 6. The future work is recommended in section 7.

In the past, some researchers have reported work related to shape memory alloy (SMA) and other similar actuators to develop the bio-mimetic fingers but IPMC artificial muscle based finger related work is limited. Pfeiffer et al. (1999) have designed artificial limbs and robot prostheses that are lightweight, compact and dexterous. This mimics the human anatomy and maintains a high lifting capability. EMG control is used for SMA actuated fingers in robot prostheses. DeLaurentis & Mavroidis (2002) have discussed the design of a 12 degreeof-freedom (DOF) SMA actuated artificial hand where the SMA wires are embedded intrinsically within the hand structure. Cocaud & Jnifene (2003) have investigated the use of artificial muscles as SMA actuators for robot manipulators. A solution is established in order to determine the optimal position of a muscle in various musculoskeletal configurations. Herrera et al. (2004) have also designed and constructed a prosthesis where linear actuators are used for designing the mechanical system and EMG sensors are introduced for designing the electrical control system. Bundhoo et al. (2005, 2008) have reported the design of artificially actuated finger by SMA towards development of bio-mimetic prosthetic hands. Different finger joints are actuated through SMA wires via EMG and the relationship between elongation/contraction of the SMA wires & the finger joints have been obtained. O'Toole & McGrath (2007) have also focused on mechanical design of a 12 DOF SMA actuated artificial hand. The SMA material is used for combination of high strength

**2. Prior research related to EMG applications** 

Some of the researchers have focused on the design of a biomechatronic robotic hand using EMG. Cheron et al. (1996) have found the relationship between EMG and the arm kinematics through dynamic recurrent neural networks (DRNN) method whereas Hudgins et al. (1997) have focused on a new control scheme, based on the recognition of naturally myoelectric signal patterns, transfers the burden of multifunction myoelectric control from the amputee to the control system. Rosen et al. (2001) have developed a myosignal-based exoskeleton system. This is implemented in an elbow joint, naturally controlled by the human. The human–machine interface is set at the neuromuscular level, by using the neuromuscular signal (EMG) as the primary command signal for the exoskeleton system. The EMG signals along with the joint kinematics are fed into a myoprocessor (Hill-based muscle model) which in turn predicts the muscle moments on the elbow joint. Banks (2001) has given remarkable effort towards design and control of an anthropomorphic robotic finger with multi-point tactile sensation whereas Light et al. (2002) have emphasized on intelligent multifunction myoelectric control of hand prostheses. Peleg et al. (2002) have extracted multiple features via EMG signal from hand amputees which is selected by help of a genetic algorithm. Fukuda et al. (2003) have developed a prosthetic hand where humanassisting manipulator system based on the EMG signals is utilized. Wheeler (2003) has presented a neuro-electric interface method for virtual device control. The sampled EMG data is taken from forearm and then is fed into pattern recognition software that has been trained to distinguish gestures from a given gesture set. Krysztoforski et al. (2004) have given remarkable effort towards recognition of palm finger movements on the basis of EMG signals with the application of wavelets.

Crawford et al. (2005) have used EMG signals for classifying in real-time with an extremely high degree of accuracy in a robotic arm-and-gripper. A linear support vector machines (SVM) based classifier and a sparse feature representation of the EMG signal are used. Hidalgo et al. (2005) have proposed a design of robotic arm employing fuzzy algorithms to interpret EMG signals from the flexor carpi radialis, extensor carpi radialis and biceps brachii muscles. The control and acquisition systems are composed of a microprocessor, analog filtering, digital filtering & frequency analysis, and finally a fuzzy control system. Mobasser & Hashtrudi-Zaad (2005) have estimated rowing stroke force with EMG signal using artificial neural network method from upper arm muscles which is involved in elbow joint movement, sensed elbow angular position and velocity. Gao et al. (2006) have focused on acquiring the data from the upper limb of the body for robotic arm motion using EMG whereas Frigo et al. (2007) have detected EMG signal from voluntarily activated muscles which is controlled for functional neuromuscular by electrical stimulation. A comb filter (with and without a blanking window) is applied to remove the signal components synchronously correlated to the stimulus. Roy et al. (2007) have compared the performance of different sEMG signal at various conditions. These performances depend on the electromechanical stability between the sensor and its contact with skin. Zollo et al. (2007) have put a remarkable effort on the control system of biomechatronic robotic hand and on the optimization of the hand design in order to obtain human like kinematics and dynamics. By evaluating the simulated hand performance, the mechanical design is iteratively refined. The mechanical structure and the ratio between numbers of actuators to the number of DOF have been optimized. Yagiz et al. (2007) have developed a dynamic model of the prosthetic finger where a non chattering robust sliding mode control is applied to make the model follow a certain trajectory. Wege & Zimmermann (2007) have shown the potential of EMG control for a hand exoskeleton device. The device has been developed with focus on support of the rehabilitation process after hand injuries or strokes. Itoh et al. (2007) have studied the hand finger operation using sEMG during crookedness state of the finger. Two electrodes (Ag/AgCl electrodes) are sticked randomly on the forearm muscles and the intensity of EMG signals at different muscles is measured for each crooked finger.

Design and Control of an EMG Driven IPMC Based Artificial Muscle Finger 367

development procedure of bio-mimetic robot hand and its control scheme where each robot hand has four under-actuated fingers, which are driven by two linear actuators coupled together. Dalley et al. (2009) have given emphasis of an anthropomorphic hand prosthesis that is intended for use with a multiple-channel myoelectric interface. The hand contains 16 joints, which are differentially driven by a set of five independent actuators. Hu et al. (2009) have presented a comparison between electromyography-driven robot and passive motion device on wrist rehabilitation for chronic stroke patients. By comparative study, it was found that the EMG-driven interactive training had a better long-term effect than the

Blouin et al. (2010) have focused on control of arm movement during body motion as revealed by EMG whereas Luo & Chang (2010) have explored a feasibility study on EMG signal integrated with multi-finger robot hand control for massage therapy applications. The forearm EMG of a person massaged by the human hands is recorded and analyzed statistically. Khokhar et al. (2010) have showed the potential of EMG applications where SVM classification technique is suitable for real-time classification of sEMG signals. This technique is effectively implemented for controlling an exoskeleton device. Huang et al. (2010) have designed a robust EMG sensing interface for pattern classification. The aim of this study was to design sensor fault detection (SFD) module through the sensor interface to provide reliable EMG pattern classification. This module monitors the recorded signals from individual EMG electrodes and performs a self-recovery strategy to recover the classification performance when one or more sensors are disturbed. Naik et al. (2010) has studied the pattern classification of myo-electric signal during different maximum voluntary contractions using BSS techniques for a blind person whereas Artemiadis & Kyriakopoulos (2010 & 2011) have presented a switching regime model for the EMG-based control of a robot arm where decode the EMG activity of 11 muscles has a continuous representation of arm motion in the 3-D space. The switching regime model is used to overcome the main difficulties of the EMG-based control systems, i.e. the nonlinearity of the relationship between the EMG recordings and the arm motion, as well as the non-stationary of EMG signals with respect to time. Vogel et al. (2011) have demonstrated the robotic arm/hand system that is controlled in real time in 6 dimension Cartesian space through measured human muscular activity via EMG. DLR Light-weight Robot III is used during demonstration of impedance control. Li et al. (2011) have presented a robot control system using four different gestures from an arm. These are achieved by EMG signal using phase synchrony features. The phase synchrony analysis using the recent multivariate extensions of empirical mode decomposition (MEMD) is carried out. Joshi et al. (2011) have focused on brain-muscle-computer interface using a single sEMG signal. Initial results show that the human neuromuscular system can simultaneously manipulate partial power in two separate frequency bands of a sEMG power spectrum at a single muscle site. Matsubara et al. (2011) have proposed an interface to intuitively control robotic devices using myoelectric signals. Through learning procedure, a set of myoelectric signals is captured from multiple subjects in the system and it can be used as an adaptation procedure to a new user after only a few

continuous passive movement (CPM) treatment.

interactions.

Hao et al. (2008) have studied the design of pneumatic muscle actuator based robotic hand where its compliance and dexterity handling are attempted. A single finger is controlled by fuzzy & PID controller and comparative studies are discussed. Murphy et al. (2008) have explored the micro electro-mechanical systems based sensor for mechanomyography system whereas Saponas et al. (2008) have also explored the feasibility on muscle-computer interaction methodology that directly senses and decodes human muscular activity rather than relying on physical device actuation or user actions. Andrews (2008) has determined an effective approach to finger movement classification in typing tasks using myoelectric data which are collected from the forearm. Cesqui et al. (2008) have explored the use of EMG signals for post-stroke and robot-mediated therapy. In this work, a pilot study has been reported under young and healthy subjects where experiments are conducted to determine whether it is possible to build a static map to cluster EMG activation patterns for horizontal reaching movements. Chen et al. (2008) have implemented an EMG feedback control method with functional electrical stimulation cycling system (FESCS) for stroke patients. The stroke patients often suffer from low limbs paralysis. By designing the feedback control protocol of FESCS, the physiological signal is recorded with help of FPGA biomedical module, DAC and electrical stimulation circuit. Lee et al. (2009) have described a development procedure of bio-mimetic robot hand and its control scheme where each robot hand has four under-actuated fingers, which are driven by two linear actuators coupled together. Dalley et al. (2009) have given emphasis of an anthropomorphic hand prosthesis that is intended for use with a multiple-channel myoelectric interface. The hand contains 16 joints, which are differentially driven by a set of five independent actuators. Hu et al. (2009) have presented a comparison between electromyography-driven robot and passive motion device on wrist rehabilitation for chronic stroke patients. By comparative study, it was found that the EMG-driven interactive training had a better long-term effect than the continuous passive movement (CPM) treatment.

Computational Intelligence in Electromyography Analysis – 366 A Perspective on Current Applications and Future Challenges

Hidalgo et al. (2005) have proposed a design of robotic arm employing fuzzy algorithms to interpret EMG signals from the flexor carpi radialis, extensor carpi radialis and biceps brachii muscles. The control and acquisition systems are composed of a microprocessor, analog filtering, digital filtering & frequency analysis, and finally a fuzzy control system. Mobasser & Hashtrudi-Zaad (2005) have estimated rowing stroke force with EMG signal using artificial neural network method from upper arm muscles which is involved in elbow joint movement, sensed elbow angular position and velocity. Gao et al. (2006) have focused on acquiring the data from the upper limb of the body for robotic arm motion using EMG whereas Frigo et al. (2007) have detected EMG signal from voluntarily activated muscles which is controlled for functional neuromuscular by electrical stimulation. A comb filter (with and without a blanking window) is applied to remove the signal components synchronously correlated to the stimulus. Roy et al. (2007) have compared the performance of different sEMG signal at various conditions. These performances depend on the electromechanical stability between the sensor and its contact with skin. Zollo et al. (2007) have put a remarkable effort on the control system of biomechatronic robotic hand and on the optimization of the hand design in order to obtain human like kinematics and dynamics. By evaluating the simulated hand performance, the mechanical design is iteratively refined. The mechanical structure and the ratio between numbers of actuators to the number of DOF have been optimized. Yagiz et al. (2007) have developed a dynamic model of the prosthetic finger where a non chattering robust sliding mode control is applied to make the model follow a certain trajectory. Wege & Zimmermann (2007) have shown the potential of EMG control for a hand exoskeleton device. The device has been developed with focus on support of the rehabilitation process after hand injuries or strokes. Itoh et al. (2007) have studied the hand finger operation using sEMG during crookedness state of the finger. Two electrodes (Ag/AgCl electrodes) are sticked randomly on the forearm muscles and the intensity of

EMG signals at different muscles is measured for each crooked finger.

Hao et al. (2008) have studied the design of pneumatic muscle actuator based robotic hand where its compliance and dexterity handling are attempted. A single finger is controlled by fuzzy & PID controller and comparative studies are discussed. Murphy et al. (2008) have explored the micro electro-mechanical systems based sensor for mechanomyography system whereas Saponas et al. (2008) have also explored the feasibility on muscle-computer interaction methodology that directly senses and decodes human muscular activity rather than relying on physical device actuation or user actions. Andrews (2008) has determined an effective approach to finger movement classification in typing tasks using myoelectric data which are collected from the forearm. Cesqui et al. (2008) have explored the use of EMG signals for post-stroke and robot-mediated therapy. In this work, a pilot study has been reported under young and healthy subjects where experiments are conducted to determine whether it is possible to build a static map to cluster EMG activation patterns for horizontal reaching movements. Chen et al. (2008) have implemented an EMG feedback control method with functional electrical stimulation cycling system (FESCS) for stroke patients. The stroke patients often suffer from low limbs paralysis. By designing the feedback control protocol of FESCS, the physiological signal is recorded with help of FPGA biomedical module, DAC and electrical stimulation circuit. Lee et al. (2009) have described a Blouin et al. (2010) have focused on control of arm movement during body motion as revealed by EMG whereas Luo & Chang (2010) have explored a feasibility study on EMG signal integrated with multi-finger robot hand control for massage therapy applications. The forearm EMG of a person massaged by the human hands is recorded and analyzed statistically. Khokhar et al. (2010) have showed the potential of EMG applications where SVM classification technique is suitable for real-time classification of sEMG signals. This technique is effectively implemented for controlling an exoskeleton device. Huang et al. (2010) have designed a robust EMG sensing interface for pattern classification. The aim of this study was to design sensor fault detection (SFD) module through the sensor interface to provide reliable EMG pattern classification. This module monitors the recorded signals from individual EMG electrodes and performs a self-recovery strategy to recover the classification performance when one or more sensors are disturbed. Naik et al. (2010) has studied the pattern classification of myo-electric signal during different maximum voluntary contractions using BSS techniques for a blind person whereas Artemiadis & Kyriakopoulos (2010 & 2011) have presented a switching regime model for the EMG-based control of a robot arm where decode the EMG activity of 11 muscles has a continuous representation of arm motion in the 3-D space. The switching regime model is used to overcome the main difficulties of the EMG-based control systems, i.e. the nonlinearity of the relationship between the EMG recordings and the arm motion, as well as the non-stationary of EMG signals with respect to time. Vogel et al. (2011) have demonstrated the robotic arm/hand system that is controlled in real time in 6 dimension Cartesian space through measured human muscular activity via EMG. DLR Light-weight Robot III is used during demonstration of impedance control. Li et al. (2011) have presented a robot control system using four different gestures from an arm. These are achieved by EMG signal using phase synchrony features. The phase synchrony analysis using the recent multivariate extensions of empirical mode decomposition (MEMD) is carried out. Joshi et al. (2011) have focused on brain-muscle-computer interface using a single sEMG signal. Initial results show that the human neuromuscular system can simultaneously manipulate partial power in two separate frequency bands of a sEMG power spectrum at a single muscle site. Matsubara et al. (2011) have proposed an interface to intuitively control robotic devices using myoelectric signals. Through learning procedure, a set of myoelectric signals is captured from multiple subjects in the system and it can be used as an adaptation procedure to a new user after only a few interactions.

Recently, Ahmad et al. (2012) have presented a review report on different techniques of EMG data recording where condition of an ideal pre-amplifier, signal conditioning and its amplification are discussed. Sun et al. (2012) have conducted an isokinetic exercise to realize the characteristics of femoral muscles in human knee movement through EMG where a mechanical model of muscle for human knee movement is established. Qi et al. (2012) have developed algorithms for muscle-fatigue detection and muscle-recruitment patterns in routine wheel chair propulsion scenarios, e.g., daily practice where for analysis purpose two speeds of muscular behavior are chosen. Gandole (2012) has developed an artificial intelligent model using focused time lagged recurrent neural network (FTLRNN) method with a single hidden layer. FTLRNN method reduces noise intelligently from the EMG signal. Chan et al. (2012) have developed an assessment platform for upper limb myoelectric prosthetic devices using EMG. The assessment platform consists of an acquisition module, a signal capture module, a programmable signal generation module and an activation & measurement module. The platform is designed to create a sequence of activation signals from EMG data captured from a patient.

Design and Control of an EMG Driven IPMC Based Artificial Muscle Finger 369

IPMC based artificial muscle finger

Object

is to detect the voltage from human muscles since human muscles generate few millivoltages when they are contracting or expanding during movement. For transferring this voltage signal to actuate the artificial muscle finger, it needs the amplification setup which is discussed in section 4. Therefore, IPMC based artificial muscle finger is activated using EMG

**Figure 1.** Schematic diagram of IPMC artificial muscle finger based micro gripper driven by EMG

Plastic based finger Holder

Human finger EMG electrode

**Figure 2.** Schematic diagram of the actuation mechanism of IPMC (Chen et al., 2011)

During development of an EMG driven IPMC based artificial muscle finger, a typical IPMC strip (Procured custom made from Environmental Robots Inc., USA) is used which has a thin (approximately 200 μm) perfluorinated ion exchange base polymer membrane (Nafion-117) with metal electrodes of platinum (5–10 μm) fused on either side. As a part of the manufacturing process, this base polymer is further chemically coated with metal ions that comprise the metallic composites. It responds in wet/dry condition. An IPMC is usually kept in a hydrated state to ensure proper dynamic operation. When the material is hydrated, the cations will diffuse toward an electrode on the material surface under an applied electric

and it allows holding an object for micro assembly operation.

Amplification setup

An EAP actuator based design for IPMC fingers have been discussed by Biddiss & Chau (2006). This shows the potential of electroactive polymeric sensors within an operating range of voltage (±3V) whereas Kottke et al. (2007) have reported on how to stimulate and activate a non-biological muscle such as an IPMC. Lee et al. (2006, 2007) have also demonstrated the potential of an IPMC actuating system with a bio-mimetic function using EMG signals. A mean absolute method is used for achieving the filtered EMG signal. Aravinthan et al. (2010) have designed a multiple axis prosthetic hand using IPMC. EMG signal through programmable interface controller (PIC) is sent to the IPMC prosthetic material to perform the required actions. By doing experiments, the potential of prosthetic hand using IPMC is shown. After that, we have also demonstrated actuation of IPMC through EMG via forearm muscles where potential of IPMC based micro robotic arm has been shown for lifting the object (Jain et al., 2010a; 2010b; 2011; 2012). For further application of EMG driven system, we are discussing detailed analysis of EMG signal control point view of IPMC based artificial muscle finger for micro gripper in this chapter.
