**3.3. Model for controlling the EMG signal for IPMC based artificial muscle finger**

For acquiring the data from said muscles, EMG electrodes are placed for measuring the electric potential produced by voluntary contraction of muscle fiber on the human finger. The frequency range of the EMG signal is within 4 to 900 Hz. The dominant energy is concentrated in the range of 95 Hz and amplitude of voltage range is ±1.2 mV according to muscle contraction and the voltage function *Vin(t)* in term of signal sample time (*t*) is given below,

$$V\_{in}(t) = V\_{ino} \sin(2\pi ft) \tag{1}$$

Where, *Vino* is amplitude of EMG voltage (±0.0012 V); *f* is frequency of EMG signal (95 Hz).

In Laplace domain, EMG input signal is written as

Computational Intelligence in Electromyography Analysis – 370 A Perspective on Current Applications and Future Challenges

field. Inside the polymer structure, anions are interconnected as clusters providing channels for the cations to flow towards the electrode (Chen et al., 2011). This motion of ions causes the structure to bend toward the anode as shown in Fig. 2. An applied electric field affects the cation distribution within the membrane, forcing the cations to migrate towards the cathode. This change in the cation distribution produces two thin layers, one near the anode and another near the cathode boundaries. The potential is generated by changing the potential electric field on cluster of ionic strips that provides the actuation of the strip.

To examine the bio-mimetic behavior of IPMC based artificial muscle finger, it is important to study the physiological structure of human finger. An internal structure of human index finger is shown in Fig. 3. The index finger is actuated by three intrinsic muscles and four extrinsic muscles. The intrinsic muscles consist of two interosseous (IO 1 and IO 2) muscles & one lumbrical (LU) muscle and four extrinsic muscles connected through long tendons i.e. extensor digitorum communis (EDC), extensor indicis proprius (EIP), flexor digitorum superficialis (FDS) and flexor digitorum profundus (FDP) (Bundhoo & Park, 2005). For heavy lifting & holding purpose, EDC and EIP are responsible in tendon network. Consequently, EMG electrodes are placed at these two positions on the human finger so that we can achieve direct actuation of IPMC based artificial muscle finger through said muscles.

**3.2. Basic tendon of index finger for identification of EMG signal** 

**Figure 3.** Basic tendons of the index finger (Bundhoo & Park, 2005)

**finger** 

**3.3. Model for controlling the EMG signal for IPMC based artificial muscle** 

For acquiring the data from said muscles, EMG electrodes are placed for measuring the electric potential produced by voluntary contraction of muscle fiber on the human finger. The frequency range of the EMG signal is within 4 to 900 Hz. The dominant energy is concentrated in the range of 95 Hz and amplitude of voltage range is ±1.2 mV according to muscle contraction and the voltage function *Vin(t)* in term of signal sample time (*t*) is given below,

$$V\_{in}(s) = \frac{2.96}{s^2 + 3.51 \text{e5}} \tag{2}$$

Using these parameters, the circuit for filtered EMG signal is designed using MATLAB SIMULINK software as shown in Fig. 4.

**Figure 4.** Block diagram of EMG signal behaviour from human index finger

In block diagram, the active EMG signal is taken from index finger muscle and uniform noise is considered. The electric potential is first amplified with gain 32 dB and then band pass filter (BPF) is used within specified frequency range (4 to 900 Hz). Using two band stop filters (BSF and BSF1), noise signal (60 Hz) that arises due to AC coupled power is eliminated. The signal is then passed through an amplifier with gain 60 dB. Subsequently, three integrators (Integrator, Integrator1 and Integrator2) are used for achieving better damped signal. The output of EMG signal with sampling time of 10-4 seconds after filtering is shown in Fig. 5.

**Figure 5.** Acquired data from finger muscles via EMG signal

Thus, the total output duration of EMG signal for sampled data is 0.1 second. The general solution of acquired EMG voltage (*VEMG*) through curve fitting method is obtained as given below,

$$V\_{EMG}(t) = \sum V\_0 \sin(2\pi f\_0 t + \delta\_0), \quad 0 \le t \le 0.1 \tag{3}$$

Design and Control of an EMG Driven IPMC Based Artificial Muscle Finger 373

It is found that the zero-poles have real value on both sides of the real axis which does not meet the Nyquist stability criteria. Also from Fig. 6, gain cross-over frequency (GCF=2.53 rad/s) is greater than phase cross-over frequency (PCF=1.18 rad/s), indicating that this

For achieving the stable EMG signal, different configurations of PID are analysed through SISO control tool of MATLAB software. By applying PD control with proportional control gain factor (*Kp=1*) and derivative control gain factor (*Kd=1*), EMG voltage *VEMGfinal 1(s)* is

1 5 2 -9 -4 2

**Figure 7.** Root-locus and bode plot behaviour of human finger through PD controller

In case of PI controller, the unit proportional control gain factor (*Kp=1*) and integrator control gain factor (*Ki=1*) parameters are used. After applying this control system, EMG voltage

**x 105**

**Phase (deg)**

**-300 -250 -200 -150 -100 -50 0 50**

**M** 

**agnitude (dB)**

P.M.: 0.62 deg Freq: 0.0108 rad/sec

G.M.: -0.389 dB Freq: 0.0108 rad/sec Unstable loop

2 2 -9 -4 2

The bode plot and root locus for this system are plotted as shown in Fig. 8. From this figure, it shows that the data from EMG signal is again unstable but the response has a better


**10-2 <sup>100</sup> <sup>102</sup> <sup>104</sup> <sup>106</sup> <sup>0</sup>**

**Frequency (rad/s)**

 **Bode Plot Editor**

3.92e s (s +1.17e ) (s +3.51e ) ( ) (s+0.724)(s + 9.91e s + 1.17e )(s -0.724s+1.90) *V s EMGfinal* (7)

The root locus and bode plot are shown in Fig. 7. This indicates that the obtained data from EMG signal is unstable but through this control system the data is converging towards

2 -4 2 5

(s +1.17e )(s +3.516e ) ( ) (s+2.546e )(s - 9.91e s+1.17e ) (s +1.38s + 1.38) *V s EMGfinal* (6)

voltage obtained from EMG signal is unstable.

=

**Root Locus Editor** 

obtained as given below,

stability from initial condition.

*VEMGfinal 2(s)* is obtained as given below,

**-600**

**-400**

**-200**

**Im ag Axis**

**0**

**200**

**400**

**600**

=

**-4 -3 -2 -1 0 1 2**

**Real Axis**

prospect of converging towards stability than previous configuration.

where, *V0* is the amplitude of EMG voltage of sine function (±0.0012 V); *f0* is average frequency of sine function (4.7 Hz); δ*<sup>0</sup>* is initial phase difference of sine function (1.66 rad); *t* is signal sample time in second.

After adjustment of root mean square (RMS) value of this sine function, the EMG voltage *VEMGrms(s)* after filtering is written as

$$V\_{EMG\_{mm}}\text{(s)} = \frac{1.16\,\text{e}^{-5}\,\text{(s}^2 + 1.17\,\text{e}^{-4})}{\left(\text{s}^2 - 0.014\,\text{s} + 1.07\,\text{e}^{-4}\right)\left(\text{s}^2 + 0.014\,\text{s} + 1.07\,\text{e}^{-4}\right)}\tag{4}$$

For controlling purpose, single input single output (SISO) control system tool in MATLAB is used and the output *VEMG(s)* data is obtained. The initial overall transfer function of EMG voltage *VEMGinitial (s)* is obtained through output signal from (4) to input signal from (2) and is given as

$$V\_{EMGinitial} \text{(s)} = \frac{3.92 \text{e}^{-6} \left(\text{s}^2 + 1.17 \text{e}^{-4}\right) \left(\text{s}^2 + 3.5 \text{e}^5\right)}{\left(\text{s}^2 - 0.0146 \text{s} + 1.07 \text{e}^{-4}\right) \left(\text{s}^2 + 0.0146 \text{s} + 1.07 \text{e}^{-4}\right)} \tag{5}$$

After that, Nyquist criterion is applied to check the stability of EMG signal. The root-locus and bode scheme are plotted as shown in Fig. 6.

**Figure 6.** Root-locus and bode plot behaviour of EMG signal via finger in initial condition (Jain et al., 2011)

It is found that the zero-poles have real value on both sides of the real axis which does not meet the Nyquist stability criteria. Also from Fig. 6, gain cross-over frequency (GCF=2.53 rad/s) is greater than phase cross-over frequency (PCF=1.18 rad/s), indicating that this voltage obtained from EMG signal is unstable.

Computational Intelligence in Electromyography Analysis – 372 A Perspective on Current Applications and Future Challenges

frequency of sine function (4.7 Hz);

*VEMGrms(s)* after filtering is written as

is signal sample time in second.

below,

given as

2011)

**-600**

**-400**

**-200**

**0**

**Imag Axis**

**200**

**400**

**600**

Thus, the total output duration of EMG signal for sampled data is 0.1 second. The general solution of acquired EMG voltage (*VEMG*) through curve fitting method is obtained as given

where, *V0* is the amplitude of EMG voltage of sine function (±0.0012 V); *f0* is average

After adjustment of root mean square (RMS) value of this sine function, the EMG voltage

For controlling purpose, single input single output (SISO) control system tool in MATLAB is used and the output *VEMG(s)* data is obtained. The initial overall transfer function of EMG voltage *VEMGinitial (s)* is obtained through output signal from (4) to input signal from (2) and is

 δ

( ) ( )


*<sup>s</sup> V s* (4)

( )( )

**10-2 <sup>100</sup> <sup>0</sup>**

**Frequency (rad/s)**

**Bode Plot Editor** 


( ) ( )

s 0.0146s 1.07e s 0.0146s 1.07e *V s EMGinitial* (5)

2 -4 2 -4 3.92e s 1.17e s +3.5e

−+ ++

+

After that, Nyquist criterion is applied to check the stability of EMG signal. The root-locus

**Figure 6.** Root-locus and bode plot behaviour of EMG signal via finger in initial condition (Jain et al.,

**x 10-3**

**45 90 135 180 <sup>225</sup>** P.M.: 0 deg Freq: 1.18 rad/sec

**-100**

**-50**

G.M.: 13.3 dB Freq: 2.53 rad/sec Unstable loop

**0**

**Magnitude (dB)**

**50**

**100**

**Phase (deg)**

− + +

0 00 ( ) (2 ), *V t V sin f t EMG* 0 0.1 ≤ ≤*t* (3)

*<sup>0</sup>* is initial phase difference of sine function (1.66 rad); *t*

= + π

<sup>+</sup> <sup>=</sup>

1.16e ( 1.17e ) ( ) s 0.014s 1.07e s + 0.014s 1.07e *rms EMG*

δ

=

**Root Locus Editor** 

**-5 0 5**

**Real Axis**

( )

and bode scheme are plotted as shown in Fig. 6.

For achieving the stable EMG signal, different configurations of PID are analysed through SISO control tool of MATLAB software. By applying PD control with proportional control gain factor (*Kp=1*) and derivative control gain factor (*Kd=1*), EMG voltage *VEMGfinal 1(s)* is obtained as given below,

$$V\_{EMGfinal1}(\mathbf{s}) = \frac{(\mathbf{s}^2 + 1.17\mathbf{e}^4)(\mathbf{s}^2 + 3.516\mathbf{e}^5)}{(\mathbf{s} + 2.546\mathbf{e}^5)(\mathbf{s}^2 \cdot 9.91\mathbf{e}^{-9}\mathbf{s} + 1.17\mathbf{e}^{-4})\left(\mathbf{s}^2 + 1.38\mathbf{s} + 1.38\mathbf{})}\tag{6}$$

The root locus and bode plot are shown in Fig. 7. This indicates that the obtained data from EMG signal is unstable but through this control system the data is converging towards stability from initial condition.

**Figure 7.** Root-locus and bode plot behaviour of human finger through PD controller

In case of PI controller, the unit proportional control gain factor (*Kp=1*) and integrator control gain factor (*Ki=1*) parameters are used. After applying this control system, EMG voltage *VEMGfinal 2(s)* is obtained as given below,

$$V\_{EMGfinal2}(\mathbf{s}) = \frac{3.92 \mathbf{e}^{-6} \text{ s} \left(\text{s}^2 + 1.17 \mathbf{e}^{-4}\right) \left(\text{s}^2 + 3.51 \mathbf{e}^5\right)}{\left(\text{s} \star 0.724\right) \left(\text{s}^2 + 9.91 \mathbf{e}^{-9} \text{s} + 1.17 \mathbf{e}^{-4}\right) \left(\text{s}^2 \cdot 0.724 \mathbf{s} + 1.90\right)}\tag{7}$$

The bode plot and root locus for this system are plotted as shown in Fig. 8. From this figure, it shows that the data from EMG signal is again unstable but the response has a better prospect of converging towards stability than previous configuration.

#### Computational Intelligence in Electromyography Analysis – 374 A Perspective on Current Applications and Future Challenges

Design and Control of an EMG Driven IPMC Based Artificial Muscle Finger 375

 **Bode Plot Editor** 

**10-5 <sup>100</sup> <sup>105</sup> <sup>1010</sup> -360**

**Frequency (rad/s)**

control system is that it is stable with least amount of noise. This signal is sent to artificial

**Magnitude (dB)**

**Figure 10.** Root-locus and bode plot behaviour of EMG signal via finger through PID controller (Jain et

**Phase (deg)**

**x 105**

**-300 -250 -200 -150 -100 -50 0 50**

**-270 -180 -90 0 90** G.M.: -13 dB Freq: 1 rad/sec Stable loop

P.M.: 36.6 deg Freq: 1.34 rad/sec

In order to examine the bio-mimetic behaviour of IPMC based artificial muscle finger through EMG signal, EMG electrodes (Ag/Agcl based) are positioned at EDC and EIP muscles on the index finger. The input EMG signal from the muscle movement varies in the range of ±1.2mV (which is observed through oscilloscope). But activation of IPMC based artificial muscle finger needs the voltage of ±3V and current rating of 50-200 mA which is only possible by amplification of voltage and current. For desired output voltage to the artificial muscle finger, EMG signal is transferred through analog-digital convertor (ADC) card and PXI system (PXI-1031 along with NI-6289) in real time environment. EMG signal through electrodes are sent to an input channel at specified ports of the ADC card. Then this signal is amplified through amplification factor of 2550 using express VI of Labview 8.5 and sent to DAC output. But DAC output signal cannot provide enough current (50-200 mA) to drive an IPMC based artificial muscle finger. For achieving this current rating, the current amplification is done using customized IPMC control circuit by combining operational amplifier (Model: LM-324), transistor (Model: TIP 122) and resistances (1kΩ and 10Ω). Noise interference is eliminated by enabling low-pass filtering with PID control system to achieve the stability during operation of the artificial muscle finger as shown in Fig. 11. The IPMC

Now, in order to prevent the abrupt physio-chemical change of the IPMC nature and subsequent shortening of the actuation operating time of the IPMC material due to irreversible electrolysis (caused when the voltage applied across the two faces of an IPMC exceeds a maximum limit), two warning flags are used. One warning flag is placed at input EMG signal and another warning flag is placed at output voltage of DAC card where IPMC control circuit is connected. This limited voltage imposed on the IPMC based artificial muscle finger aborts

muscle finger for holding the object.

**4. Experimental testing setup** 

**-600**

**-400**

**-200**

**0**

**Imag Axis**

**200**

**400**

**600**

**-3 -2.5 -2 -1.5 -1 -0.5 0 0.5**

**Real Axis**

**Root Locus Editor** 

size 40mm × 10 mm × 0.2 mm is used for testing purpose.

al., 2011)

**Figure 8.** Root-locus and bode plot behaviour of EMG signal via finger through PI controller

After that, PID controller is used where proportional control gain factor (*Kp*= 0.5), integrator control gain factor (*Ki*= 1) and derivative control gain factor (*Kd*=1) are given in compensator to attain the stability of EMG voltage. For applying the PID control system, the SIMULINK block diagram is modified as shown in Fig. 9.

**Figure 9.** Block diagram of EMG signal via human finger after applying PID control system

Thereafter, the model is again simulated in MATLAB software, upon which, the data from finger muscles shows the all zero-poles in left hand side of real axis which satisfies the Nyquist criteria shown in Fig. 10. The final EMG voltage *VEMGfinal (s)* is acquired as given below,

$$V\_{EMGfinal}(\mathbf{s}) = \frac{\mathbf{s}\left(\mathbf{s}^2 + 1.17\mathbf{e}^{-4}\right)\left(\mathbf{s}^2 + 3.51\mathbf{e}^5\right)}{\left(\mathbf{s} + 2.54\mathbf{e}^5\right)\left(\mathbf{s} + 1.52\right)\left(\mathbf{s}^2 + 1.83\mathbf{e}^{-8}\mathbf{s} + 1.17\mathbf{e}^{-4}\right)\left(\mathbf{s}^2 - 0.14\mathbf{s} + 0.90\right)}\tag{8}$$

Also, GCF (1rad/s) is less than PCF (1.34 rad/s). Hence, a stable EMG voltage data is achieved. This filtered EMG signal is stable enough to provide necessary voltage signal across the IPMC for proper functioning during operation. The major advantage of this control system is that it is stable with least amount of noise. This signal is sent to artificial muscle finger for holding the object.

**Figure 10.** Root-locus and bode plot behaviour of EMG signal via finger through PID controller (Jain et al., 2011)

#### **4. Experimental testing setup**

Computational Intelligence in Electromyography Analysis – 374 A Perspective on Current Applications and Future Challenges

**Root Locus Editor** 

block diagram is modified as shown in Fig. 9.

**-0.8 -0.6 -0.4 -0.2 <sup>0</sup> 0.2 0.4 0.6 -600**

**Real Axis**

**-400**

**-200**

**0**

**Imag Axis**

**200**

**400**

**600**

=

( )

**Figure 8.** Root-locus and bode plot behaviour of EMG signal via finger through PI controller

**Figure 9.** Block diagram of EMG signal via human finger after applying PID control system

criteria shown in Fig. 10. The final EMG voltage *VEMGfinal (s)* is acquired as given below,

Thereafter, the model is again simulated in MATLAB software, upon which, the data from finger muscles shows the all zero-poles in left hand side of real axis which satisfies the Nyquist

+

Also, GCF (1rad/s) is less than PCF (1.34 rad/s). Hence, a stable EMG voltage data is achieved. This filtered EMG signal is stable enough to provide necessary voltage signal across the IPMC for proper functioning during operation. The major advantage of this

( )( ) ( )( )( ) ( )

2 -4 2 5

+ + + + −+

s 2.54e s 1.52 s 1.83e s 1.17e s 0.14s 0.90 *V s EMGfinal* (8)

5 2 -8 -4 2 s s 1.17e s +3.51e

After that, PID controller is used where proportional control gain factor (*Kp*= 0.5), integrator control gain factor (*Ki*= 1) and derivative control gain factor (*Kd*=1) are given in compensator to attain the stability of EMG voltage. For applying the PID control system, the SIMULINK

**10-2**

P.M.: -39.8 deg Freq: 1.56 rad/sec

G.M.: 73.3 dB Freq: 79.3 rad/sec Unstable loop

**-300 -250 -200 -150 -100 -50 0 50**

**M agnitude (dB)**

**Phase (deg)**

**100**

**<sup>104</sup> <sup>0</sup>**

 **Bode Plot Editor** 

**Frequency (rad/s)**

**102**

In order to examine the bio-mimetic behaviour of IPMC based artificial muscle finger through EMG signal, EMG electrodes (Ag/Agcl based) are positioned at EDC and EIP muscles on the index finger. The input EMG signal from the muscle movement varies in the range of ±1.2mV (which is observed through oscilloscope). But activation of IPMC based artificial muscle finger needs the voltage of ±3V and current rating of 50-200 mA which is only possible by amplification of voltage and current. For desired output voltage to the artificial muscle finger, EMG signal is transferred through analog-digital convertor (ADC) card and PXI system (PXI-1031 along with NI-6289) in real time environment. EMG signal through electrodes are sent to an input channel at specified ports of the ADC card. Then this signal is amplified through amplification factor of 2550 using express VI of Labview 8.5 and sent to DAC output. But DAC output signal cannot provide enough current (50-200 mA) to drive an IPMC based artificial muscle finger. For achieving this current rating, the current amplification is done using customized IPMC control circuit by combining operational amplifier (Model: LM-324), transistor (Model: TIP 122) and resistances (1kΩ and 10Ω). Noise interference is eliminated by enabling low-pass filtering with PID control system to achieve the stability during operation of the artificial muscle finger as shown in Fig. 11. The IPMC size 40mm × 10 mm × 0.2 mm is used for testing purpose.

Now, in order to prevent the abrupt physio-chemical change of the IPMC nature and subsequent shortening of the actuation operating time of the IPMC material due to irreversible electrolysis (caused when the voltage applied across the two faces of an IPMC exceeds a maximum limit), two warning flags are used. One warning flag is placed at input EMG signal and another warning flag is placed at output voltage of DAC card where IPMC control circuit is connected. This limited voltage imposed on the IPMC based artificial muscle finger aborts the execution of the program when the warning flag has a high output. The flow chart for actuation of IPMC based artificial muscle finger is shown in Fig. 12. During operation, artificial muscle finger bends in a similar manner as that of the index finger. For generating force, this finger is held in cantilever configuration on the fabricated work bench. A load cell is used to collect the data at different angles of the index finger. The current and voltage analysis of the human muscles are also done through oscilloscope. Thus an IPMC artificial muscle finger based micro gripper driven by EMG is developed and the holding behaviour is demonstrated.

Design and Control of an EMG Driven IPMC Based Artificial Muscle Finger 377

Show warning

Abort execution

Show warning

Abort execution

No

No

**Figure 12.** Flow chart of for actuation IPMC based artificial muscle finger using EMG signal

End

Activate IPMC based artificial muscle finger in micro gripper to hold the object

Feed amplified data in another DAQ Assistant inside the LABVIEW

Add steps in process flow diagram for amplification of voltage (within ±3V) along with acquired input signal from EMG

Start

Take EMG signal from human finger via EMG electrodes an input port of ADC using DAQ and NI-PXI system

Develop process flow diagram using LabVIEW to acquire input signal from EMG with ±1.2mV

> Is input range within ±1V

Yes

Is output range within ±10V

Send amplified signal to DAC

Connect to IPMC control circuit

Yes

**Figure 11.** Basic testing layout for actuation of IPMC based artificial muscle finger
