*4.2.2. EMG based control method with targeted muscle reinnervation*

Targeted muscle reinnervation (TMR) is a surgical procedure which is developed to increase the number of psychologically appropriate control inputs available for use with a prosthetic device [42]. A surgical procedure is used to transfer the motor neurons of the residual limb to a remaining set of muscles. After denerverting, the target muscles motor nerves can be reinnervated. These reinnervated muscles are capable of providing EMG to a prosthesis controller with more accurate motion intention. For a transhumeral amputee, the goal is to transfer the median and distal radial nerves to target muscle segments creating hand open and close signals. For the shoulder/humeral neck disarticulated amputees all four brachial plexus nerves are transferred to the target muscles [42]. It should be noted that for cases with left shoulder disarticulation the interference of electrocardiogram (ECG) causes an effect to the TMR EMG signals [42]. This could be eliminated using a second-order high pass filter with 60Hz cutoff frequency, which results in 80% elimination of the ECG signals and 20% elimina‐ tion of the EMG signals producing an improved law signal-to-noise ratio output.

been done using Linear Discriminant Analysis (LDA) and it provides nine features for the collected EMG data set. Then both k-Nearest Neighbor (kNN) and Library Support Vector Machine (LIBSVM) have been used for the classification. Use of majority voting technique has

> **Elbow extension**

Redial N. to P.head Major, inf. Stemal

> Native remnant triceps

Native remnant triceps/post. Deltoid

Native Triceps, long and medial head

\*Unexpected result: subject indicates opening hand with a movement of thumb abduction, a median nerve function

**Table 2.** Summary of TMR subjects including level of amputation, surgical sites used as inputs and any other prosthetic

Manus hand shown in Fig. 19 as a replacement to a lost hand for an amputee is capable of generating 3DoF, reproducing the grasping function for the user [23]. The hand is controlled via EMG signal of the residual muscle of the user. The simultaneous operation of the joints has been realized. It is achieved using a differently approached pattern recognition technique [23]. In this technique, single muscle EMG signal is used to generate the grasping commands. In order to do this a command language comprising of three EMG bits has been developed. Each

long head Native Triceps

long head Native Triceps

Median N. to upper stemal head

Median N.to P.Major, stemal head and P.Minor

Ulnar N.to remnant P.Major stemal head segment

Median N. to Biceps, short head

Median N. to Biceps, short head

Median N. to Biceps, short head

**Hand close Hand open Wrist rotation**

Recent Trends in EMG-Based Control Methods for Assistive Robots

Toggle between Hand Open/Close control using socket-mounted FSR

http://dx.doi.org/10.5772/56174

259

Two-site control with two socket mounted FSRs

Single-site control with one socket mounted FSR

Single-site control with harness linear potentiometer

Single-site control with harness linear potentiometer

Single-site control with harness linear potentiometer

Median N. to upper stemal head\*

Distal Radial N. to Thoracodorsal N (Serratus Anterior)

> Radial N to P.Minor

Distal Radial N. to Brachialis

Distal Radial N. to Brachialis

Distal Radial N. to Brachialis, lateral head

resulted in providing smoothed classification accuracies.

**Elbow flexion**

Musculocutaneo us N. to P.Major, Clav.Head

Musculocutaneo us N. to P.Major, Clav.Head

Musculocutaneo us N. to P.Major, Clav.Head

long head

\*\*Humeral neck level amputation clinically fit as a shoulder disarticulation level amputee

**Level of amputation**

Shoulder Disartic.\*\*

TH-A Trans-humeral Native Biceps,

TH-B Trans-humeral Native Biceps,

TH-C Trans-humeral Native Biceps,

controls implements in the post-TMR prosthetic fitting [42]

*4.2.4. EMG based control method of manus hand*

SD-A Shoulder Disartic.

SD-C Shoulder Disartic.

**Subject identifier**

SD-B

Boston Digital arm [43], Otto Bock electric wrist rotator [44] and an electric terminal device (hook or hand, depending on subject preference) are fitted to six TMR subjects, three trans‐ humeral (TH) amputees, two shoulder disarticulation (SD) amputees and one humeral neck transhumeral amputee. After attaching the patient with the prosthetic device the patient should be guided and continuously monitored to achieve better outcome from the TMR. Even though the training process of the prosthetic user with TMR EMG control is much more easily compared to the pre-TMR, EMG controlling since the TMR itself provides an initiative to the user for training. The control algorithm of the system is comprised of proportional controlling of velocity according to the amplitude of the EMG signal. Table 2 shows summary of infor‐ mation of TMR subjects in the experiment [42].

Even though TMR is capable of providing more sites to extract EMG signals, it is a surgical process and involves invasive procedures, which may cause user discomfort. Therefore, TMR should be replaced by a hybrid control mechanism developed using other signals such as EMG of residual muscles and the EEG signals.

## *4.2.3. EMG based control of prosthetic finger*

Rami *et al*[16] have proposed an improved control algorithm for a control of a prosthetic finger. Using the developed controller different finger postures of the prosthetic hand can be con‐ trolled. The main EMG pattern recognition setup is shown in Fig. 18. The controller is mainly developed considering the main ten motions of the hand; flexion of each finger, pinching of thumb combined with each and every finger and clenching of the fist..

Data was collected using two EMG channels. For the feature extraction the windowing was done using a disjoint windowing scheme, which consumes less computer resources due to its simplicity. Various feature sets have been extracted; Waveform length, Hjoth Time Domain (TD) Parameters, Slope Sign Changes, Number of Zero Crossings, Sample Skewness and Auto Regressive (AR) Model parameters were selected [46]. The two feature sets from the two channels resulted in a large one feature set. Dimensionality reduction of the feature set has been done using Linear Discriminant Analysis (LDA) and it provides nine features for the collected EMG data set. Then both k-Nearest Neighbor (kNN) and Library Support Vector Machine (LIBSVM) have been used for the classification. Use of majority voting technique has resulted in providing smoothed classification accuracies.


\*Unexpected result: subject indicates opening hand with a movement of thumb abduction, a median nerve function

\*\*Humeral neck level amputation clinically fit as a shoulder disarticulation level amputee

**Table 2.** Summary of TMR subjects including level of amputation, surgical sites used as inputs and any other prosthetic controls implements in the post-TMR prosthetic fitting [42]

#### *4.2.4. EMG based control method of manus hand*

*4.2.2. EMG based control method with targeted muscle reinnervation*

258 Electrodiagnosis in New Frontiers of Clinical Research

mation of TMR subjects in the experiment [42].

of residual muscles and the EEG signals.

*4.2.3. EMG based control of prosthetic finger*

Targeted muscle reinnervation (TMR) is a surgical procedure which is developed to increase the number of psychologically appropriate control inputs available for use with a prosthetic device [42]. A surgical procedure is used to transfer the motor neurons of the residual limb to a remaining set of muscles. After denerverting, the target muscles motor nerves can be reinnervated. These reinnervated muscles are capable of providing EMG to a prosthesis controller with more accurate motion intention. For a transhumeral amputee, the goal is to transfer the median and distal radial nerves to target muscle segments creating hand open and close signals. For the shoulder/humeral neck disarticulated amputees all four brachial plexus nerves are transferred to the target muscles [42]. It should be noted that for cases with left shoulder disarticulation the interference of electrocardiogram (ECG) causes an effect to the TMR EMG signals [42]. This could be eliminated using a second-order high pass filter with 60Hz cutoff frequency, which results in 80% elimination of the ECG signals and 20% elimina‐

tion of the EMG signals producing an improved law signal-to-noise ratio output.

Boston Digital arm [43], Otto Bock electric wrist rotator [44] and an electric terminal device (hook or hand, depending on subject preference) are fitted to six TMR subjects, three trans‐ humeral (TH) amputees, two shoulder disarticulation (SD) amputees and one humeral neck transhumeral amputee. After attaching the patient with the prosthetic device the patient should be guided and continuously monitored to achieve better outcome from the TMR. Even though the training process of the prosthetic user with TMR EMG control is much more easily compared to the pre-TMR, EMG controlling since the TMR itself provides an initiative to the user for training. The control algorithm of the system is comprised of proportional controlling of velocity according to the amplitude of the EMG signal. Table 2 shows summary of infor‐

Even though TMR is capable of providing more sites to extract EMG signals, it is a surgical process and involves invasive procedures, which may cause user discomfort. Therefore, TMR should be replaced by a hybrid control mechanism developed using other signals such as EMG

Rami *et al*[16] have proposed an improved control algorithm for a control of a prosthetic finger. Using the developed controller different finger postures of the prosthetic hand can be con‐ trolled. The main EMG pattern recognition setup is shown in Fig. 18. The controller is mainly developed considering the main ten motions of the hand; flexion of each finger, pinching of

Data was collected using two EMG channels. For the feature extraction the windowing was done using a disjoint windowing scheme, which consumes less computer resources due to its simplicity. Various feature sets have been extracted; Waveform length, Hjoth Time Domain (TD) Parameters, Slope Sign Changes, Number of Zero Crossings, Sample Skewness and Auto Regressive (AR) Model parameters were selected [46]. The two feature sets from the two channels resulted in a large one feature set. Dimensionality reduction of the feature set has

thumb combined with each and every finger and clenching of the fist..

Manus hand shown in Fig. 19 as a replacement to a lost hand for an amputee is capable of generating 3DoF, reproducing the grasping function for the user [23]. The hand is controlled via EMG signal of the residual muscle of the user. The simultaneous operation of the joints has been realized. It is achieved using a differently approached pattern recognition technique [23]. In this technique, single muscle EMG signal is used to generate the grasping commands. In order to do this a command language comprising of three EMG bits has been developed. Each bit has defined three digital levels. Accordingly, an input comprises of three muscle contrac‐ tions to generate three EMG levels and 27 different commands. From the practical point of view, 18 out of 27 commands were selected for the implementation. However, no information relevant to the pattern recognition has been published by the authors. Table 3 shows the main commands of the 3-bit command language extracted from [45]. According to the controller mechanism no simultaneous operation could be realized using the MANUS hand. Therefore instead of the 3-bit command language it would be better to use a proper pattern recognition method with a higher number of EMG inputs. This could also be coupled with an EEG or other biological signal, in order to achieve higher effectiveness from the controller.

**Functional command 3-bit pattern Remarks**

Close, gripping mode "1", Preset 0 to 250 gr. Total

Close, gripping mode "1", Preset 251 to 500 gr. Total

Close, gripping mode "2", Preset 0 to 250 gr. Total

Close, gripping mode "2", Preset 251 to 500 gr. Total

Close, gripping mode "3", Preset 0 to 300 gr. Total

Close, gripping mode "1", Preset total pressure.500gr., or

Available 111 Available 112

*4.2.5. EMG based ANN controller for a transhumeral prosthesis*

that of the EMG data and average joint angles were extracted.

controls implements in the post-TMR prosthetic fitting [45]

Close, gripping mode "1", Preset total pressure.500gr., or

pressure, or until "Stop"

pressure, or until "Stop"

pressure, or until "Stop"

pressure, or until "Stop"

pressure, or until "Stop"

until "Stop"

until "Stop"

Stop 100 Constant, pre-set, compulsory Default position 200 Constant, pre-set, compulsory Calibration 212 Constant, pre-set, compulsory Rotate to right, until "Default" or until "Stop" 210 Individually adapted, recommended

Rotate to left, until "Default" or until "Stop" 101 Individually adapted, recommended

**Table 3.** Summary of TMR subjects including level of amputation, surgical sites used as inputs and any other prosthetic

The EMG based Artificial Neural Network (ANN) controller was developed to realize elbow flexion/extension and forearm supination/pronation based on the artificial neural network (ANN) techniques using EMG signals as the input. Fig. 20 shows the basic concept of the design of the controller. EMG signals are fed into the controller from seven muscles: brachialis, biceps, medial head of triceps, posterior deltoid, anterior deltoid, clavicular and pectoralis major. Raw EMG signals from the muscles are amplified, alternating current coupled, low-pass filtered and recorded. Recorded EMG signals are processed offline. They are filtered, windowed and features extracted. Several features are extracted from 320 samples, 128ms sample time rectangular windows with 50 percent overlap between adjacent segments which provides a sample time of 64ms. The features, mean absolute value, waveform length, number of zero crossing and number of slope sign changes are extracted from each window, which generate a four-element feature set for each EMG channel. Simultaneously motion capturing data from the Optotrak Certus motion capture system is resampled such that its sample time matched

Close, gripping mode "4", Preset: arch, or until "Stop" 202 Individually adapted

"preset"

Recent Trends in EMG-Based Control Methods for Assistive Robots

http://dx.doi.org/10.5772/56174

261

211 Individually adapted

221 Individually adapted

222 Individually adapted

120 Individually adapted

121 Individually adapted

122 Individually adapted

201 Individually adapted

"preset"

**Figure 18.** Block diagram of experimental evaluation of the EMG-pattern recognition system [46]

**Figure 19.** MANUS hand [23]


**Table 3.** Summary of TMR subjects including level of amputation, surgical sites used as inputs and any other prosthetic controls implements in the post-TMR prosthetic fitting [45]
