This is a simple EMG Controller for a bionic hand prosthesis

Dependencies:   mbed-dsp mbed NOKIA_5110

Galileo Bionic Hand - ST Nucleo Example


This is an example of a simple hybrid sEMG(surface Electromyography) activated controller for the Galileo Bionic Hand Prosthesis implemented with the Galileo EMG Shield. The user has to select the desired posture by sending a special character through serial port (115200 baud rate) and then perform it through sEMG activation by detecting contraction on flexor muscles of the forearm. Contractions on forearm extensor muscles releases the posture and allows the return to the default or rest posture.

Special characters:

  • '1', for "Power Grip" selection
  • '2', for "Point" selection
  • '3', for "Pinch Grip" selection
  • '4', for "Lateral Grip" selection
  • '5' for "Tripod Grip" selection


Galileo EMG Shield - 3 Channels for surface EMG

The electrodes for sEMG are placed on the skin surface and are connected to the input of a precision instrumentation differential amplifier based on Texas Instruments (TI) INA122, then its output is passed through an active low pass filter (LPF) based on TI OPA335 in order to sense the action bio-potentials of the muscular fibers with an output signal span in the range of 0 to 3.3V and a bandwidth between 0 to 500 Hz. The circuit is built-in on a custom PCB with 3 sEMG channels and a bias voltage reference output (1.25 V); furthermore, it is pin compatible with Arduino pin compatible ARM Cortex M4 boards and mbed platform ecosystem (ST Nucleo F411RE, Freescale FRDM K64F, NXP LPCXpresso 4337, ST Discovery F746NG, Renesas GR-Peach, etc), which is ideal for the single ended input of a microcontroller in addition to contributing to low cost development kits. /media/uploads/julioefajardo/shieldbrd.png /media/uploads/julioefajardo/image1.jpg

Electrodes Placement and Connection

Standard surface mounted Ag/AgCl electrodes with wet conductive gels are placed on palmaris longus and extensor digitorum muscles (third channel could be placed on Flexor carpi Ulnaris), focusing only on below elbow disarticulation. These electrodes have been well-characterized and most of its properties are well understood, except for some properties as drifting and low-frequency noise. Nevertheless, with proper preparation of the skin, the sEMG signal is excellent.

Disposable electrodes and snap leads information:

Proper placement of disposable electrodes for two channels of surface EMG is shown below:

/media/uploads/julioefajardo/electrodes.png /media/uploads/julioefajardo/mucles.png

Customizable Postures

You can customize the actions by modifying PWM values (microseconds) on FingerPosition function (values depends on the way that the hand was built it). The prosthesis has five fingers and a thumb rotation mechanism and five actuators in order to perform multiple types of grasping. Wrist rotation will be implement later.

The servo motors have to be connected as shown below:


Function Declaration and Usage Examples

include the mbed library with this snippet

void FingerPosition(float32_t thumb_us, float32_t index_us, float32_t middle_us, float32_t pinky_us, float32_t thumbrot_us);

FingerPosition(2400, 600, 600,2400,2400);   //Close
FingerPosition(2400,2400, 600,2400,2400);   //Point
FingerPosition(2400, 600,2400, 600,2400);   //Pinch
FingerPosition(2400, 600, 600,2400, 600);   //Lateral
FingerPosition(2400, 600, 600, 600,2400);   //Tripod
FingerPosition(1000,2400,2400, 600, 600);   //Open

Serial Oscilloscope Feature

This feature easily allows to watch and log to file the data using serial oscilloscope software (115200 baud rate).

  • Serial_Osciloscope(TRUE,RAW) to watch raw signals, FALSE deactivate this feature
  • Serial_Osciloscope(TRUE,RECTIFIED) to watch rectified signals, FALSE deactivate this feature
  • Serial_Osciloscope(TRUE,SMOOTH) to watch smooth signals, FALSE deactivate this feature


Universal Real-Time Software Oscilloscope GUI information:

Nolia 5110 LCD

Nokia 5110 display implementation for visual feedback will be add later, we have to modify libraries and fonts in order to improve the functionality. The main idea is to change of action by pressing a push button and change thresholds using a potentiometer.



Videos, bill of materials and tutorials to build the Galileo EMG Shield and the Galileo Bionic Hand will be posted soon, more information on:



File content as of revision 1:9472990d8bb4:

#include "mbed.h"
#include "arm_math.h" 

Ticker EMG_Sampler;
DigitalOut myled(LED1);
AnalogIn   Ref(A0);
AnalogIn   E1(A1);
AnalogIn   E2(A2);
AnalogIn   E3(A3);
float32_t EMG1, EMG2, EMG3;
float32_t samples[100];
float32_t samples2[100];
float32_t samples3[100];
float32_t abs_output[100];
float32_t abs_output2[100];
float32_t abs_output3[100];
float32_t samplesfi[128]; 
float32_t samplesfi2[128];
float32_t samplesfi3[128];
float32_t mean = 0.0f, mean2 = 0.0f, mean3 = 0.0f;
uint8_t COCO = 0;

void ADC_Sampler() {
    EMG1 = (*3.3f;
    EMG2 = (*3.3f;
    EMG3 = (*3.3f;
    uint32_t m = __get_PRIMASK();
    for(int j=99;j>0;j--) {
        samples[j]=samples[j-1];                    //Fill Array
        samples2[j]=samples2[j-1];                  //Fill Array
        samples3[j]=samples3[j-1];                  //Fill Array
    //copy 25 samples on array
    for(int j=99;j>0;j--){ 
    COCO = 1;

float32_t h[29] = {
int main() {
  pc.baud(115200);                                  //Serial com at 115200 bauds
  EMG_Sampler.attach(&ADC_Sampler, 0.001);          //1ms ticker for ADC Sampler
  myled = 1;
  //Thumb = 0;
  while(1) { 
        //mean for threshold
        arm_abs_f32(samples, abs_output, 100);       //rectifier
        arm_abs_f32(samples2, abs_output2, 100);     //rectifier
        arm_abs_f32(samples3, abs_output3, 100);     //rectifier
        arm_conv_f32(abs_output,100,h,29,samplesfi);//low pass filter
        arm_conv_f32(abs_output2,100,h,29,samplesfi2);//low pass filter
        arm_conv_f32(abs_output3,100,h,29,samplesfi3);//low pass filter
        arm_mean_f32(samplesfi, 128, &mean);
        arm_mean_f32(samplesfi2, 128, &mean2);
        arm_mean_f32(samplesfi3, 128, &mean3); 
        if ((mean>0.009f)&&(mean2>0.007f)&&(mean3>0.007f)) {
            myled = 0;
        else {
            myled = 1;
        COCO = 0;