EEG Module and Arm Prosthesis V2

For this project we'll need to do a lot of things, since it's very complicated. It's the second version of my proshtetic hand. The first one had a lot of problems and right now it doens't even work! This time I've made everything more thoughtfully and even made a 3D model of the hand by myself. The prosthetic arm is controlled by your brainwaves! Isn't it cool, right? EEG reads you alpha braiwaves (but you can also read beta brainwaves and analize them on your own) and then an algorithm processes the signal and rotates servos. Right now, I'm still waiting for gears, so there would be updates in the future. But the main thing works, so you are free to start making this project. All videos of things working are in the "Results" step.

We need a lot of stuff. To make it simplier, I devided parts into sections, so, it'll be easier for you to navigate.

EEG:

Electrodes (I used Gold Cup electrodes), AD620ANZ, 2xTL084CN, resistors (560, 2x22k, 2x12, 2x270k, 220k, 47k, 2x180k, 3x100k, 220), 1k variable resistor, non polarized capacitors (5x220nF, 100nF, 10nF, 25nF), polarized capasitors (2x10uF, 1uF), 2xbattaries (9V).

Prosthetic arm:

LiPo battaries (11.1V, 2800mA*h), ESP32, 5xServos (MS-1.3-9), LM2594N, 1N5820, 63uH 3A inductor, 1uF non polarized capasitor, polarized capasitors (180uF 35V, 470uF 35V), gears (exact sizes will be a bit later).

Here's the circuit of the most important part - electroencephalography. I'll explain what's going on here.

We need alpha brainwaves, which have frequency from 8 to 13 Hz. I also made it possible to read beta brainwaves (12-30 Hz), since my EEG filters out the noise which is not between 7 and 31 Hz. Firstly, the signal is amplified 89 times. Then it goes to 60 Hz notch filter, which filters out the biggest noise due to power lines interference. The next one is 7Hz high pass filter. It filters out the noise made by skin. Then comes 31 low pass filter. It just gets rid of anything above our frequency range. Then there's 1 Hz high pass fiter just for extra attenuation of unwanted noise and resistor in parallel with capasitor, which provide extra filtering (~160 Hz low pass filter). Then comes another amplification, but this time with variable resistor. It gives 83-455 gain. The last filter is another 60 Hz notch filter to filter out the noise which is left.

Now about electrode placement. You have 3 electrodes (+, -, and GND). GND electrode is needed to be placed at A1 spot, which is at earlobe. Electrodes + and - should be placed at Fp2 and O2 areas (there shouldn't be any difference between placing + electrode at Fp2 and - O2 or otherwise).

Now a bit simplier circuit. Let's start with powering.

I used step-down converter to get 5V from 11.1V to not burn down our components. This step-down converter produces enough current to power both ESP32 and all 5 Servos.

Then we connect OUTPUT from EEG and Servos to ESP32 as shown on the picture.

The code is also simple. It just finds average value of the signal and then compares it to constant value. If it's bigger than constant value, then servos start moving.

Here I'll explain why I used 30:1 gearbox specifically. The length of one finger is ~0.08052m. My servo has 0.127 Nm torque. We're aiming for ~30N grip strength, so we need to calculate our torque for the finger. Torque = F*l = 30*0.08052 = 2.4156Nm. To find gear ratio I assumed some losses, so our efficiency coeficent would be 85%. Gear ratio = Torque/(torque of the servo * efficiency coeficent) = 2.4156/(0.127*0.85) = ~22.4. I round this number up, so it would be definetly enough. That's how you get 30:1 gear ratio.

I'm still finishing design of the hand, so I'm just showing how it supposed to look like.

Here you can see that EEG work perfectly fine. I'll share with other results as soon as possible.