Design of Robotic Hand - EMG control

Hi there,

I am currently researching the necessary requirements for the development of a “human-controlled” robotic hand. The two notable methods are EEG and EMG. (I don’t want the control to be mimicked by attaching flex sensors to a glove). My goal is to achieve single finger movement. My first impression is that EEG is going to be far too difficult to reproduce meaningful signals from an EEG headset, despite the numerous electrodes used and filters. Thus I have decided to pursue EMG control.

I have seen a number of commercially available development boards such as the Muscle Sensor v3(discontinued although can be sourced) and the upgrade of that being the Myoware sensor. I assume there’s no possible way to control each finger individually using one Myoware sensor as the range of feedback values from the Myoware sensor won’t be large enough to accurately distinguish between the different finger movements?? (Due to the complexity of the muscle groups in the forearm) Would it be a better option to use at least two Myoware sensors(to have additional measurements) on two parts of the forearm, and then compare consecutive values between sensor outputs to distinguish the different finger movements?

Any input will be greatly appreciated. Thanks in advance. :slight_smile:

Yes. Good idea. Do that.

When you have an Arduino question that we can help you with, we'll be ready.

How can only two sensors be sufficient for getting signals to the many muscles in a hand?

Ideally, I wouldn't use only two but many more, that was a suggested approach in order to make signals distinguishable due to the numerous muscle groups in the forearm as well as the fact that they are closely grouped. Also, the sensors are rather pricey in my country thus I am trying to gain individual finger control while minimizing cost.

Putting further thought into this, it may be a better option to have a common ground electrode and then have multiple electrode channels for the different muscle groups in the arm - which avoids using the myoware sensors and then requires the development of boards to process signals and so on.

Thanks for the replies! Really appreciate the input.

If ever, I'd try a 3D image of the hand, from which all movements can be determined. Two cheap cameras may be sufficient for making 3D images, plus image processing software for tracking the position of (visible or virtual) markers on the hand. This task requires much computing power, can not be accomplished with a small Arduino. One or more Arduinos; depending on the number of muscles; can be used to make the robotic hand perform the observed moves.