from the exercise.

Okay, so as you have downloaded your ones, so we will check them after I run my code,

we will control your code. And so we have exercise two, and this is the exercise.

All right, so we need to talk to me. This is the practical was just to make you a little bit acquainted with spike training, as you've seen also in the industry lecture.

So the data that we acquire, the goal is to take this, this, you know, to a physiological signal with many action potential, and then to derive the structures from the signal that are these are usually,

most of the time is done by the convolving these electrophysiological thing at the unique spike train from individual nanos.

So taking the actual potential from these high density breeds of electrons and converting converting this signal into a binary code.

So series of zeros and one, where zeros and one, they basically respond to the all of non response of the nanos.

So, so the first task was just for was just, you know, to convert once you once this data has been decomposed.

And we can look at this in another lecture, but basically this was done with a blind source separation procedure. There are different procedures there.

Also, for example, now we are trying to use AI to do this, but, but blind source separation or spike sorting algorithms, they can identify this unique action potential.

So then now I just already gave to you some data that was decomposed, and this was done by blind source separation, and you had these time instance of firing of each nanos, and you had to calculate which the actual, you know, to find the actual potential to see the binary spike train.

And, and then do some classic correlation analysis. So let's start with the for task one, which was to convert in one and zeros.

So, I will firstly go through the code that we prepared and then we go to the one that you have uploaded on the website. So let's load the practical, and then we do so basically, as you can see here we have an index with zeros, and then we run this code.

To be honest, I cannot read it.

Yeah, you cannot read it.

Because I have a very big screen. Can you make it.

I mean the left part of the window with the current folder and stuff is pretty big and also the workspace.

Let me see if I do this.

Can, does this improve anything?

Maybe if you just increase the font size.

Yeah, can we do it. How do you do it in Matlab.

So, so so so so so I probably have to increase it for all right.

So, plot references.

Fonts.

Yeah, it's a bit too big.

Okay.

Okay, does it work now. Yes.

So let's dock it.

Okay, so we load it.

And then here you can see, it looks so weird.

So the firing so basically we just have a metrics that we initialize with zeros, that is equal to the size of the motor unit pulses, the rows are equal to the size of the motor unit pulses, and then the length is equal to our reference signal.

And because the reference signal in this much this can be also, this can also be the EMG signal.

So you basically it has the whole length of the chunk of data.

And it's just to speed it up, because Matlab works better when when you initialize.

And then you have, you have basically a for loop that goes from one to the size of motor pulses, and then you get the times of individual motor units, and the times are basically the times are basically the firing of this unit.

And then for each motor unit, you take this time, and you add a one, and this creates you basically a matrix, where you have, for example on row one, motor unit one, you have the old zeros and one for that motor unit.

Okay, so for that neuron.

And one is one, the neuron is firing. Okay, and you can see it here.

So if we plot, this is how, if we plot, this is plotted just one, and this is plotting all of them, and you can see this that sometimes you this has four, and this is because some neurons are perfectly synchronized.

So the firing of the neurons are very similar. So this is something, so maybe I will ask questions to you.

So, you know, if you have one signal, and you have two neurons that are perfectly synchronized, how can the composition algorithm or any, you know, any processing that you want to do.

Why do you think we are able to identify two neurons that fire completely synchronously, because even if you have, what is, why we're able to do this.

Anyone wants to take this.

So, the reason is because, and this is also related to the industry lecture that you've seen that they're using so many electrons now in the brain, so many. So this is because by having many, here you have 64 channels, by having many, you have a special representation.

So if these two neurons, one is here, one is here, and they fire together, they still have special listing. So you can recognize them. But this is a problem that you know back in the day they had extensively, because they were using only one source.

And if you have only one source, and you have two neurons firing together, these are merged, you know, and you will never be able to remove this superimposition of the action potential.

So by having high dimensional grids of electrons, we improve that.

And you can see that this happens consistently that you know two neurons fired together. And this is quite common for the spinal cord, because the spinal cord has a very strong, the motor neurons in the spinal cord are highly synchronized.

So very high synchronized activity. And this is because, you know, it's functionally important, because you have one muscle, and the muscle will be weird, you know, if you would activate it in a different way, you would have a, you could have tremor, for example,