 Hi, I am Raj. These are my payments. And we are going to, what we are doing is, we are getting the bank glucose level. I mean, bank glucose level. And we get the data. And we convert it into the PC, which produces a PCM signal, right? That too, sorry, waveform 2, the waveform 2 PCM signal. We draw a graph with that signal, okay? The basic idea behind this application is, we have peripheral device which has some sensor, actually it has a sensor, in which the sensor electrode has ferrocinate and our blood will pass into the first layer, that is, it will be given as a counter electron and it will pass to the next electron, so that the two conducting plates will be there, so that it will produce some electrons and the flow of electrons. We will get some of the PWM, that is, pulse width modulated output. We will get some pulse width modulated output from this peripheral device. Absolutely we miss to bring the peripheral device. Anyway, our good luck, we bring samples of the peripheral device and we go for a scratch that we need to optimize the contagion area because we are giving the PWM, that is, pulse width modulated signal in the audio into the mobile phone. We have the audio jack. So, when it enters into the audio jack, it will be as analog input and we need to convert again into the raw signal and we need to show it as a graph and to store the value in the SD card or somewhere else, we need to encode it into the PCM signal. So, in this type of conversion, we have got various quantizing. First thing is, whenever we are converting a signal for ADC or DAC or something else, what other conversion takes place? Sure, there will be some data loss. The same thing happens here. While sampling, we need to quantize and while quantizing we have some quantizing error. So, while coming here and we have gone for the scratch and absolutely our entire idea is to build the app, but when we come to this hack night, our good luck, we have built a optimizing quantizing algorithm here. That's the thing. Quantizing error algorithm has been proposed in this hack night and we are happy to say that 20% age of error we have reduced here. In such case, we have taken a sampling PWM signal. What I have told you before, I will bring the PWM signal. In the PWM signal, the major part rolls with the mid-values, that is 0.5. And we said the PWM signal, the algorithm generation credits will go to my friend and he will explain more about it. Well, as he told, we had to convert a PWM which was received as an analog signal into a PCM for storing in SD card and then back to PWM for plotting in future. See, one analog digital will have a lot of loss of data, but two will hardly get about 50% of the data. So I came, I was like, okay, we can use something like as simple as Nyquist sampling ratio, but then as we searched on it, it is kind of difficult because sampling in amplitude is not as easy as sampling in time. So then I thought, okay, why should we search what someone has already done? So we sat together, all of us, and then we brainstormed and we came up with an algorithm and this has got really good results. So once the computer is loaded, I'll just show it to you. Actually, we have more than 10 lakhs of samples in PWM signal what we sampled. And we compared this which first we have done this PWM signal sample by the MATLAB so that, yeah, this is a story behind us creating this algorithm. First of all, we created the samples. What we brought is, we have put it in the MATLAB and we have tried out it and we got the white points and the NS and the Nyquist ratio what you have been given as the sampling rate of 44,000. Then after some things, we have done some modulation technologies and inbuilt that we showed the algorithms that what we have derived now step by step. In first step, we have divided it by 1000 and we have rounded off the values and again we have done some of the police things and again we got the wrong outputs then again we go through the other ways and we found some of the rounding values and at last we came to the MATLAB is working fine when our team came for a video capture we got the output in the MATLAB then we went into the Java using the dandroid we got struck up in the modulation that is mod mod of some values are given there and we got struck up in that and we don't know the function that is IEEE function. Is it actually easier to use a digital device or a digital device? See the thing is like you told already PWM to PCM converters there are a lot of ICs which do this but like I told you there is a lot of data loss I mean any A2D device however ideal it is will have a data loss because you are dropping the potential between two converters so it's a passive element there's going to be a data loss and we actually went through this see the basic idea of this is whenever there's a conversion the maximum error is always in the midpoint so 1 minute I'll just show you two things one is the actual data set but that is not really good to get a feel of then I'll show you something this is the signal like you told this is the signal you get from the peripheral device like you see there are so many samples you can hardly make out anything from it and this is the reconstructed signal which you can see is almost same I mean you might say that how will I know whether it's same or not but you can see here the mean squared error if I just look the normalized value is going to be something like 7.62 in 10 power minus 4 but based on an algorithm now we are able to achieve something like 6.22 into 10 power minus 4 for this again these are just numbers you can only believe it if you see it normalized so I'll just show you a little bit of plot which you can appreciate just once again so this is a sine wave from 0 to 5 as you can see it now I will show you the normalized sine wave which is going to be something like this which is the quantized sine wave which is aegakutum minar for you and now after our data sampling this is what we get from the picture you can see that this anyone can see and you can tell this is more like a sine wave than a kutum minar so this was the result of our work from about 4 to 5 hours of continuous brainstorming and we hope to implement this algorithm for my name future thanks thank you