hello this is version here and this is my uh intel iot project on a wearable fall detection monitoring and alert system for the elderly now this is generally one on the belt and i've made it so as to detect a fall if any uh that an individual undergoes and if such a fall under if such a fall is detected it immediately sends the message over to the doctor and also sends an sms to the nearest king that is the relatives and the doctor that the patient has fallen in addition to this it also has a pedometer that measures the number of steps a person takes because the number of steps a person takes indicates the health of the person directly so if you look at the components i've used an arduino uno microcontroller here along with that an adxl 355 accelerometer which does the bulk of the processing that is it takes most of the inputs from the motion and it actually extrapolates that to give the pitch and the roll as well as the acceleration on all three axis it also has a battery so as to give wireless power along with a hc-05 bluetooth module because it is a prototype you have used a bluetooth module to transfer the data serially through a virtual bluetooth serial port to the computer it also consists of a piezo buzzer and a switch and a three color led just to show the state of the person i've made it wearable as a belt using a velcro strap i've used here and now we should look at how it works let's start a bit more about how we use acceleration to detect all this i shall now present you two important pieces of data that you extract from the accelerometer the first one is the net acceleration that is the root of the sum of the squares of the acceleration along all three axis that is the x y and z axis which can be uh taken and processed with your arduino i ide and the second one there is the picture of the roll the pitch and the roll is uh what gives you your orientation that is if your patient is lying down or if your patient has fallen down and we find this about two axis using simple trigonometry for the pitch and the roll and it's the confluence of the combination of these two factors that in fact give you the patient's stability or to indicate or to get a better algorithm of all detection algorithms such now i have written a certain algorithm by myself to detect a fall and i found this accurate about 97 and the algorithm goes as such the first thing is that you find the net acceleration of the patient and when a person undergoes a fall theoretically he should be in zero g state that is there should be a very big drop in acceleration since this isn't a an ideal theoretical free fall we shall give it a threshold as say it goes below 0.8 g so that would be your first threshold this is your first checkpoint your second checkpoint is after that fall he must make an impact to the ground upon making impact the ground there is a sharp spike in the acceleration value and here we've used the threshold using my trial and error to be greater than 1g to 1.1 g and we check this within a span of say 2 seconds the next is that after the person has fallen down you need to note his orientation and in this case we should uh not involve those cases where he's sitting down or involve those cases where he's slightly tilted that he should legitimately have fallen down and for this i've taken a threshold to say greater than 60 degrees and the time lag between this checkpoint and the previous one must be very very minimal which uh actually this eradicates all your false signals given due to jumps are given due to just uh small stammers while you walk as such and all these generate signals here false positives that we can eradicate using these checkpoints the next one is that when the user has fallen you have to wait for a period of say three seconds because again there's a window of a chance that there's a false alarm there and after these three seconds we can safely assume that the person has fallen and we can indicate that the fall has been sensed now to properly establish this and to confirm we wait for 20 seconds and in 20 seconds if the person does not deactivate the fall detection through this manual switch over here then help must be called for and that is when the fall detection comes into play this is the arduino sketch that i've written for this have used software serial to transmit the data to the bluetooth module and these are the computations that one must do in order to convert the raw data of the accelerometer values into the useful net acceleration as well as the pitch on the road this is the function that we use to detect a step which is rather quite elaborate thus reducing any error that you would face and this is the function that detects the distress that is if a person has fallen through a series of states as you can see here this is the processing sketch that takes serial data values from the arduino and displays it graphically it also has the integration with uh the internet that is it has the tempo in the twilio integration that is used to access the api to send the sms to the phone and it also graphically displays all this through text and with the graph as well for running the program this is the window that is displayed on the top left is the graph showing the pitch and role of the individual on the top right is the bio data of the patient at the bottom left is the condition of the patient that is either stable or in the fall detected yet and finally is the parameter which displays the number of steps taken by him next we shall see how this reacts when someone falls now we shall look at the device in action i've strapped it around as a belt over here as you can see and now the condition in the monitor here says that i'm stable it says the number of steps i've taken and it shows that shows you the bio data of the patient this is generally what we would keep in the doctor's office and we must have a stable wi-fi connection in order to send an sms using this so here's me taking a few steps that are indicated and now we will check it detect the fall now it is blue and now after three seconds it indicates that the fall has been sensed and it waits for a period of 20 seconds in this period this is to confirm whether the person has undergone a legitimate fall now it is confirmed and it turns red where the displaces fall confirmed and it says there's a message been sent this sms will be received on the mobile phone as i will show you shortly the device also doesn't generate any false alarms for example jumps are not detected and lying down is not sensed as a fall so as a result the accuracy that i have determined is about 95 to 97 which is pretty good for a prototype you