WEARABLE FALL DETECTION,

MONITORING AND ALERT SYSTEM FOR

THE ELDERLY

A project submitted as a part of the Intel IoT Internship conducted in

association with the Electronics and Communication (ECE) department
of NIT Trichy

By

Sudharshan Suresh
110113086

A student of
INSTRUMENTATION AND CONTROL ENGINEERING
NATIONAL INSTITUTE OF TECHNOLOGY
TIRUCHIRAPALLI- 620015

DECEMBER 2014
 ACKNOWLEDGEMENT

I am deeply grateful to Dr. B. Venkataramani, Ms. R. K. Kavitha

and Ms. R. Thilagavathy for their guidance, constant support and
assistance in project work. I would like to thank Intel for sponsoring the
program and enabling us to learn over the winter vacations. I would like
to thank the department of Electronics and Communication Engineering
of NIT Tiruchirapalli for providing us with components and a space to
work.

In addition, I would also like to thank our 3rd year mentors for helping us
during the initial stages of the internship.

Last but not the least; I give my sincere thanks to friends and family, who
have assisted me throughout the course of this month.

  2  
 ABSTRACT

Loss of consciousness and fall-induced injury is one of the leading problems faced by
the elderly. They are the leading cause for the transfer of senior citizens from the
comfort of their residence to assisted care environments. They also account for 40%
of injury related deaths and lack of timely action can prove fatal. On the personal side,
such sporadic events stifle the independence of the individuals, requiring a caretaker
to maintain constant surveillance.

My aim is to design a system to sense and alert doctors and nearest of kin of falls and
other related incidents of distress. It is also to provide the functionality of a
pedometer, to constantly monitor and relay the activities of the individual. The device
mustn’t be cumbersome, but rather wearable, so as to remove any social stigma
related to the possession of such a device. It must have a high degree of reliability and
differentiate activities of daily life from genuine falls. This would require a signal
plotting and thresholding-based approach to arrive upon a fall-detection algorithm.

With the dawn of the IoT (Internet of Things) era, it is ideal that such a device be
constantly relaying data via such a wireless network. This ensures seamless
connectivity between the patient, doctor and relatives – and swift response.

Keywords: Fall detection, elderly, monitoring, IoT

  3  
 TABLE OF CONTENTS
Title Page no.
INTRODUCTION AND PROJECT MOTIVATION
Status Quo………………………………………………………………………5
Falls and their impact…………………………………………………………...6
The detection system……………………………………………………………7

BASIC FLOWCHART………………………………………………………..9

COMPONENT DESCRIPTION……………………………………………...10

SOFTWARE AND LIBRARY PACKAGES REQUIRED…………………12

OBTAINING MEANINGFUL DATA FROM ACCELEROMETER……..13

FALL DETECTION ALGORITHM………………………………………....15

PEDOMETER FUNCTIONALITY…………………………………………..16

CIRCUIT DIAGRAM………………………………………………………….17

WORKING OF SYSTEM……………………………………………………...18

PROCESSING + TEMBOO + TWILIO………………………………………22

ARDUINO CODE……………………………………………………………….23

PROCESSING CODE…………………………………………………………..30

SAMPLE TEST DATA………………………………………………………… 36

REFERENCES…………………………………………………………………..37

  4  
 INTRODUCTION AND PROJECT
MOTIVATION
Status Quo

We live in a generation that is pressed for time and that, arguably, has a lack of
compassion. As a result of this negligence, the old and infirm are unfortunately
forgotten. The institutions of assisted care environments mushroomed as a result of
this alarming trend. Along with a rapid growth in aging population there is also a
decrease in the population of caretakers.

In India for example, there is a tremendous increase in the elderly population, with a
current estimate of 90 million over the age of 60. This is a result of a disproportionate
growth among age groups, with the 60+ population increasingly far more rapidly than
the 15-60 age group. The age dependence ratio of the country is now at a perilous
0.58. Studies show almost a quarter of such elders have poor health conditions, as
one could expect. This means that an approximate 22.5 million individuals are at risk
from occasional health complications.

To quote from the World Health Organizations ‘Aging in India’ report:

“The UN defines a country as ‘ageing’ where the proportion of people over 60

reaches 7 percent. By 2000 India will have exceeded that proportion (7.7%)and is
expected to reach 12.6% in 2025.”

This poses an alarming disadvantage to the elderly from a societal point of view. India
has a long and cherished culture of respect and devotion towards elders. However,
with the dawn of the nuclear family era, some communities find it hard to fully
commit to the senior citizens. This has heralded the growth of NGO’s and senior care
centers such as HelpAge India. But still, this begs the question- is it possible to attend
to our elders, regardless of the time? No, that is a far more pressing issue. It is indeed
time India took to caring for senior citizens 24/7 by embracing technology.

  5  
Falls and their impact

Falls are one of the most prominent causes of unintentional injury. They are not as
simple as what you may perceive them to be. They are coded as E880-E888 in
International Classification of Disease-9 (ICD-9), and as W00-W19 in ICD-10, which
include a wide range of falls including those on the same level, upper level, and other
unspecified falls. For the sake of it, we shall even define it according to WHO
literature-
“Inadvertently coming to rest on the ground, floor or other lower level, excluding
intentional change in position to rest in furniture, wall or other objects”.

Falls are responsible for 40% of all injury related deaths and need immediate medical
attention. The reason for a fall may be heart problems, loss of consciousness, fatigue,
exhaustion, diseases and loss of balance. Although falls occur through all age groups,
the major chunks of incidents are those of senior citizens. Approximately 28-35% of
people aged 65 and above fall every year. This figure only increases for the age group
of 70+ with 32-42%.

The major talking point is the need for swift medical aid. 50% of all injury-type
hospitalizations for the concerned age group are fall related. Thirty-two percent of
elderly people aged over 75 years have ever fallen at least once a year, and among
them, 24% have been seriously injured. Approximately 3% of all fallers lie
unattended for periods greater than 20 minutes, leading to insufficient medical action.
An upsetting figure is that 40% of nursing home admissions are directly linked with
incidents of falls.

This highlights two major points-

i. Falls require immediate medical attention- and the only way one can establish
that is by conceiving a sensing method.
ii. Falls greatly limit the independence of an individual. He or she will not have
the confidence to venture or stay alone for extended periods of time. This is
primarily due to the fear of medical complications surfacing.

  6  
The detection system
Now that the root issue has been established, we shall identify a method of sensing.
The major criteria one has in mind for detection of falls are -

i. Non-invasive
ii. Wearable system
iii. No false-positives
iv. No delay time
v. Common route of communication
vi. Interactivity

In the world of technology, wearable have always had a special appeal. This is due to
the ease of access and usage, as well as the seamless nature of the product. One does
not bear the brunt of social stigma by wearing such a device, thus satisfying the user.
Efficiency of the system is paramount, as the message must be delivered from point A
to point B in a matter of seconds. An established mode of communication must be
employed; in this case the Short Message Service (SMS) can reach people the
quickest. False alarms can be both extremely annoying and wastes both time and
money. So in order to ensure accuracy, the fall-detection algorithm must be perfected
and the device must be made interactive.

Figure  1:  Human  body's  center  of  

gravity

Taking all the above factors into consideration, I would propose the use of an
accelerometer-based sensing device around the waistline. This would not only mimic
the look of a commonplace belt, but is also the location of the center of gravity
(Figure 1). As a result, readings that one obtains from the sensor will be accurate and
trustable.

  7  
Communication can be made possible via wireless means to a central database. This
database is to be located at the doctor’s end, giving a constant stream of information
related to the patient. In case of any issues detected, a suitable alert is issued both at
the central display, as well as via SMS to the doctors and relatives. SMS is the most
rudimentary, ever-accessible means of communication in the modern world and
cannot be missed by any individual.

One uses the Internet of things to communicate from machine-to-machine. We can

employ various available modules such as Wifi-modules (eg: ESP 8266), Zigbee
protocol, Ethernet shields, etc. However for the purpose of prototyping, we employ a
Bluetooth serial transfer protocol between the microcontroller and the computer.

Wirel  

Wireless  methods     SMS    

  8  
 BASIC FLOWCHART

Relay  data  
Detect  change  in   Is  it  a   regarding  the  fall  
acceleration  along   fall  or   being  sensed  to  
an  ADL?     FALL  
with  inclination     the  central  
database  
 

ADL    
ADL  
No  problem    

ADL  

False   Yes    
Alarm    
Wait  for  response  
Response?     from   patient’s  side  
(20  secs)  
Distress    
confirmed.   No    
Help  is  on  
its  way  
 

  9  
 COMPONENT DESCRIPTION

I. Arduino Uno:
The Uno is a microcontroller board that is based on the ATmega328. It is widely
preferred for its ease-of-use, programming capabilities and large pool of online
documentation. It operates on very low current of 40-50 mA and has a logic level of 0
to 5V. It possesses 14 digital I/O pins ( 6 with PWM functionality). 6 analog inputs,
onboard LEDs, a reset switch and ceramic resonator. It can be supplied power through
USB connection or external power jack. It has a memory size of 32 KB,
programmable via Arduino IDE.

One could otherwise employ IoT ready boards such as the Texas Instruments
cc3200 or Arduino Yun or IoT implementable boards such as the Intel Galileo.

II. ADXL 335 (Accelerometer)

In order to sense angular position and acceleration we use the ADXL 335 MEMS-
based accelerometer. It is a 3 axis system which gives analog output corresponding to
the X, Y and Z axes of the orthogonal coordinate system.

  10  
The advantage of this sensor are twofold-
i. Extremely low noise
ii. Low power consumption of approximately just 320 micro Amperes.

The module has an entire sensing range of +/-3g.

III. HC-05 Bluetooth Module

For reasons of prototyping, we shall use this Bluetooth module to interface our device
with the computer.

The HC-05 is a class 2 bluetooth module, capable of easy connectivity with android
platforms or serial port communication with the computer. It serves as an efficient
replacement for wired solutions, offering mobility. It operates on 3.3V DC logic with
a current drawing capacity of 50mA.

IV. Piezo buzzer

V. Push Button Switch
VI. RGB 3 color LED

  11  
 SOFTWARE AND LIBRARY PACKAGES
REQUIRED
I. Arduino IDE
This is an open-source Arduino environment that enables us to type and
upload code to the board via serial connection. Available for Windows,
Mac OS X and Linux, this environment is the ubiquitous go-to software
for Arduino projects. The environment is structure using Java and based
off Processing, another software we shall be using.

II. Processing
 
Processing  is  an  open-­‐source  programming  language  and  IDE  that  is  aimed  
towards  computer  programming  in  a  visual  context.  Each  project  is  referred  to  
as  a  sketch  –  providing  computational  and  graphics  functions.  This  allows  us  to  
receive  instant  visual  feedback  to  any  process.  A  major  advantage  processing  
possesses  is  the  wide-­‐array  of  3rd  party  libraries  available  for  use.  Similar  to  
Arduino,  it  too  is  built  upon  the  Java  model.    
 
III. Temboo
Temboo  is  a  virtual  library  that  is  our  key  to  the  Internet  of  Things.  With  
streamlined  access  to  over  100+  APIs  across  various  programming  languages,  
this  intuitive  web-­‐based  coding  framework  helps  us  perform  machine-­‐to-­‐
machine  communication.    We  are  required  to  create  an  ID  for  their  service  and  
download  the  library  package  for  Processing.  Using  this,  one  can  transmit  text  
messages  to  cellular  device  via  an  online  server.      
 
IV. Twilio
Twilio  is  a  cloud-­‐based  communication  portal  that  allows  developers  to  make  
calls  and  send  SMS  programmatically.  The  web  service  API  can  be  accessed  used  
the  choreos  of  Temboo  and  billing  is  based  on  usage.  For  this,  we  can  use  a  trial  

  12  
account  on  Twilio  that  gives  us  limited  access  to  generating  text  messages  to  a  
pre-­‐determined  number.    
 
 

OBTAINING MEANINGFUL DATA

FROM THE ACCELEROMETER

Accelerometers measure the acceleration, most likely due to motion of a body.

However, when the accelerometer is fixed, only the gravity pulling down on it is
sensed. The device detects linear acceleration along three perpendicular axes. If one
was to sample the x, y and z axes data, they would get an accurate idea regarding the
orientation of the object.

The analog values that the microcontroller receives are the acceleration quantities
along each axis. However, one must first calibrate their device to ensure accuracy
keeping the device in a position such as the one shown in the figure. We know that
both X and Y-axis readings must be 0 as it is parallel to the ground while the Z-axis
reading must be the maximum. Adjust the correcting factor to obtain this.

In the above orientation, the values in g’s must be – (0g, 0g, 1g). Simple mathematical
manipulation tells us to divide the corrected analog value by the maximum value.

  13  
Thus, we have successfully converted the voltage values into g’s value.

The net acceleration is the root of the sum of the squares of all 3 values.

Anet = 𝑨𝒙𝟐 + 𝑨𝒚𝟐 + 𝑨𝒛𝟐

Net acceleration is an incredibly valuable quantity for fall detection. When a body
undergoes free-fall, the phenomenon of weightlessness will occur. This implies that
during a fall, the Anet will tend towards 0g. It is true an ordinary fall is different from
free-fall; but Anet will still be substantially lesser than 1g.

The final quantities of interest we obtain are those of orientation – pitch and roll.

These values give us an understanding of orientation of the user, which is vital when
it comes to detecting falls. The acceleration along each axis can be used, along with a
bit of trigonometry, to find out the pitch and roll angles.

!" !"
Pitch is 𝑡𝑎𝑛 !! ( ). Roll is 𝑡𝑎𝑛!! ( ).
!" ! !!" ! !" ! !!" !

  14  
 FALL DETECTION ALGORITHM
It was an extremely difficult process to generalize falls based on only two parameters.
However after many trial and error sessions and a whole lot of signal analysis, I
present to you my algorithm:

To safely conclude that a given action is a fall, you must eliminate the other activities
of daily life. These include lying down, sitting, jumping, running, squatting, hopping,
etc. Therein lies a major difficultly, because many of these actions give similar analog
readings as that of falling. Thus assigning simple thresholds will never do the job.
What I have done is create a series of checkpoints that must all be satisfied to
conclude that a certain event is a fall.

Sample  graph  showing  net  acceleration  on  fall

Checkpoint 1: As elucidated in the previous section, there will be a fall in the Anet
value when the fall commences. Thus if the value goes below a specific threshold (in
this case, 0.8g) it passes the first checkpoint.

Checkpoint 2: Following the fall, the user makes impact with the ground or surface.
A sharp spike in the Anet readings characterizes this. For our purposes, we consider

  15  
this as greater than 1g. The time lag between the checkpoints must be no greater than
2 seconds.

Checkpoint 3: The pitch and roll of the user is checked and compared with a
predefined threshold (In this case, greater than 60 degrees). The time lag between
checkpoint 2 and 3 must be very small – in the order of milliseconds.

Checkpoint 4: There is now a good chance that the user has fallen. We now check for
a period of 3 seconds whether the user gets up, by setting a threshold on inclination. If
3 seconds have elapsed, we can assume a fall has been sensed.

Checkpoint 5: We now wait for a period of 20 seconds for the user to disable the
alert, in case of false alarm. If 20 seconds have elapsed, we confirm the fall and
required action is taken.

PEDOMETER FUNCTIONALITY

The number of steps and distance travelled by an individual is an indicator of the

health of the person. To identify a step taken, we use a series of thresholds, similar to
that used in the fall detection.

The first criterion is that the user must be in ‘stable’ state. This means that he must be
upright and not in the process of falling.
The second criteria it checks for is a spike in the net acceleration as a result of foot
contact. Each time a step is taken, we detect this and increment the number of steps
taken.

  16  
 CIRCUIT DIAGRAM

  17  
 WORKING OF SYSTEM

Finished  device

Velcro  strap  and  main  frame  

  18  
We have highlighted the fall-detection algorithm that we shall use in the previous
section. This has been translated into code (as you will see in the code section) and
burned into the microcontroller. The other modules and components are integrated
with the MCU as you have seen in the circuit diagram. Here is the breakdown-

ADXL335: Connected to the analog pins 0, 1, 2 and 3 of the Arduino. They receive
the analog values of acceleration and necessary computation is performed on them.
The unit is powered with the +5V and the GND pins.

Accelerometer  in  my  device

The unit is position perpendicular with the ground and fixed onto the inside of the
wearable unit. It is calibrated in accordance to this position.

HC-05 Bluetooth Module: It is connected to the pins 10 and 11, which function as
software serial pins of the Arduino. One line transmits data to the module while the
other line receives data. The module is powered by 3.3V pin and the GND.

  19   to  frame
Bluetooth  module  attached  
To connect the Bluetooth module for serial communication, we must identify and
open both the incoming and outgoing ports. For this project I have opened the
outgoing port with TeraTerm while the Processing sketch handles data from the
incoming port.

RGB 3-color LED:

It consists of 4 pins, with one being a common cathode, while the other three power
the red, green and blue LEDs respectively. We connect digital pin 2, 4 and 5 to R, G
and B respectively. By using the digitalWrite functions we can change the color of
the LED based on the state. Thus it is green when the user is table, blue when fall is
sensed and red when fall is confirmed.

Piezo buzzer:
The buzzer is connected between the pin 9 and ground. We use the tone() function to
generate a pwm pulse of specified frequency at that pin. When a fall is first sensed,
we activate a 250Hz pulse and when confirmed, a 4000Hz pulse.

Piezo  buzzer  &  switch  circuit

Push button:
It is wired between pin 12 and ground, and pulled up by default. When it is pressed,
the pin is grounded- triggering an event. The event in this case is the false alarm
signal.

  20  
Implementation as wearable:

It can be used as a belt-type accessory tied around the waist with the help of the
Velcro strap. In the prototype, frontward falls may damage the structure, although a
compact, rugged casing can be used instead.

Belt-­‐like  wearable

 
Fall  testing

  21  
 PROCESSING + TEMBOO + TW  ILIO

As already explained in the previous sections, Temboo is a powerful portal to access

APIs. Using their web service, I was able to obtain the processing code snippet
required for integrating with a programmable virtual number on Twilio. The function
RunSendSMSChoreo() initiates the response with the text I have supplied (In this
case being a fall alert).  

Text  message  received

  22  
 ARDUINO CODE
The Arduino receives the data, performs calculations and outputs the data via
Bluetooth serial connection. In addition it controls the LED and piezo buzzer.

/*

Fall Detection, Alert and Action monitoring system

for the elderly-- An Intel IoT project by Sudharshan
Suresh
December 2014

The circuit:

ADXL335 (Accelerometer)
-----------------------
analog 0: accelerometer self test
analog 1: z-axis
analog 2: y-axis
analog 3: x-axis
GND: ground
+5V: vcc

HC-05 (Bluetooth Module)

------------------------
+3.3V: Vcc
GND: ground
Digital Pin 10(RX): TX
Digital Pin 11(TX): RX

Piezobuzzer
-----------
Digital Pin 9(PWM): +ve
GND: -ve

3-color LED
-----------
Digital Pin 2: RED
Digital Pin 3(GND): Ground
Digital Pin 4:GREEN
Digital Pin 5:BLUE

Push Button
-----------
Connected between Digital Pin 12 and GND

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Arduino Uno @ COM18---- via Arduino IDE

  23  
 Standard Serial over bluetooth link @ COM21
(Incoming)---- via Processing
Standard Serial over bluetooth link @ COM23
(Outgoing)---- via TeraTerm

*/

#include <SoftwareSerial.h>

// Accelerometer pin description

const int xpin = A3; // x-axis
const int ypin = A2; // y-axis
const int zpin = A1; // z-axis

float Xval, Yval, Zval; // acceleration in g's

along each axis
float Anet, pitch, roll; // Net acceleration ( 1
when stationary), pitch and roll convey orientation
of accelerometer

unsigned long time = 0; // variable to compute time

differences
unsigned long steps = 0; // number of steps taken
unsigned long steptime; // variable to compute time
between steps

boolean stepcheck = false;

boolean firstfoot = false;

int state = 0; // condition of subject ( 0

indicates stable and 5 indicates fallen )
int count =0;
int buttonrel = 0;
int button = 0;
int ALERT = 0; // fall, perhaps?
int blinkstate = 0;

//below are the calibration values for the 3 axes,

such that in stable condition the values are (1g,
0g, 0g)
float xZero = 505;
float yZero = 510;
float zZero = 525;

SoftwareSerial mySerial(10, 11); //RX, TX

void setup()
{
mySerial.begin(9600); // initialize the
software serial communication

  24  
 pinMode(12, INPUT_PULLUP); // Pull-up switch
connection
pinMode(9, OUTPUT); // PWM output for
piezobuzzer
pinMode(13, OUTPUT);
digitalWrite(13, LOW);

//RGB LED Pins

pinMode(2, OUTPUT); //Red
pinMode(3, OUTPUT); // Ground
pinMode(4, OUTPUT); //Green
pinMode(5, OUTPUT); //Blue

digitalWrite(2, LOW);
digitalWrite(3, LOW);
digitalWrite(4, HIGH);
digitalWrite(5, LOW);

steptime = millis();
}

void loop()
{
analogReference(EXTERNAL); // Enable Aref @ 3.3V

// Compute acceleration values in g's

Xval = (analogRead(xpin)- xZero)/102;
Yval = (analogRead(ypin) - yZero)/102;
Zval = (analogRead(zpin) - zZero)/102;
Anet = sqrt(sq(Xval) + sq(Yval) + sq(Zval));

// Pitch and roll caluculation

pitch = atan(Xval/sqrt(pow(Yval,2) +
pow(Zval,2)));
roll = atan(Yval/sqrt(pow(Xval,2) +
pow(Zval,2)));

//convert radians into degrees

pitch = pitch * (180.0/PI) - 90;
roll = roll * (180.0/PI);

// Serial transfer of required variables to

Processing
mySerial.print(pitch);
mySerial.print(",");
mySerial.print(roll);
mySerial.print(",");
mySerial.print(state);
mySerial.print(",");

  25  
 mySerial.println(float(steps));

pedometer(); // Function to sense steps taken by

wearer
distress(); // Manually operated distress-
signal can be sent via pushbutton hold
fall_detect(); // FALL DETECTION FUNCTION

delay(20); // Delay between passing values

}

void pedometer()
{
stepcheck = !stepcheck; // To sense every
alternate iteration

if((Anet>1.2)&&(stepcheck==true)&&(state==0)&&(first
foot==false)) // Sense initial spike in acc. reading
{
firstfoot = true;
steptime = millis();
}

if((Anet<0.9)&&(((millis()-
steptime)/1000)<2)&&(firstfoot == true)) // Sense
subsequent drop in acc. reading
{
steps++;
firstfoot = false;
}

else if((((millis()-
steptime)/1000)>2)&&(firstfoot == true)) //
maximum permissible time between above 2 conditions
firstfoot = false;
}

void distress()
{
if ((state==0)&&(digitalRead(12) == LOW))
// detects button PRESS
{
button++;
if(button>70)
// detects button HOLD
{

  26  
 tone(9,4000);
// Generate piezotone for distress
state = 42;
}
}

else if((state==0)&&(digitalRead(12) == HIGH))

// detects button release
{
buttonrel++;
if(buttonrel>10)
{
button = 0;
buttonrel = 0;
}
}
}

void fall_detect()
{
if(state ==0)
{
digitalWrite(2, LOW);
digitalWrite(5, LOW);
digitalWrite(4, HIGH);
}

if((Anet<0.8)&&(state==0)&&((abs(pitch)<60)&&(a
bs(roll)<60))) // Initial drop in
acc. due to free-fall like condition
{
state = 1;
time = millis();
}

if(state == 1)
{
if(Anet>1) // Increase
in acceleration due to impact of fall
{
state = 2;
time = millis();
}

if(((millis()-time)/1000)>2) // max.
permissible time between states 1 and 2
state = 0;
}

if(state == 2)

  27  
 {
if((abs(pitch)>60)||(abs(roll)>60)) //
Senses a fall in any orientation by factoring pitch
and roll
{
state = 3;
time = millis();
}
if((millis()-time)>100) // max.
permissible time between states 2 and 3
state = 0;
}

if(state == 3)
{ digitalWrite(4, LOW);
digitalWrite(5, HIGH );
if((abs(pitch)<60)&&(abs(roll)<60)) //
Detects getting-up and prevents false alarm
{
count++;
if(count>20)
{
state = 0;
count = 0;
}
}

if(((millis()-time)/1000)>3) // 3
second window after fall between states 3 and 4
{
state = 4;
ALERT = 1;
tone(9,250); // Fall, perhaps?
}
}

if((state ==4)&&(ALERT==1))
{
digitalWrite(5, (blinkstate) ? LOW: HIGH);
digitalWrite(2, (blinkstate) ? HIGH: LOW);
blinkstate = !blinkstate;

if(digitalRead(12) == LOW) //
Checks for button press to indicate false alarm
{
ALERT = 0;
state = 0;
}
if(((millis()-time)/1000)>20) // if
patient is unresponsive for more than 20 secs, fall
is confirmed-- necessary action taken via processing

  28  
 {
state = 5;
tone(9,4000); // Generate
piezotone for distress
digitalWrite(5, LOW);
}
}
}

  29  
 PROCESSING CODE
The processing sketch acts as the central monitoring unit for the patients. By
graphical means, we represent-

• Graphs of pitch and roll and the angle data for the same.
• Patient bio-data, time and date.
• Condition of patient
• Number of steps taken (Pedometer)

Processing  window

/*Processing sketch to process and interpret serial

data received via bluetooth module. Serves as the
central reference for doctors/ hospitals giving
alerts for possible falls, loss of consciousness and
also number of steps taken by patient. Also requires
active Internet service to send message via
3rd party server.
*/

// for serial data interpretation

import processing.serial.*;
Serial myPort;

  30  
 /*Temboo is a cloud-based platform with 2000+
processes for APIs, databases, and more. The
Processing + Temboo library enables us to connect
with TWILIO,
a cloud communication company that allows
developers to send/receive SMS and phone calls
programmatically. This solution is cost efficient,
effective and
is an alternate to a bulky, power hungry GSM
module. */

import com.temboo.core.*;
import com.temboo.Library.Twilio.SMSMessages.*;

int graphPara = 2; // Graph parameters -- pitch

and roll

float[] values = new float[graphPara+2];

//total of 4 serial variables
int[] min = new int[graphPara];
// minimum value to graph
int[] max = new int[graphPara];
// max value to graph
color[] valColor = new color[graphPara];
// graph line color

float effHeight; // effective height of

graph window

int xPos = 1; // horizontal position of

the graph
String status = "Stable";
int flag = 0;

// Create a session with your Temboo account details

TembooSession session = new TembooSession("suddhus",
"myFirstApp", "412da22525b64a10bd01b819434f15de");

void setup()
{
size(1260, 400); // Window dimensions
effHeight = 300 / graphPara;

// List all the available serial ports:

printArray(Serial.list());
// Port [0] is generally found to be the incoming
bluetooth serial line, if no USB COM is active:
myPort = new Serial(this, Serial.list()[0], 9600);
// don't generate a serialEvent() until you get a
newline character:

  31  
 myPort.bufferUntil('\n');

background(0);

// initialize pitch:
values[0] = 0;
min[0] = -90;
max[0] = 90;
valColor[0] = color(255, 0, 0); // red

// initialize roll:
values[1] = 0;
min[1] = -90;
max[1] = 90;
valColor[1] = color(0, 255, 0); // green

// state variable
values[2] = 0;

// steps variable
values[3] = 0;
}

// empty draw() loop

void draw()
{
}

// Everything occurs in the serialEvent() loop

void serialEvent(Serial myPort) // get the
ASCII string:
{
fill(0);
stroke(255);
rect(630,0,630,300);

textSize(40);
fill(255);
text("Patient Name: XYZ", 750,80);

String time = hour() + ":" + minute() + ":" +

second();
String date = day() + "/" + month() + "/" +
year();
text(time , 750, 130);
text(date, 750, 180);

textSize(45);

  32  
 fill(0);
rect(0,300,630,100);

//Check state of patient

if(status == "Stable")
fill(0,255,0);
else if(status == "FALL sensed")
fill(0,0,255);
else
fill(255,0,0);

//display appropriate state message

String condition = "Condition: " + status;
text(condition, 0, 360);

textSize(20);
text(int(values[2]), 1230, 20);

// Send SMS if fall Confirmed/ Distress

if((values[2] ==5)||(values[2]==42))
{
if(flag==0)
runSendSMSChoreo();
textSize(40);
fill(128, 0, 128);
text("SMS SENT", 750, 230);
flag = 1;
}

textSize(50);
fill(0);
rect(630,300,630,100);

//Display number of steps taken

fill(0,0,255);
String steps = "Pedometer : " + int(values[3]);
text(steps,640,360);

String inputString = myPort.readStringUntil('\n');

noFill();
stroke(255);
textSize(12);
rect(0,0,630,300);

if (inputString != null)
{
// trim off any whitespace:
inputString = trim(inputString);

  33  
 // splits the serial string on the delimiters
and assigns it to a float array, here the delimiter
can be a ,--[comma] , --[space] , \t--[tab]
values = float(splitTokens(inputString, ",
\t"));

//check if you recieved the whole packet and map

the data
if (values.length >= graphPara)
{
for (int i=0; i<graphPara ; i++)
{
fill(50);
noStroke();
rect(0, effHeight*i+1, 70, 12);
fill(255);
text((values[i]), 2, effHeight*i+10);

// map to the range of effective screen

height:
float mappedVal = map(values[i], min[i],
max[i], 0, effHeight);

// draw lines:
stroke(valColor[i]);
line(xPos, effHeight*(i+1)- mappedVal + 2,
xPos, effHeight*(i+1) - mappedVal-2);

// draw dividing line:

stroke(255);
line(0, effHeight, 630,effHeight);

//dotted line (origin):

stroke(64,64,64);
for( int j=0, k=0;j<=610;j=j+20)
{
if((k%2)==0)
{
line(j,75,j+20,75);
line(j,225,j+20,225);
}
k++;
}

// reset graph if at the edge:

if (xPos >= 630)
{
xPos = 0;

  34  
 // erase graph window with translucent black
fill(0,200);
rect(0,0,630,300);
}

else
xPos++; // increment graph position

}
}

print(values[2]);
if(values[2]==5)
status = "FALL confirmed";

else if(values[2] ==4)

status = "FALL sensed";

else if(values[2] == 0)
status = "Stable";

else if(values[2] == 42)

status = "Distress";
}

// Function to generate SMS

void runSendSMSChoreo()
{
// Create the Choreo object using your Temboo
session
SendSMS sendSMSChoreo = new SendSMS(session);

// Set credential
sendSMSChoreo.setCredential("AlertSMS");

// Set inputs

// Run the Choreo and store the results

SendSMSResultSet sendSMSResults =
sendSMSChoreo.run();

// Print results
println(sendSMSResults.getResponse());

  35  
 SAMPLE TEST DATA

Activity No. of tests Fall detected Not detected

Front fall 5 5 0
Backward fall 5 5 0
Side right fall 5 5 0
Side left fall 5 5 0
Sitting 5 0 5
Jumping 5 0 5
Lying down 5 1 4

𝟑𝟒
Accuracy = ×𝟏𝟎𝟎 = 𝟗𝟕. 𝟏𝟓%
𝟑𝟓

  36  
 REFERENCES

http://www.who.int/ageing/publications/Falls_prevention7March.pdf

http://www.futurehealthsystems.org/publications/heterogeneity-in-self-assessed-health-
status-among-the-elder.html

http://www.helpageindia.org/about-us.php

http://www.gjmedph.org/uploads/r2-vo2no4.pdf

http://bildr.org/2011/04/sensing-orientation-with-the-adxl335-arduino/

http://physics.rutgers.edu/~aatish/teach/srr/workshop3.pdf

http://www.arduino.cc/

https://temboo.com/

https://www.twilio.com/

  37