Fall Detection System for Elders

Among the elderly population, falls are one of the most common causes of death and injury. More than 30% of people over 65 years old fall each year and the prevalence increases for people above 80 years old. Even a minor fall can severely affect the physical and mental health of an elder due to the

Project Title

Fall Detection System for Elders

Project Area of Specialization

Artificial Intelligence

Project Summary

Among the elderly population, falls are one of the most common causes of death and injury. More than 30% of people over 65 years old fall each year and the prevalence increases for people above 80 years old. Even a minor fall can severely affect the physical and mental health of an elder due to the fear of falling again. 

Thus, the caregiving process and the quality of life of older adults can be improved by adopting systems for the automatic detection of falls. We are developing a reliable and efficient fall detection system that will monitor the movements of patients, recognize a fall, and automatically send a request for help to the caregivers. To reduce the problem of false alarms, the system will include novel techniques for the recognition of those activities of daily living that could be erroneously detected as falls (such as sitting on a sofa or lying on a bed). To limit the intrusiveness of the system, a small external sensing unit (worn on the wrist) will be used for the acquisition of movement data.

The second part of this project is an Application that continuously collects accelerometer data preprocess it and applies the fall detection algorithm, and sends help message to emergency contacts when a fall is detected. The app will be used either as a stand-alone system or with bluetooth enabled wearable sensor systems.

Project Objectives

• The main purpose of this application is to develop an automatic system for old persons specifically and for everyone generally who are not able to tell others that they are in problem now and need any kind of medical or physical aid. It will automatically send an alert like a message or call to our emergency contact number to let them know they are in an emergency and need their assistance as soon as possible.
• Our main focus is to develop a reliable system with reduces the number of false alarms which doesn’t miss any positive alarm when patient get caught by an emergency.
• It will automatically inform our emergency contact number that we are facing any unhandled situation and provide us aid as soon as possible.
• Our next focus is to improve the accuracy of our application. By training different machine learning models we anticipate achieving an accuracy of around 97% by implementing random forest classifiers. We will also improve its sensitivity and specificity.
• Reducing the number of false alarms is also our main target so that the app doesn’t miss any positive alarm and the survival chances of the patient can be increased by providing them aid as soon as we get a message from our patient number.
• The next target is to develop our app in such a way that it is simple and easy to use for elders. They can understand its functionality and working easily. They can select provided options according to their choices like sensor and alarm duration.
• We will make it possible that app doesn’t consume much battery as compared to other apps. For this purpose, we implemented two methods one is the threshold method and the next is machine learning based algorithm. First threshold method is implemented and checked if it is really a fall or not. If it is not a fall the machine learning method is not implemented. In this way, most of the events are discarded after the threshold method and machine learning method is not implemented every time. It reduces the battery consumption of the app.

Project Implementation Method

Fall Detection

Our system uses a two-step fall detection approach composed of TBM and Random-Forest. Data coming from Shimmer sensor will be filtered by a low pass Butterworth filter. Then the first step of the algorithm relies on the TBM for determining fall-like events and efficiently discarding most fall-like activities of daily learning. The threshold value 3-g will be selected as it is widely reported in the literature and is small enough to avoid false negatives, as even low impact falls have peak acceleration greater than 3-g value.

For the implementation of second step of the algorithm, 15 features will be extracted from data and then app random forest will be applied to give the final result (either fall or not).

App Development

We will design a standalone and user-independent fall detection system that will actively run in the background and uses a two-step algorithm to analyze subject movement. Upon fall detection, the application will trigger the SP to vibrate and an alert cancellation page appears on the screen. Unless canceled within a specified time period (default setting of 30 s) by the user, a sound alarm will be activated followed by a help text message containing location information being sent to specified emergency contacts.

The GUI of the app will consist of four screens: the main page, settings, fall alert cancellation, and feedback. The layout will be designed to facilitate usability by the elderly with an overall focus on reducing the battery power consumed particularly for unnecessary computations.

Benefits of the Project

Falls are responsible for majorly 90% of hip and wrist fractures and 60% of head injuries. Besides these injuries, long-lie situation (i.e., remaining on the ground for a long time) is another outcome of the fall that has serious consequences such as dehydration, hypothermia, and even death. Moreover, frequent incidence of falls in the elderly may provoke fear of falling, which in turn, deteriorates their confidence in living independently and being socially active. Thus, the development of automated, reliable, and prompt fall detection systems is vital to guarantee immediate assistance in case of falls, especially those involving long lies, and minimize severe health complications.

Commercial fall detection devices are expensive and charge a monthly fee for their services. (Life Alert Company in US charges $198 per month). In comparison, the cost of our product will be low.

Technical Details of Final Deliverable

Hardware

We will implement the system using a Shimmer3 wireless sensor, produced by Shimmer. The Shimmer3 Consensys IMU Development Kit provides an intuitive solution in body worn applications. The Shimmer3 IMU comes with integrated 9 DoF + altimeter inertial sensing via accelerometer, gyroscope and magnetometer, each with selectable range. The device includes 802.15.4 and Bluetooth radios. The latter has been used for communication with the smartphone.

Software

The software on the smartphone side runs on top of the Android operating system, while the software running on the external sensing unit has been written using the ConsensysBASIC software.

App will be developed using Android Studio IDE with min API 17 and target API level 29. The GUI of application will consist of four screens: the main page, settings, fall alert cancellation,and feedback.

1) Configuration and Control: The main page of the application will serve three functions: 1) start/stop the fall detection process; 2) application configuration settings; and 3) summarized display of critical/crucial settings. Once the user starts the sensing process, a notification icon will appear on the top left corner of the screen in the notification bar. The icon will remain visible as long as the application is running in the background. The settings page can be used to customize personal details, location of the SP being carried, fall detection service priority, and settings related to fall alert notification. Settings will be saved using the Shared Preferences class, which permits previous settings to persist over multiple sessions as well as after SP reboot.

2) Fall Detection Service: The algorithm proposed for fall detection will be implemented as an Android service, using the Intent Service class, that can run continuously in the background irrespective of the application. The intent service will run the algorithm in its own separate worker thread without blocking the main UI thread that otherwise can make the application nonresponsive. Once the service is activated, it will instantly acquire the PARTIAL_WAKE_LOCK, which prevents the CPU from going into sleep mode when the phone is idle. The incoming sensor data values are stored in a linked-list queue following the first-in-first-out (FIFO) discipline and maintains a data history of up to 6 s. steps of the TBM algorithm use this data to detect fall like events. Upon detection, the Random Forest algorithm will then be executed.

3) Notification System: The system will retrieve the last known geographical position of the user based on the available location providers (like GPS, Network) and sends the most accurate result in the form of an alert text message to the prespecified emergency contact number.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Medical

Other Industries

IT , Health

Core Technology

Artificial Intelligence(AI)

Other Technologies

Wearables and Implantables

Sustainable Development Goals

Good Health and Well-Being for People

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com