Health Secure Radar: Software Defined Radio Based Micro-Doppler Radar for Healthcare and Security Applications

Aging and chronic diseases become prominent societal challenges and pose a significant strain on life quality. The resolving of these challenges can be shifted to understanding the life patterns of individuals. For example, vital signs can be used for monitoring medical problems and activity recogni

Project Title

Health Secure Radar: Software Defined Radio Based Micro-Doppler Radar for Healthcare and Security Applications

Project Area of Specialization

Electrical/Electronic Engineering

Project Summary

Aging and chronic diseases become prominent societal challenges and pose a significant strain on life quality. The resolving of these challenges can be shifted to understanding the life patterns of individuals. For example, vital signs can be used for monitoring medical problems and activity recognition can be used to forecast and prevent many chronic diseases. However, capturing such information is a challenging task. Many diverse approaches have been proposed from both industry and academia for in-home healthcare informatics and can be generally divided into wearable sensors and non-contact sensors. Currently, wearable sensors are considered as the primary technique as a source of rich information pertaining to health. However, wearable sensors are not suitable for everyone. Some patient groups like people with skin diseases, Parkinson’s, and infants are discouraged from wearing such sensors. Thus, the non-contact sensors may be the only solution for those who cannot wear any device on their body. Our project aim is the use of radar systems for indoor monitoring of people to infer information regarding the health status of vulnerable people such as the elderly and people with cognitive or physical disabilities. Radar can provide a contactless and unobtrusive tool to perform this monitoring, with no need of using invasive cameras or wearable devices. Micro-Doppler indicates the small modulations on the radar echoes generated by movements and rotations of various parts of humans e.g., the small swinging movements of limbs and torso oscillations. Human micro-Doppler signatures extracted from radar sensors have been investigated in recent years for a variety of applications, such as the classification of different activities performed by people (walking, running, carrying objects), and healthcare/assistive purposes (e.g., early detection of fall events involving elderly people). With the use of this technology, our project is also aiming at security applications. For example, the use of conventional and mini drones has been increased considerably for a variety of applications. However, this poses a serious threat to sensitive installations and gatherings owing to their potential use for malicious activities. The radar micro-Doppler signature-based approach is a new method for non-cooperative target identification, recognition, and classification. These drones are difficult to detect as they fly at a very low speed and altitude and they have a very low radar cross-section. Due to their stability and robustness, micro-Doppler characteristics can be used to extract unique signatures of a target.

Project Objectives

The main project objectives that will be addressed in this project are the following:

• To design an optimum waveform for detection of micro doppler signatures for indoor monitoring of humans to infer information regarding the health status of vulnerable people such as elderly and people with cognitive or physical disabilities. Moreover, to detect fall detection and drone detection for health and security applications respectively
• To design algorithms that can help differentiate different types of complex motions such as vibration, rotation, or acceleration. 
• To develop novel signal processing algorithms for locating and monitoring human activities like walking, running, fall detection, heartbeat, breathing rate.
• To develop novel signal processing algorithms for locating and monitoring drones for military and security applications.
• To develop novel algorithms for the classification of micro doppler signatures using deep learning tools.
• To use software-defined radios for Implementation and validation of the developed algorithms.

Project Implementation Method

For obtaining micro doppler signatures data of different drones and objects, we use 2 USRP-2921 (Universal Software Radio Peripheral) Radars for transmitting and receiving the signal which is programmed through GNU Radio in Linux. This USRP sends and receives backscattering signals from the target object. The data obtained from backscattering signals are then processed in MATLAB in which we use time-frequency analysis i.e., Short Time Fourier Transform (STFT), and obtain the spectrogram of the backscattering signals. The STFT captures the unique signatures of different materials and can be used to classify different objects. After this, we use deep learning models for the classification of the micro doppler signals while comparing them with various training data sets. The deep learning model we are using is Convolutional Neural Networks (CNNs). For deep learning work, we are using Google Colaboratory. 

Benefits of the Project

During the last 30 years, Pakistan has undergone extreme transformations with respect to population and economic conditions. As a hazard-prone country, with more people living in high-risk areas than ever before, it is increasingly important to pro-actively address natural and manmade hazards and the cumulative risks. In this project, we aim to develop a complete prototype that will allow detection to detect the presence of drones and humans using software-defined MIMO radars and signal processing. Humans usually produce movements that are periodic in nature for example movement of the chest due to inhaling and exhaling. This project will also benefit in various military applications like recognition of air and ground moving targets, through-wall detection, fall detection, and detection of humans using their breathing and walking motions. 

Technical Details of Final Deliverable

An electromagnetic wave is transmitted from a software-defined radio transmitter connected to a directional antenna. The transmitted signal gets reflected and scattered. In some cases, it penetrates through the object. The reflected signal is received by the radar receiver. The received signal is then processed to investigate the variations of the micro-Doppler signatures of the target. These signatures help us classify the targets based on their unique micro-Doppler signatures. Thereafter, a two-dimensional image of the target is generated which contains information about the target’s lateral extent and its location. For micro-Doppler signatures, we used USRP-2921 which is an SDR (Software Defined Radio) programmed through GNU Radio in Linux. This USRP sends and receives backscattering signals from the target object. Researchers have also been using FPGA (Field Programmable Gate Array ) based SDRs with GNU Radio to analyze and extract micro-Doppler signatures because FPGA offers good signal processing with advantages like small size, lightweight, and low power. Moreover, the FPGA-based SDR offers wider bandwidth and a user-friendly MATLAB interface but FPGA-SDR is a hardware-based process and it is costly. We have used USRP based SDR for extracting micro-Doppler signatures for its low-cost advantage. The disadvantage of using USRP based SDR is that it induces an additional time varying delay in the results. So, we have to calibrate the USRP for every run, but this delay is compensated by using a calibration scheme based on a frequency domain matched filter. For the transmitter and receiver, the sampling rate is set to 200KHz. The carrier frequency on which the transmitter and receiver are working is set to 2.4GHz. The gain of the transmitter is 20dB and that of the receiver is 15dB. The transmitting waveform is a 1KHz Sine Wave. For deep learning, we are using Convolutional Neural Networks (CNNs) as they are widely used in image classification. We are implementing CNNs in Google Colaboratory as it is a user-friendly environment. As we know, training a model takes time so Google Colaboratory provides us with an environment in which we can use GPUs to efficiently run and train our network.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Health

Other Industries

Security

Core Technology

Artificial Intelligence(AI)

Other Technologies

Others

Sustainable Development Goals

Good Health and Well-Being for People, Sustainable Cities and Communities

Required Resources

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