Network as Sensor
(2025) EITM02 20241Department of Electrical and Information Technology
- Abstract
- The thesis introduces a concept of leveraging advanced 5G, or even future 6G networks, as powerful sensing systems. By utilizing communication network infrastructure, it explores the potential of deploying transmitter and receiver nodes at different locations, thus enabling the detection of Non-Line-of-Sight (NLOS), passive, dynamic objects transforming the network into dual-purpose communication and sensing platforms.
This thesis presents a novel approach to human fall detection in an indoor environment and cyclist tracking in an outdoor environment, using advanced signal processing techniques. For robust fall detection, key scenarios like walking, sit-to-stand transitions, falling, and static postures are crucial indicators that the... (More) - The thesis introduces a concept of leveraging advanced 5G, or even future 6G networks, as powerful sensing systems. By utilizing communication network infrastructure, it explores the potential of deploying transmitter and receiver nodes at different locations, thus enabling the detection of Non-Line-of-Sight (NLOS), passive, dynamic objects transforming the network into dual-purpose communication and sensing platforms.
This thesis presents a novel approach to human fall detection in an indoor environment and cyclist tracking in an outdoor environment, using advanced signal processing techniques. For robust fall detection, key scenarios like walking, sit-to-stand transitions, falling, and static postures are crucial indicators that the network must accurately identify to ensure its effectiveness and reliability in real-world applications. Utilizing MATLAB for simulations, the research integrates the beam steering capability of a custom-designed patch microstrip antenna array operating at a chosen set of carrier frequencies, the ray tracing technique and machine learning algorithms. The antenna arrays, including both transmit and receive units, are designed to optimize detection capabilities of passive objects.
Two optimally chosen waveforms are passed through a simulated propagation environment, including multi-path effects, and are subsequently received and processed. Channels and noise are modeled using Ray Tracing Channels (RTC) and Additive White Gaussian Noise (AWGN). The machine learning process involves segmenting the signal data, assigning labels based on predefined categories, and training a Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) using Doppler signatures to ensure robust performance. The network architecture includes layers designed for feature extraction and classification, optimized using the Adam optimizer.
The results demonstrate the efficacy of the proposed system in accurately detecting falls, with high classification accuracy confirmed by confusion matrix analysis and Cumulative Distribution Function (CDF) plot for positioning error. This work contributes to the field of health monitoring by providing a reliable and efficient method for human fall detection. (Less) - Popular Abstract
- Being able to see and analyze objects or occurrences without needing to make direct physical contact with them but just by using signals has penned up new paradigms of imagination. It could function like sixth sense, enhancing awareness of the surroundings. Bistatic and multistatic sensing technology in Integrated Sensing and Communication (ISAC) exemplify the state-of-the-art technique in remote sensing technology, offering hitherto unseen capabilities for viewing and understanding the environment.
To achieve bistatic sensing, the transmitter and the receiver are deliberately positioned at different locations in contrast to monostatic sensing. By analyzing the way signals are reflected off of things and back to the receiver, we may... (More) - Being able to see and analyze objects or occurrences without needing to make direct physical contact with them but just by using signals has penned up new paradigms of imagination. It could function like sixth sense, enhancing awareness of the surroundings. Bistatic and multistatic sensing technology in Integrated Sensing and Communication (ISAC) exemplify the state-of-the-art technique in remote sensing technology, offering hitherto unseen capabilities for viewing and understanding the environment.
To achieve bistatic sensing, the transmitter and the receiver are deliberately positioned at different locations in contrast to monostatic sensing. By analyzing the way signals are reflected off of things and back to the receiver, we may extract crucial information about their properties. This technique offers up a wide range of applications, such as surveillance, environmental monitoring, and medical imaging, and it completely changes the way to extract information about surroundings.
Multistatic sensing goes one step farther by necessitating the deployment of several transmitters and receivers in varied configurations. With such configurations, it is possible to collect richer and more detailed data, which gives the
unique ability to precisely identify minute details in our surroundings. Multistatic sensing provides a broad and potent tool for sensing NLOS, passive and dynamic targets, whether it is for controlling vehicular traffic or human fall detection using reflected and diffracted rays.
These innovative technologies have a great deal of potential for improving security and safety as well as expanding scientific knowledge and exploration. Bistatic and multistatic sensing provide an insight into a future in which it will be possible to see and comprehend the surroundings with a level of clarity and depth never before possible, by utilizing the power of electromagnetic waves, machine learning and complex signal processing algorithms. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9186965
- author
- Aravinda Prabhu, Nisha Prabhu LU and Köhn, Miranda
- supervisor
- organization
- course
- EITM02 20241
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- report number
- LU/LTH-EIT 2025-1044
- language
- English
- id
- 9186965
- date added to LUP
- 2025-03-26 16:31:02
- date last changed
- 2025-03-26 16:31:02
@misc{9186965,
abstract = {{The thesis introduces a concept of leveraging advanced 5G, or even future 6G networks, as powerful sensing systems. By utilizing communication network infrastructure, it explores the potential of deploying transmitter and receiver nodes at different locations, thus enabling the detection of Non-Line-of-Sight (NLOS), passive, dynamic objects transforming the network into dual-purpose communication and sensing platforms.
This thesis presents a novel approach to human fall detection in an indoor environment and cyclist tracking in an outdoor environment, using advanced signal processing techniques. For robust fall detection, key scenarios like walking, sit-to-stand transitions, falling, and static postures are crucial indicators that the network must accurately identify to ensure its effectiveness and reliability in real-world applications. Utilizing MATLAB for simulations, the research integrates the beam steering capability of a custom-designed patch microstrip antenna array operating at a chosen set of carrier frequencies, the ray tracing technique and machine learning algorithms. The antenna arrays, including both transmit and receive units, are designed to optimize detection capabilities of passive objects.
Two optimally chosen waveforms are passed through a simulated propagation environment, including multi-path effects, and are subsequently received and processed. Channels and noise are modeled using Ray Tracing Channels (RTC) and Additive White Gaussian Noise (AWGN). The machine learning process involves segmenting the signal data, assigning labels based on predefined categories, and training a Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) using Doppler signatures to ensure robust performance. The network architecture includes layers designed for feature extraction and classification, optimized using the Adam optimizer.
The results demonstrate the efficacy of the proposed system in accurately detecting falls, with high classification accuracy confirmed by confusion matrix analysis and Cumulative Distribution Function (CDF) plot for positioning error. This work contributes to the field of health monitoring by providing a reliable and efficient method for human fall detection.}},
author = {{Aravinda Prabhu, Nisha Prabhu and Köhn, Miranda}},
language = {{eng}},
note = {{Student Paper}},
title = {{Network as Sensor}},
year = {{2025}},
}