Modeling and Characterization of a Propagation Channel in a Body-Centric Nano-Network

Researchers have been trying to find smart health solutions that will allow people to continuously monitor their health through applications connected to the internet. One possible solution involve using nano-machines to create body-centric networks. However, the biggest challenge using nano-machines are how they would communicate with the outside world. To investigate this, I have in this thesis developed a multi-layer channel model for human skin tissues and investigated how signals at the terahertz frequency band interact with skin biomaterial. The model is built upon analytical equations describing electromagnetic propagation in a dielectric medium were the electromagnetic properties of human tissue were collected from different sources. The model were implemented in a flexible matlab program able to simulate different numbers of layers from a library with either fixed or random depths. The human skin model used to gather results were chosen to consist of 4 layers of epidermis, dermis, blood and hypodermis, and the depth of the layers were chosen to vary between typical values for the human body. This Matlab based multi-layer channel model was validated by a similar model made in CST Studio Suite, both for a single layer as well as for a 2-layer scenario. Results from the Matlab program showed that the path loss is significantly affected by frequency and material. The expected path loss could therefore vary significantly, however for a human skin model with depths of 1.23, 3.76, 0.21 and 1.38 mm respectively, the path loss was approximately between 250-350 dB for frequencies of 0.5-1.5 THz at the end distance. Finally, numerical analysis were used on 10 data sets created from the multi-layer channel model in order to develop a simple interpolation equation able to describe path loss through the human skin with varying tissue layer depths. The equation had an average mean error of 4.08 \% and a maximum mean error of 6.61\% against 90 different random data sets