Neural network-based path loss prediction for digital TV macrocells
Path loss prediction in radio wave propagation models are often categorized as theoretical/physical, empirical or a hybrid combination. Theoretical propagation models rely more on the physical behavior of radio waves while empirical models are based on actual field strength measurements in a particu...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | text |
| Published: |
Animo Repository
2016
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/faculty_research/1735 https://animorepository.dlsu.edu.ph/context/faculty_research/article/2734/type/native/viewcontent |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Summary: | Path loss prediction in radio wave propagation models are often categorized as theoretical/physical, empirical or a hybrid combination. Theoretical propagation models rely more on the physical behavior of radio waves while empirical models are based on actual field strength measurements in a particular environment. Consequently, the equations for theoretical models are based on physics while those for empirical models are based on statistical analysis of the gathered data. While physical models can be adapted to any type of environment, they are known for their computational complexity since they consider the path profile to each and every point in a given area. Empirical models are attractive for their computational efficiency but they may only be suited to the specific place where the measurements were conducted. This paper aims to propose and ascertain the viability of using an alternative neural network (NN) model to predict path loss. The approach may be loosely categorized as empirical since actual field strength measurements should be basis for the prediction model. For purposes of preliminary analysis, however, the field strength measurements will be replaced with results from a Longley-Rice model simulation, which is often used for TV transmission. The simulation data will be used to generate training and validation for the neural network. Once neural network fitting using feedforward backpropagation is achieved, the neural network-based propagation model is shown to give more accurate results compared to more familiar propagation models, such as Free Space and Egli, while having the advantage of adaptability to arbitrary environments. © 2015 IEEE. |
|---|