Speaker
Description
Radio telescopes like the Giant Metrewave Radio Telescope (GMRT) in Pune detect very weak radio waves from sources such as pulsars and galaxies. Man-made signals from mobile phones, Wi-Fi and other electronic devices often disturb these waves. This is called radio frequency interference (RFI), and it reduces the clarity and accuracy of the data. Traditional methods like fixed filtering and manual flagging are slow and cannot adapt when the type of interference changes.
In this work, I used an Artificial Neural Network (ANN) to identify and reduce this noise. A smooth, weak signal was generated to represent a cosmic source, and sinusoidal noise was added at frequencies similar to a mobile phone and a Wi-Fi router. A feed-forward network with two hidden layers of 64 neurons was trained on pairs of noisy and clean signals. It learned to tell the sharp, strong interference apart from the weak astronomical signal and to reconstruct the clean signal.
Compared with traditional filtering, the ANN gave better noise removal, clearer signals and faster processing. Its accuracy was about 90–98%, against 70–80% for the older methods. Since the network can be retrained, it can also adapt to a changing radio environment.
The study used simulated signals, so the next step is to test the method on real GMRT data and on deeper models for real-time use. The results suggest that neural networks can be a simple, low-cost way to improve data quality in radio astronomy.
Other Fields of Research (Track) for Abstract
computational & signal processing
| Nature of work | Computational |
|---|