Classifying 2D Spectrograms of Deep Space Radio Signals
As a part of an online course ,I did a project for classifying 2D spectrograms of space signals.Simply put,building and training a CNN in keras. I thought it would be a good idea to document it as I (as my friend proved) suffer from short-term memory loss (like Dory :) )
Original signals are not 2D spectrograms. They are time series. I classified the signals by using the 2D data collected by the antennas of SETI (Search for Extra Terrestrial Intelligence) institute. It is important to distinguish between the various signals that we receive because we don’t want to waste a lot of time in a radio wave hoping to communicate with aliens when all it was , was nothing but a mere noise.
I classified 3 types of signals:
0 : Squiggle wave
1 : Narrow wave
2 : Noise
3 : Narrow band drd
Loading and Preprocessing SETI data
Random images of the radio signals in gray scale from the training data
Summary of the trained model
After the training an accuracy of 0.760 and loss of 0.372 was obtained
This is my end result . I feel happy to have completed this even if it’s a small step. No matter how quarantined we are , we can never quarantine our ideas and creativities of our minds.