Some preprocessing steps to refine the raw signal data are mentioned below:
- Noise filtering - You can use lowpass, bandpass, or any other filtering option
- Clutter removal - You can try Moving Target Indication filters or background subtraction to remove stationary objects
- Range processing - Convert time-domain signal to frequency-domain and analyze range using Fast Fourier Transform, Hamming, or Hann
- Doppler processing - Use FFT along the pulse repetition interval to get Doppler frequency shift
- Normalization - Scale the features to have mean and standard deviation of zero and one respectively
- Data Augmentation - Generate more training data by introducing minor variations
- Dimensionality reduction - Principal Component Analysis
- Training/Validation split
- etc.
Hope this helps!
