Transfer Learning for RF Domain Adaptation – Synthetic Dataset
A synthetic dataset designed to evaluate transfer learning performance for RF domain adaptation in the publication Assessing the Value of Transfer Learning Metrics for RF Domain Adaptation. The dataset contains a total of 13.8 million examples, with 600k examples each of 22 modulation schemes (given below) and AWGN noise (200k each for training, validation, and testing); 512 raw IQ samples per example. For each example, the signal-to-noise ratio is uniformly selected from the range [-10dB, 20dB], and the frequency offset is uniformly selected from the range [-10%, 10%] of sample rate. Further details can be found in the publication.
Modulation schemes included:
· PSK of order 2, 4, 8, and 16
· DPSK of order 4
· QAM of order 16, 32, and 64
· APSK of order 16 and 32
· FSK with 5k and 75k carrier spacing
· GFSK with 5k and 75k carrier spacing
· Narrowband and wideband FM
· Double sideband, double sideband suppressed carrier, lower sideband, and upper sideband AM
This dataset was generated using Python wrappers around liquid-dsp (https://github.com/jgaeddert/liquid-dsp), and is saved in SigMF format such that each example is saved in an individual ‘.sigmf-data’ file with an associated ‘.sigmf-meta’ file of the same name. The ‘.sigmf-data’ file contains the interleaved raw IQ samples in binary format and can be read using the numpy.load() function. The ‘.sigmf-meta’ file contains all metadata parameters used to generate the example including the number of samples, modulation type, signal-to-noise ratio, frequency offset, and filtering parameters, is in json format ,and can be read using json.load(). Further details and code examples for loading the dataset can be found at https://github.com/gnuradio/SigMF.