Using YOLOv9 Deep Learning Model for Automatic Detection and Classification of VHF Signals

Using YOLOv9 Deep Learning Model for Automatic Detection and Classification of VHF Signals

연구 분야: Infrastructure

학회: 2024 International Symposium on Electronics and Telecommunications (ISETC)

The present work aims to highlight the results of using a convolutional neural network algorithm, namely You Only Look Once (YOLO) v9 in classification of very high frequency (VHF) emissions based on spectrograms recognition. A low-cost software-defined radio (SDR) transceiver with as low as 8 bits conversion capability was used to record four different types of VHF signals. Images of the spectrograms were used for training, validation, and testing of the dataset. The correct classification percentages were very good for all four classes of investigated radio emissions, with an average of 83.4%, even if the number of spectrograms per emission type used for training was as low as 150 and for validation 30. In these conditions, it resulted that YOLOv9 proves excellent capabilities in identification and classification of VHF signals received by low-resolution, low sensitivity, limited dynamic range and reduced spectrogram accuracy provided by the cheap SDR platform.