The transmitted FM signal is received by using a simple RTL-SDR dongle, which provides the baseband FM modulated signal. This signal is demodulated with GNURadio and then a deep learning model classifies the audio segments into three different classes : music, speech, and ads.

Developed with a RTL-SDR USB dongle, GNURadio, gr-osmosdr, and PANNs, Pretrained Audio Neural Networks for Audio Pattern Recognition.