LOS/NLOS Classification for UAV Communications: A Time-Frequency Multimodal Learning Approach

In communication networks, distinguishing between line-of-sight (LOS) and non-line-of-sight (NLOS) is critical for optimizing signal quality, coverage and network reliability. It influences network design, frequency band selection, and technology choices to ensure efficient and reliable communications. In this paper, we design an noval multimodel detection method to distinguish LOS and NLOS communication scenarios. This method combines time-domain channel characterization with frequency-domain spectrogram analysis, Time-Frequency Multimodal Detection (TFMD). Importantly, we show that the channel quality indicator (CQI) and the downlink coding and modulation scheme (DL MCS) are of paramount importance in the characterization of the signal. Moreover, we leverage the deep learning framework, 'you only look once' (YOLO), which shows great value in the detection of signals. We perform a precise comparative evaluation of unimodal and multimodal datasets, with a special focus on the accuracy and classification capabilities. The results highlight the significant advantages of multimodal detection methods in distinguishing between LOS and NLOS states, which achieve an accuracy of over 98%.