SCSMN: A Signal-Channel Features Separation with MMSE Network for UAV Signal Feature Extraction

As an emerging device in the electromagnetic space, the safety issues during the operation of Unmanned Aerial Vehicles (UAVs) have attracted great attention. Some malicious behaviors tamper with information, making it difficult to determine the true identity, and even without considering the channel effect, the identification effect of radio frequency fingerprints is not satisfactory. With the maturity of artificial intelligence technology and in response to the above problems, this paper proposes a Signal-Channel feature Separation based on the MMSE Network (SCSMN). The network considers the influence of the channel and mainly separates the input signal into signal and channel time through convolution operation. Sequence, parameter correction is performed based on the MMSE algorithm, and various losses are formed through cross-classifier classification to guide the network for accurate recognition. Experiments show that SCSMN can reduce the impact of the channel on signal recognition and achieve accurate identification of UAVs.