SDR implementation of a DSSS detector based Light CNN for cognitive radio

Abstract Direct Sequence Spread Spectrum (DSSS) communication has gained significant traction in both civil and military communications and applications, owing to its high processing gain, anti-jamming capability, and low probability of interception. Due to the wide bandwidth utilized in transmission, the power spectral density of a DSSS signal may drop below the noise level. Consequently, detecting such a signal poses challenges in cognitive radio networks. In this paper, the detection of DSSS signals in a blind scenario without any preprocessing of the received signals is conducted using a proposed deep learning model. This model, named DET-CNN, is built upon a Convolutional Neural Network (CNN). Through simulations and experimental analysis, the detection performance of DET-CNN in various scenarios has been assessed and evaluated, contrasting it with traditional methods such as : energy detection, autocorrelation detection methods, goodness of fit and models from the literature across different Signal-to-Noise Ratios (SNRs) in various channels. The results indicate that the detection performance of the proposed model surpasses the existing methods, whether based on conventional techniques or deep learning models. Our model achieved a detection probability of 100% at SNRs of −12 dB and −10 dB with synthetic and real data, respectively.