Optimizing Signal Detection in MIMO Systems: AI vs Approximate and Linear Detectors

Artificial intelligence has transformed multiple input multiple output (MIMO) technology into a promising candidate for six-generation networks. However, several interference signals impact the data transmission between various antennas; therefore, sophisticated signal detection techniques are required at the MIMO receiver to estimate the transmitted data. This paper presents an optimized AI-based signal detection technique called AIDETECT for MIMO systems. The proposed AIDETECT network model is developed based on an optimized deep neural network (DNN) architecture, whose efficiency lies in its lightweight network architecture. To train and test the AIDETECT network model, we generate and process the data in a suitable form based on the transmitted signal, channel information, and noise. Based on this data, we calculate the received signal at the receiver end, where the received signal and channel information were integrated into the AIDETECT network model to perform reliable signal detection. Simulation results show that at a 20-dB signal-to-noise ratio (SNR), the proposed AIDETECT technique achieves between 97.33% to 99.99% better performance compared to conventional MIMO detectors and is also able to accomplish between 25.34% to 99.98% better performance than other AI-based MIMO detectors for the considered performance metrics. In addition, due to lightweight network architecture, the proposed AIDETECT technique has also achieved much lower computational complexity than conventional and AI-based MIMO detectors.