ASeO-CNN: Active Search Optimization-enabled Convolutional Neural Network for Channel estimation on Fifth Generation wireless channels

Abstract

Channel estimation remains a crucial requirement to ensure efficient spectrum usage, enhance the network performance, and reliable data transmission in fifth-generation wireless communication systems. The prevailing channel estimation methods struggle with reduced latency, pilot overhead, and higher computational complexity in dynamic wireless communication environments. Henceforth, the research proposes an Active Search Optimization-enabled Convolutional Neural Network(ASeO-CNN) for channel estimation. Through leveraging the intrinsic complex feature extraction characteristics, the model acquires the complex channel features from the orthogonal frequency division multiplexing signals. Meanwhile, the Active Search Optimization(ASeO) employed optimizes the model hyperparameters and helps in enhancing the convergence speed and channel estimation accuracy. Furthermore, the Quadrature Amplitude Modulation(QAM) scheme helps in reliable data transmission through evaluating the complex channel characteristics significantly. Experimental findings with a 20dB Signal to Noise Ratio demonstrate the improved channel estimation performance by achieving 5.32 x 10-5 Bit Error Rate, 0.97 correlation, 0.13 Mean Absolute Error, 0.040 Mean Squared Error, and 0.933 R-Squared metrics.