On the Structured Design for Efficient Machine Learning Based PRACH Preamble Detection
Du Hui Yang, Sang Won Choi · 2023
In this paper, we consider a structured design for efficient PRACH preamble detection in conjunction with devised Machine Learning (ML) methodology. Specifically, the main contribution of this paper is to provide the receiver structure such that all the learned weights and biases in the ML are maintained irrespective of timing offset which varies in the actual PRACH preamble detection situation. Furthermore, by considering an enhanced input signal called Complex power Delay Profile (CDP) rather than Power Delay Profile (PDP), the proposed PRACH preamble detection scheme is shown to have not only efficient structure but also enhanced detector performance in the sense of detection and false alarm probabilities. For that, we show the effectiveness of the proposed detector structure by leveraging mathematical signal interpretation. Consequently, we show that there exists SNR gain where the proposed PRACH preamble detector satisfies the corresponding 3GPP performance requirement in comparison with the existing typical one, which validates the proposed structured design from the perspective of detector performance.