Machine Learning-Based PRACH Preamble Detection: Efficient Structured Design Perspective

In this paper, we propose a structured design for efficient detection of the Physical Random Access CHannel (PRACH) preamble, integrating a novel Machine Learning (ML) methodology. Specifically, our main contribution is the development of a receiver structure that preserves all learned weights and biases from the ML methodology, despite variations in timing offset commonly encountered in PRACH preamble detection scenarios. In addition, we introduce an enhanced input signal, termed the Complex Power Delay Profile (CDP), in place of the traditional Power Delay Profile (PDP). This enhancement not only streamlines the structural design but also improves the detection performance in terms of probability of detection and false alarm rates.We demonstrate the efficacy of our proposed detector structure by leveraging mathematical signal interpretation. Our results reveal a Signal-to-Noise Ratio (SNR) gain, confirming that the proposed PRACH preamble detector meets the corresponding the 3rd Generation Partnership Project (3GPP) performance requirements, thereby validating our structured design in terms of detector performance.