Machine Learning for MPC Extraction and False Detection

Published

July 8, 2025

Author(s)

Howard Dai, Jack Chuang, Jian Wang, Samuel Berweger, David Griffith

Abstract

Multipath component (MPC) extraction is critical for channel modeling and joint communications and sensing (JCAS). Super-resolution algorithm known as CLEAN-Space-Alternating Generalized Expectation (CLEAN-SAGE) is widely used for MPC extraction. Because the algorithm is based on maximum likelihood estimation (MLE), it can suffer performance degradation and false detection events when a model mismatch occurs between the theoretical model and the received signal. Moreover, the complexity of CLEAN-SAGE makes real-time or near-real-time applications difficult to support because the algorithm iterative extracts parameters of MPCs with gradient search. In this paper, we propose two machine learning (ML)-based solutions to address these issues: a classification model for false signal detection and a regression model for direct MPC parameters extraction. The classification model uses a convolutional neural network (CNN) to predict truth or false detection from beamformed received signal, which can be used alongside CLEAN-SAGE to reduce false detections. Our model achieves greater than 90% classification accuracy on both simulated data and a smaller set of real-world examples collected by NIST 140 GHz phased array channel sounder. The regression model uses multiple CNN based encoder and decoder and several fully connected linear layers to directly predict numbers of MPCs and ranges and azimuth angles of each MPCs from a heatmap. We leverage the NIST Wi-Fi JCAS simulation bench to generate large amounts of heatmap training data in the signal-to-noise ratio (SNR) range of –10 dB to 25 dB. Our model achieves comparable regression accuracy and a less than 1 % false detection rate in multi-target scenarios

Proceedings Title

SPIE Defense + Commerical Sensing: Signal Processing, Sensor/Information Fusion, and Target Recognition XXXIV

Conference Dates

April 13-17, 2025

Conference Location

Orlando, FL, US

Conference Title

SPIE Defense + Commerical Sensing

Pub Type

Conferences

Download Paper

Local Download

Keywords

Deep Learning, NextG, Joint communications and sensing, Channel Sounding

Citation

Dai, H. , Chuang, J. , Wang, J. , Berweger, S. and Griffith, D. (2025), Machine Learning for MPC Extraction and False Detection, SPIE Defense + Commerical Sensing: Signal Processing, Sensor/Information Fusion, and Target Recognition XXXIV, Orlando, FL, US, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=958600 (Accessed October 27, 2025)