An ECA–ResNet-Based Intelligent Communication Scenario Identification Algorithm for 6G Wireless Communications

Abstract

The sixth generation (6G) wireless communication envisions global coverage, all spectra, and full applications, which correspondingly creates many new communication scenarios. As the foundation of 6G communication system design, network planning, and optimization, more intelligent scenario identification algorithms are necessitated in wireless channel modeling to automatically match suitable parameters for various scenarios. With channel statistics and the efficient channel attention (ECA) mechanism, we propose an improved residual network (ResNet) to identify scenarios in the 6G space–air–ground–sea framework. Datasets from both channel measurements and 6G pervasive channel model (6GPCM) simulations are collected to establish a scenario channel characteristic database, including the numbered scenarios and channel statistical properties such as root mean square (RMS) delay spread (DS), RMS angle spread (AS), and stationary distance/time/bandwidth, etc. During the training and verification process, the proposed algorithm is optimized for 29 scenarios, and the identification accuracy of the proposed ECA–ResNet is higher than the convolutional neural network (CNN) and recurrent neural network (RNN). Finally, the cumulative distribution functions (CDFs) of RMS AS and RMS DS for interoffice main road, office outdoor, office, and industrial Internet of Things (IIoT) scenarios are verified according to the measurement data.