# Multi-User Matching and Resource Allocation in Vision Aided Communications

**Authors:** Weihua Xu, Feifei Gao, Yong Zhang, Chengkang Pan, Guangyi Liu  
**arXiv:** [2304.09075](https://arxiv.org/abs/2304.09075)  
**Date:** 2023-04-18  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.09075v1/

## Key points

- The paper proposes a user matching method for environments with a variable number of objects.
- Visual perception is an effective way to obtain the spatial characteristics of wireless channels and reduce overhead for communication systems.
- 3D detection is applied to extract all environmental objects from images taken by multiple cameras.
- A deep neural network (DNN) is designed to estimate the location distribution of users using images and beam pairs at multiple moments, thereby identifying users from all extracted environmental objects.
- The paper presents a resource allocation method based on taken images that reduces time and spectrum overhead compared to traditional methods.
- Simulation results show that the proposed user matching method outperforms existing methods, while the proposed resource allocation method achieves a 92% transmission rate of traditional methods but with significantly reduced time and spectrum overhead.
- The authors are Weihua Xu, Feifei Gao, Yong Zhang, Chengkang Pan, and Guangyi Liu.

## Summary

The paper "Multi-User Matching and Resource Allocation in Vision Aided Communications" proposes a user matching method for environments with a variable number of objects. The authors highlight that visual perception is an effective way to obtain the spatial characteristics of wireless channels and reduce overhead for communication systems. To address the critical problem of matching radio signals with visual information from corresponding users, the authors apply 3D detection to extract all environmental objects from images taken by multiple cameras. They then design a deep neural network (DNN) to estimate the location distribution of users using images and beam pairs at multiple moments, thereby identifying users from all extracted environmental objects. Furthermore, the paper presents a resource allocation method based on taken images that reduces time and spectrum overhead compared to traditional methods. Simulation results show that the proposed user matching method outperforms existing methods, while the proposed resource allocation method achieves a 92% transmission rate of traditional methods but with significantly reduced time and spectrum overhead. The authors are Weihua Xu, Feifei Gao, Yong Zhang, Chengkang Pan, and Guangyi Liu. Overall, their work contributes to advancing vision-aided communication systems by addressing critical challenges associated with user matching and resource allocation in complex environments.