Related Experiment Video
Updated: Sep 21, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
530
MHSA-EC: An Indoor Localization Algorithm Fusing the Multi-Head Self-Attention Mechanism and Effective CSI
Wen Liu1, Mingjie Jia1, Zhongliang Deng1
1School of Electronic Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China.
Entropy (Basel, Switzerland)
|May 28, 2022
Summary
This study introduces MHSA-EC, an indoor positioning algorithm using channel state information (CSI) and a multi-head self-attention mechanism. It improves accuracy by better extracting multi-path effects and reducing long-distance point mismatches.
Area of Science:
- Wireless communications
- Signal processing
- Indoor localization
Background:
- Channel state information (CSI) is crucial for indoor positioning due to its sensitivity to multipath effects.
- Existing methods struggle to aggregate non-adjacent CSI features, limiting multipath information extraction and causing long-distance positioning errors.
Purpose of the Study:
- To propose a novel indoor localization algorithm, MHSA-EC, addressing limitations in aggregating long-distance CSI features and mitigating mismatches.
- To enhance the extraction of multipath information from CSI signals for improved positioning accuracy.
Main Methods:
- Developed the Multi-Head Self-Attention and Effective CSI (MHSA-EC) algorithm.
- Utilized the multi-head self-attention mechanism to aggregate non-adjacent CSI features effectively.
- Focused on leveraging correlations between subcarriers and antennas influenced by multipath effects.
Main Results:
- MHSA-EC demonstrated superior performance in aggregating long-distance CSI features compared to traditional methods.
- The algorithm significantly reduced mismatches for long-distance points.
- Achieved an average positioning error of 0.71 m in an office environment and 0.64 m in a laboratory.
Conclusions:
- MHSA-EC offers a stable and accurate solution for indoor localization.
- The proposed method effectively extracts multipath information, leading to enhanced positioning precision.
