Attention Mechanism and LSTM Network for Fingerprint-Based Indoor Location System

Zhen Wu,Peng Hu,Shuangyue Liu,Tao Pang

DOI: https://doi.org/10.3390/s24051398

IF: 3.9

2024-02-23

Sensors

Abstract:The demand for precise indoor localization services is steadily increasing. Among various methods, fingerprint-based indoor localization has become a popular choice due to its exceptional accuracy, cost-effectiveness, and ease of implementation. However, its performance degrades significantly as a result of multipath signal attenuation and environmental changes. In this paper, we propose an indoor localization method based on fingerprints using self-attention and long short-term memory (LSTM). By integrating a self-attention mechanism and LSTM network, the proposed method exhibits outstanding positioning accuracy and robustness in diverse experimental environments. The performance of the proposed method is evaluated under two different experimental scenarios, which involve 2D and 3D moving trajectories, respectively. The experimental results demonstrate that our approach achieves an average localization error of 1.76 m and 2.83 m in the respective scenarios, outperforming the existing state-of-the-art methods by 42.67% and 31.64%.

engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in indoor positioning systems, fingerprint - based positioning methods are affected by multipath signal attenuation and environmental changes, resulting in a significant decline in their performance. Specifically, the paper proposes a method that combines the self - attention mechanism and the Long - Short - Term Memory network (LSTM) to improve the accuracy and robustness of fingerprint - based indoor positioning systems. By integrating the self - attention mechanism and the LSTM network, the proposed method performs well in different experimental environments and can effectively deal with problems such as multipath effects, shadow effects and signal fading, thereby improving positioning accuracy and system robustness.

The main contributions of the paper include:
1. Proposing a new deep neural network that integrates the self - attention mechanism and the LSTM network for indoor positioning. This method processes RSSI values of continuous time instances and predicts the position at the last moment of the input sequence.
2. Based on LSTM, introducing the self - attention mechanism, enabling LSTM to effectively capture the dependencies between different time instances, thereby improving the extraction of location information and reducing positioning errors.
3. Through comparative analysis with several state - of - the - art methods, the experimental results show that the proposed SA - LSTM model achieves the highest positioning accuracy in both experimental scenarios, demonstrating its robustness and accuracy.

These contributions aim to solve the problem of limited performance of existing fingerprint - based indoor positioning methods in complex indoor environments and provide a more accurate and robust solution.