Supervised Machine Learning Assisted Hybrid Positioning Based on GNSS and 5G

Duong, Phuong Bich; Ghimire, Birendra; Dietmayer, Katrin; Ali, Sheikh Usman; Kim, Haider Al; Seitz, Jochen

doi:10.1109/ipin54987.2022.9918098

Cited by 3 publications

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Historically, conventional approaches have typically employed statistical methods such as thresholding, root mean square delay, signal rise time, etc. [4][5][6], or machine learning methods like Support Vector Machines (SVM), logistic regression, and others [7][8][9][10], to differentiate LOS from NLOS. However, these methods often relied on simplistic linear modeling when dealing with complex signals, resulting in suboptimal performance.…”

Section: Introductionmentioning

confidence: 99%

Non-line-of-sight ultra-wideband signal recognition using DWT-CNN

He,

Zhao,

Liu

et al. 2024

Fourth International Conference on Sensors and Information Technology (ICSI 2024)

View full text Add to dashboard Cite

With the development of technology, the application of positioning techniques has become increasingly widespread. However, in indoor or specific situations, due to the complexity of the environment, it can be challenging to apply outdoor positioning techniques. Therefore, research on indoor positioning techniques is highly necessary.Ultra-Wideband(UWB) technology, as an indoor positioning method, can address many indoor positioning scenarios. However, when facing obstacles, it can still lead to non-line-of-sight (NLOS) situations, resulting in less accurate positioning. To address this issue, this paper proposes an approach that combines Discrete Wavelet Transform (DWT) and Convolutional Neural Networks (CNN). Firstly, it extracts coefficients from Channel Impulse Response (CIR) signals using DWT. Subsequently, these coefficients are input to a CNN to explore deep features, ultimately distinguishing between Non-Line-of-Sight (NLOS) and Line-of-Sight (LOS) signals based on these features. Experimental results demonstrate that this method achieves better classification accuracy compared to other approaches.

show abstract