Convolutional Neural Networks based Denoising for Indoor Localization

Njima, Wafa; Chafii, Marwa; Nimr, Ahmad; Fettweis, Gerhard

doi:10.1109/vtc2021-spring51267.2021.9448839

Cited by 7 publications

(2 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In particular, deep learning (DL) methods, already provide a variety of advanced localization systems with high accuracy [11], [12]. However, they are data hungry requiring large labeled training databases.…”

Section: Introductionmentioning

confidence: 99%

DNN-Based Indoor Localization Under Limited Dataset Using GANs and Semi-Supervised Learning

2022

Self Cite

View full text Add to dashboard Cite

Indoor localization techniques based on supervised learning deliver great performance accuracy while maintaining low online complexity. However, such systems require massive amounts of data for offline training, which necessitates costly measurements. The essence of this paper is twofold with the purpose of providing solutions to missing data of different nature: available unlabeled data and missing unlabeled data. In both cases, we rely on a few labeled available data, which is costly yet insufficient to achieve a high localization accuracy. To address the problem of available unlabeled data, a weighted semisupervised DNN-based indoor localization approach leveraging pseudo-labeling methods in combination with real labeled samples and inexpensive pseudo-labeled samples is proposed in order to boost localization accuracy, while overcoming the high cost of collecting additional labeled data. As for the extreme case of unavailable unlabeled data, we propose an alternative localization system generating fake fingerprints based on generative adversarial networks (GANs) named 'Weighted GAN based indoor localization'. Furthermore, a deep neural network is trained on a mixed dataset containing both real collected and fake produced data samples using a similar weighting technique in order to improve location prediction performance and avoids overfitting. In terms of localization accuracy, our proposed localization approaches outperform conventional supervised localization schemes utilizing the same collection of real labeled samples. For experimental evaluations, we have tested our proposed methods on both simulated data and experimental data from the publicly available UJIndoorLoc database, which is built to test indoor positioning systems relying on Wi-Fi fingerprints. Results based on experimental data provide the localization accuracy increase compared to the classical supervised learning method using the same set of labeled collected data when using the weighted semi-supervised and the weighted-GAN approaches by 10.11 % and 8.53 %, respectively.

show abstract

Section: Introductionmentioning

confidence: 99%

DNN-Based Indoor Localization Under Limited Dataset Using GANs and Semi-Supervised Learning

2022

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, in a complex indoor environment, signal propagation is easily influenced by external factors; hence, the developed system framework is unstable. According to relevant studies, CNN can also be applied to indoor positioning systems to achieve a better positioning effect by matching the most suitable position in the way of image classification [ 22 , 23 , 24 ]. Sinha and Hwang proposed a CNN model to solve the instability of fingerprint databases built with RSS [ 25 ].…”

Section: Introductionmentioning

confidence: 99%

Improved CNN-Based Indoor Localization by Using RGB Images and DBSCAN Algorithm

Cheng

Niu

Zhang

et al. 2022

Sensors

View full text Add to dashboard Cite

With the intense deployment of wireless systems and the widespread use of intelligent equipment, the requirement for indoor positioning services is increasing, and Wi-Fi fingerprinting has emerged as the most often used approach to identifying indoor target users. The construction time of the Wi-Fi received signal strength (RSS) fingerprint database is short, but the positioning performance is unstable and susceptible to noise. Meanwhile, to strengthen indoor positioning precision, a fingerprints algorithm based on a convolution neural network (CNN) is often used. However, the number of reference points participating in the location estimation has a great influence on the positioning accuracy. There is no standard for the number of reference points involved in position estimation by traditional methods. For the above problems, the grayscale images corresponding to RSS and angle of arrival are fused into RGB images to improve stability. This paper presents a position estimation method based on the density-based spatial clustering of applications with noise (DBSCAN) algorithm, which can select appropriate reference points according to the situation. DBSCAN analyses the CNN output and can choose the number of reference points based on the situation. Finally, the position is approximated using the weighted k-nearest neighbors. The results show that the calculation error of our proposed method is at least 0.1–0.3 m less than that of the traditional method.

show abstract