A probability neural network-Jensen-Shannon divergence for a fingerprint based localization

Abdullah, Osamah Ali; Abdel‐Qader, Ikhlas; Bazuin, Bradley J.

doi:10.1109/ciss.2016.7460516

Cited by 7 publications

(4 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a result, depending on L p -norm or square Euclidean distance algorithms do not always lead to systems with high accuracy. For example, it was proved in Reference [ 7 ] that the concave-convex procedure can obtain higher accuracy than algorithms that depend on the square Euclidean distance such as the kNN and probabilistic neural network (PNN). In this section, we introduce the symmetric Hölder divergence.…”

Section: Overall Structure Of Proposed Positioning Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Convex Optimization via Symmetrical Hölder Divergence for a WLAN Indoor Positioning System

Abdullah

2018

Entropy

View full text Add to dashboard Cite

Modern indoor positioning system services are important technologies that play vital roles in modern life, providing many services such as recruiting emergency healthcare providers and for security purposes. Several large companies, such as Microsoft, Apple, Nokia, and Google, have researched location-based services. Wireless indoor localization is key for pervasive computing applications and network optimization. Different approaches have been developed for this technique using WiFi signals. WiFi fingerprinting-based indoor localization has been widely used due to its simplicity, and algorithms that fingerprint WiFi signals at separate locations can achieve accuracy within a few meters. However, a major drawback of WiFi fingerprinting is the variance in received signal strength (RSS), as it fluctuates with time and changing environment. As the signal changes, so does the fingerprint database, which can change the distribution of the RSS (multimodal distribution). Thus, in this paper, we propose that symmetrical Hölder divergence, which is a statistical model of entropy that encapsulates both the skew Bhattacharyya divergence and Cauchy–Schwarz divergence that are closed-form formulas that can be used to measure the statistical dissimilarities between the same exponential family for the signals that have multivariate distributions. The Hölder divergence is asymmetric, so we used both left-sided and right-sided data so the centroid can be symmetrized to obtain the minimizer of the proposed algorithm. The experimental results showed that the symmetrized Hölder divergence consistently outperformed the traditional k nearest neighbor and probability neural network. In addition, with the proposed algorithm, the position error accuracy was about 1 m in buildings.

show abstract

Section: Overall Structure Of Proposed Positioning Algorithmmentioning

confidence: 99%

“…Thus, fingerprinting-based localization has become the more dominant technique in IPSs and has two major phases. First, the offline phase, in which the RSS value is recorded with their coordinates at predetermined reference points (RPs) to generate a radio map database [ 7 , 8 , 9 ].…”

Section: Introductionmentioning

confidence: 99%

Convex Optimization via Symmetrical Hölder Divergence for a WLAN Indoor Positioning System

Abdullah

2018

Entropy

View full text Add to dashboard Cite

show abstract

“…One is to learn the mapping relationship between RSS and position of MT, the other is to add the difference of RSS into the input layer of previous RBF neural network. Combining with Jensen-Shannon divergence as a measure of similarity, a probabilistic neural network (PNN) localization method is proposed in [13]. By transforming the problem of localization into classification problem, the work in [14] proposes a multi-layer neural network (MLNN) method for RSS-based indoor localization.…”

Section: Introductionmentioning

confidence: 99%

Neural Network Localization With TOA Measurements Based on Error Learning and Matching

Zhang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

Due to the widespread application of location information, the neural network localization method with the advantage of high localization accuracy has received significant interests in recent years. In this paper, we present two new neural network localization methods with time-of-arrival (TOA) measurements. In order to deal with three types of error about TOA measurements such as measurement error, non-line of sight (NLOS) error, and synchronization error, the proposed methods contain an offline training stage and an online localization stage. In the offline stage, the artificial neural network (ANN) or radial basis function (RBF) neural network is utilized to train the range measurements with the output of range errors rather than the position of the mobile terminal (MT). Moreover, due to the unknown signal propagation condition whether it is the line of sight or NLOS propagation, the k-mean clustering algorithm is used to classifying the range errors into different clusters. In the online stage, the range errors are predicted and updated, and then, the linear least square algorithm with the adjusted range measurements is applied for the position estimate of MT. Comparing with the ANN or RBF neural network localization methods, the simulation results show that the proposed methods can effectively reduce the localization error, especially when the training sample is not adequate. In addition, they are insensitive to measurement error, synchronization error, and the distribution of NLOS error. Finally, the memory requirement and computational complexity about different algorithms are analyzed and compared. INDEX TERMSArtificial neural network (ANN), localization, linear least square (LLS), non-line of sight (NLOS), radial basis function (RBF), time of arrival (TOA).However, the absence of direct line of sight (LOS) propagation with satellite will lead to severe deterioration of the localization performance [1]. There are some important supplements to the satellite-based technique, such as Wireless Sensor Network (WSN), Wireless Cellular Network (WCN) and Wireless Local Area Network (WLAN) [2], [3] techniques are used to localize the MT, especially in indoor environments and some outdoor environments where the satellite signals are heavily attenuated or blocked. These supplemental techniques have common principle to localize the MT. By measuring some signal parameters between MT and Fixed

show abstract

“…The fingerprint-based technique is divided into offline and online phases. In the offline phase, the entire area of interest is divided into a rectangular set of grid points, and at each point, a site survey is taken by recording the RSS from APs, which is then stored in a database called the radio map [6][7][8][9][10]. In the online phase, the smartphone collects the RSS from the APs and sends it to the server to compare the predefined fingerprint of the offline phase with the RSS in the online phase in order to estimate the location on the grid map, as shown in Figure 1.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning Algorithm for Wireless Indoor Localization

Abdullah¹,

Abdel‐Qader²

2018

Machine Learning - Advanced Techniques and Emerging Applications

View full text Add to dashboard Cite

Smartphones equipped with Wi-Fi technology are widely used nowadays. Due to the need for inexpensive indoor positioning systems (IPSs), many researchers have focused on Wi-Fi-based IPSs, which use wireless local area network received signal strength (RSS) data that are collected at distinct locations in indoor environments called reference points. In this study, a new framework based on symmetric Bregman divergence, which incorporates k-nearest neighbor (kNN) classification in signal space, was proposed. The coordinates of the target were determined as a weighted combination of the nearest fingerprints using Jensen-Bregman divergences, which unify the squared Euclidean and Mahalanobis distances with information-theoretic Jensen-Shannon divergence measures. To validate our work, the performance of the proposed algorithm was compared with the probabilistic neural network and multivariate Kullback-Leibler divergence. The distance error for the developed algorithm was less than 1 m.

show abstract

A probability neural network-Jensen-Shannon divergence for a fingerprint based localization

Cited by 7 publications

References 13 publications

Convex Optimization via Symmetrical Hölder Divergence for a WLAN Indoor Positioning System

Convex Optimization via Symmetrical Hölder Divergence for a WLAN Indoor Positioning System

Neural Network Localization With TOA Measurements Based on Error Learning and Matching

Machine Learning Algorithm for Wireless Indoor Localization

Contact Info

Product

Resources

About