Topology Preserving Input Image for Convolutional Neural Network Based Indoor Localization

Laska, Marius; Blankenbach, Jörg

doi:10.1109/ipin51156.2021.9662471

Cited by 3 publications

(8 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Existing localization approaches have been explicitly designed to solve certain localization tasks that vary in the granularity of the estimated position. Solutions range from building/floor distinction [4], [15], to classifying predetermined zones/areas [16], [17], whereas some models are capable of estimating location coordinates (mostly 2D) [5] or predicting trajectories [18].…”

Section: Related Workmentioning

confidence: 99%

“…Multi-layer perceptron [19] either directly receives the fingerprint vectors or a low-dimensional embedding via stacked (denoising) autoencoder (S(D)AE) [4]- [7], [20]. Convolutional neural networks (CNN) can be directly applied on the fingerprint vector (1D-convolution) [21], on time-series fingerprints [21], or on 2D-input images, which are constructed from the original fingerprint vector [15], [22]. The 2D-input images are obtained by arbitrarily assigning fixed pixel position to APs [22] or as recently proposed by preserving the topology of the APs within the input image [15].…”

Section: Related Workmentioning

confidence: 99%

“…Convolutional neural networks (CNN) can be directly applied on the fingerprint vector (1D-convolution) [21], on time-series fingerprints [21], or on 2D-input images, which are constructed from the original fingerprint vector [15], [22]. The 2D-input images are obtained by arbitrarily assigning fixed pixel position to APs [22] or as recently proposed by preserving the topology of the APs within the input image [15]. The objective of proposed models is mostly a dedicated task, which manifests itself in the neural network head of the architecture.…”

Section: Related Workmentioning

confidence: 99%

Section: Backbone Architecturesmentioning

confidence: 99%

“…1) 2D-CNN with topology preserving input images 2D-CNN have been proven to work especially well for classification task within the indoor localization context [15]. The fingerprint vector is mapped to a 2D input image by either arbitrarily assigning fixed pixel positions to APs or as recently proposed by [15] via preserving the topology of the APs within the input image (TPAI). TPAI satisfy the prior of CNNs that close pixels are more related than distant ones and it was shown that it can boost the floor classification accuracy of a variety of CNN architectures [15].…”

Section: Backbone Architecturesmentioning

confidence: 99%

See 4 more Smart Citations

Multi-Task Neural Network for Position Estimation in Large-Scale Indoor Environments

Laska

Blankenbach

2022

IEEE Access

Self Cite

View full text Add to dashboard Cite

Pedestrian localization within large-scale multi-building/multi-floor indoor environments remains a challenging task. Fingerprinting-based approaches are particularly suited for such large-scale deployments due to their low requirements of hardware installments. Recently, the fingerprinting problem has been addressed by deep learning. Existing models are mostly task specific by providing floor classification or position estimation within a small area. A strategy to support localization within large-scale environments is to sequentially apply hierarchical models. This has several drawbacks including missing scalability and increased deployment complexity on smartphones. We propose a unifying approach based on training a single neural network that classifies the building/floor and predicts the position in a single forward-pass of the network. Our model classifies a grid cell and performs within grid cell regression, which solves the performance degradation of applying regression within large areas. To reduce the error in case of misclassified grid cells, we propose a novel technique called multi cell encoding learning (multi-CEL), where a model simultaneously learns several redundant position representations within an overlapping grid cell encoding. For three public WLAN fingerprinting datasets, we demonstrate that multi-CEL surpasses existing state-of-the-art multi-task learning neural networks and even outperforms regression neural networks explicitly trained for 2D-positioning by up to 17%.

show abstract