/Tensor_Sensing_for_RF_Tomography

RF Tomography imaging

Primary LanguageMATLABMIT LicenseMIT

Tensor_Sensing_for_RF_Tomography

  • [1] T. Deng, F. Qian, X.-Y. Liu, M. Zhang, A. Walid. Tensor sensing for RF tomographic imaging. IEEE ICME, 2018.

Usage

Basicly, we select IKEA 3D dataset as an example. IKEA 3D

Run ./run_TS.m to recover the unkowm tensor, you can input dct for DCT transform or fft for FFT transform. You can change the parameters such as the size of the unknown tenor, the sampling rates, the iteration times in ./TS_example.m. The main steps of the algorithm Alt-Min is in ./TS.m. And the ./toolbox contains the dependent functions.

Experiment

We compare the proposed algorithm Alt-Min with tensor-based compressed sensing [2] on 50 IKEA 3D models. Each 3D model is used to generate one ground truth tensor of size $60\times 60\times 15$, which is placed in the middle of the tensor and occupies a part of the space.

For the simulations of the wireless channel, the space of interest is divided into a set of three-dimensional voxels, and a set of RF signal nodes are uniformly deployed around the space, forming a complete tomography network. Any pair of nodes can establish a unique wireless link, and the path loss on a wireless link has three contributions: (1) Large-scale path loss due to distance; (2) Shadowing loss due to obstructions; and (3) Non-shadowing loss due to multipath [2,3]. The relevant code is in ./toolbox/generate_sampling_tensor.m and ./toolbox/one_link.m.

Fig.1 (a) and (c) are the 3D visualizations of two IKEA models, (b) and (d) are the corresponding recovery results.
Fig.2 (a) is RSEs vs sampling rates; (b) is Alt-Min with FFT; (c) is Alt-Min with DCT.
  • [2] Matsuda, Takahiro, et al. "Multi-dimensional wireless tomography using tensor-based compressed sensing." Wireless Personal Communications 96.3 (2017): 3361-3384.
  • [3] J. Wilson and N. Patwari, "Radio tomographic imaging with wireless networks," IEEE Transactions on Mobile Computing, vol. 9, no. 5, pp.621–632, 2010.