2 February 2023 in IEEE Communications Letters
Although fingerprint-based methods could achieve high location accuracy for device-free indoor localization, it requires cost-expensive massive labeling. In order to fully exploit the previously collected CSI fingerprints for various indoor localization tasks, this letter proposes a meta-learning approach for indoor localization, with which an excellent initialization solution for a new localization task is resorted and further training with a few labeled samples results in sound performance improvement over the state-of-the-art deep learning (DL) approach. By recognizing the unequal contribution of an individual training task to the target task, a novel weighted loss is introduced under the framework of meta-learning. Experiments show that the proposed meta-learning approach can gain improvement of 20% Root-Mean-Square-Error (RMSE) enhancement over the DL approach.
29 September 2021 in IEEE Communications Letters
Deep learning models are widely used in the field of Device-Free Localization (DFL), and their performances rely heavily on the labeled dataset. In this letter, a novel data preprocessing method is proposed for exploiting the full potential of a collected dataset in DFL. Firstly, we propose to employ a pix-level multimodal representation of radio images for a fusion of both amplitude and phase features from Channel State Information (CSI). Then, a novel Conditional Generative Adversarial Network with Auxiliary Classifier (AC-GAN) model is used to generate artificial samples for further expanding the collected dataset. We further employ a regression formulation to train the Convolutional Neural Networks (CNNs) for positioning. The experimental results show that using the proposed method gains approximately 18% Root-Mean-Square-Error (RMSE) improvement over the CSI amplitude-only method and 22% over the CSI phase-only method.
14 March 2022 in IEEE Wireless Communications Letters
Fingerprint-based methods have become the mainstream for Device-Free Localization (DFL), and their performances depend heavily on the labeled dataset. In this letter, a novel Channel State Information (CSI) fingerprinting scheme is proposed for fully-exploiting the power of a given dataset. First, a phase calibration approach is applied to eliminate the phase offset due to the imperfection of synchronization and further make the collected phases from the different reference points distinguishable. Then, a Structural Similarity-based (SSIM-based) augmentation method is used to generate artificial samples for expanding the given dataset. Compared to existing GAN-based methods, it is very simple and computationally more efficient. Extensive experiments demonstrate the effectiveness of the proposed method. In comparison with the initial dataset and the AC-GAN based method, using the proposed method can gain approximately 13% and 8% Root-Mean-Square-Error (RMSE) enhancement, respectively.
28 October 2022 in IEEE Communications Letters
Recently, belief propagation list (BPL) decoding of polar codes has been proposed for achieving state-of-the-art performance. By representing a polar code factor graph permutation with an equivalent bit index permutation, a single decoder can be applied on different permutated factor graphs, which is hardware-friendly. Nevertheless, the design of permutations under BPL decoding turns out to be cumbersome and highly relevant with the code structure. In this work, we propose a two-step approach for the joint design of permutations and polar code under BPL decoding. Firstly, an empirical approach was proposed for selection of good initial permutations under BPL decoding. Then, the joint optimization of permutations and polar code is implemented based on genetic algorithm. With the aid of an outer cyclic redundancy check (CRC) code, the proposed BPL scheme is shown to perform very close to CRC aided successive cancellation list (CA-SCL) decoding at practical SNRs.
4 July 2023 in IEEE Communications Letters
Recently, belief propagation list (BPL) decoding has been proposed to achieve the excellent parallel decoding performance of polar codes with high throughput and low latency. However, the performance gap between BPL decoding and state-of-the-art successive cancellation list (SCL) decoding is still significant at practical SNRs. In this letter, we propose an efficient BPL flip (EBPLF) decoding algorithm for filling this gap. Firstly, an efficient bit-flipping strategy under BPL decoding is proposed to allow multiple bits being flipped more efficiently than the existing bit-flipping strategy. Then, by exploiting the decoding diversity and reliability of each bit over the multiple factor graphs within BPL decoding, an enhanced flip set is constructed to identify the error-prone patterns in an efficient way. Extensive simulation results show that EBPLF decoding outperforms the existing BPLF scheme and approaches the performance of adaptive SCL decoding at practical SNRs with much lower average latency.