Bing Luo

Three-station Time-of-Arrival (TOA) positioning technology holds significant application value in electronic reconnaissance and unmanned aerial vehicle tracking. However, TOA measurements in real-world environments are often affected by non-line-of-sight propagation and synchronization errors, making it challenging to accurately obtain the system's noise statistical characteristics. Traditional Kalman filtering (KF) heavily relies on precise system models and noise parameters; when the pre-set noise covariance mismatches the actual environment, positioning accuracy significantly degrades. To address this issue, this paper proposes a deep Kalman network (KalmanNet)-assisted positioning algorithm based on unsupervised learning.The basic idea is that: encapsulating environmental priors and noise distributions into GRU model parameters, the Kalman Gain could be calculated based on observation variables and dynamically updated state variables. This algorithm adopts a hybrid architecture combining model-driven and data-driven approaches. To avoid performance degradation caused by the inability to collect samples or discrepancies between the training data and the real-world environment, we employ an unsupervised training method. Unlike supervised learning that relies on true state values, this paper designs an unsupervised loss function based on observation innovations, optimizing network parameters through backpropagation. Experimental results demonstrate that, even with only observation data and no true system states, the proposed algorithm can predict target states in real time. Its positioning accuracy outperforms traditional Kalman filtering and exhibits strong robustness across varying noise levels.

With the rise of smart home applications, WiFi-based posture estimation has become a research focus. Existing methods often emphasize end-to-end feature extraction while neglecting high-resolution intermediate spatial features. We observe that horizontal and vertical antennas capture electromagnetic wave distributions in different directions, while multiple subcarrier frequencies inherently enhance multipath resolution. By analyzing signal correlations among subcarriers, we achieve equivalence to virtual antennas, enabling the generation of high-resolution spatial correlation maps. Using a cascaded network with UNet, we extract human spatial transformation features and a steering vector to obtain a spatial spectrum, mapping keypoints with high precision. Tests in through-wall environments achieve an average mPCK of 0.90 at 20 pixels and 0.98 at 50 pixels, significantly improving detection accuracy.

Most existing image deblurring methods construct statistical prior to describe the difference between blur and clear image. They discard the position information and ignore pixel feature changing in deblurring, which results in inferior restoration performance for images unsatisfying corresponding assumptions. Intuitively, fuzziness of pixel belonging to different image regions will reduce along with image deblurring. This phenomenon could intrinsically describe the pixel characteristic. To this end, we analyze fuzziness of pixels and objects in a blurry image, and utilize the similarity between two fuzzy objects on image pixels to depict the blur degree of an image, which is inspired by overlap functions and overlap indices. To minimize the similarity between fuzzy objects, we introduce the non-parameters model to construct an integer programming problem. Energy minimization could significantly reduce the similarity between two fuzzy objects. Experimental results show that the proposed method can achieve better performance than the state-of-the-art blind deblurring methods on benchmark datasets and natural images.

Referring image segmentation aims to segment object in an image based on a referring expression. Its difficulty lies in aligning expression semantics with visual instances. The existing methods based on semantic reasoning are limited by the performance of external syntax parser and do not explicitly explore the relationships between visual instances. This article proposes an end-to-end method for referring image segmentation by aligning ’linguistic relationship’ with ’visual relationships’. This method does not rely on external syntax parser for expression parsing. In this paper, the expression is adaptively and structurally parsed into three components: ’subject’, ’object’, and ’linguistic relationship’ by the Semantic Component Parser (SCP) in a learnable manner. Instances Activation Map Module (IAM) locates multiple visual instances based on the subject and object. In addition, the Relationship Based Visual Localization Module (RBVL) firstly enables each instance of the image to learn global knowledge, then decodes the visual relationships between these visual instances, and finally aligns the visual relationships with the linguistic relationships to further accurately locate the target object. The experimental results show that the proposed method improves performance by 4– 9% compared with baseline method on multiple referring image segmentation datasets.

Ringing artifacts is a critical problem in image deblurring. Existing methods based on image sparsity cannot efficiently recognize the plausive edge and result in significant ringing artifacts. Those methods mostly consider the statistic prior for blur and clear images but ignore continuity of natural image in feature space. While, image intrinsic continuity could complement the sparsity information for image statistic prior. Intuitively, ringing effects should form the shake in feature space and break the piece-wise smoothness of natural image. To this end, we propose a minimizing manifold area prior (MAP) method to reduce the ringing artifacts. We first embed a two-dimension manifold into the image high-dimension feature space. Then the feature difference between blur and clear images could be described by manifold area, which could be represented via local Euclide-space. This is beneficial to proposing a fast and effective algorithm to calculate manifold area, which significantly improve the computational complexity. Experiments validate that our method is superior than state-of-the-arts in performance and time consuming, especially for ringing artifacts.

With the advent of the information age, there are more and more text data on the Internet. As the most widely distributed information carrier with the largest amount of data, it is particularly important to use text classification technology to organize and manage massive data scientifically. In this paper, a semi-supervised ensemble learning algorithm Heterogeneous-training is proposed and applied to the field of text classification. Based on the Tri-training algorithm, the Heterogeneous-training algorithm improves the traditional Tri-training algorithm by using different classifiers, dynamically updating the probability threshold and adaptively editing data. A large number of experiments show that our method always outperforms Tri-training algorithm in text classification on benchmark text data sets.