Mingxing Nie

Class-incremental learning (CIL) for time series data faces critical challenges in balancing stability against catastrophic forgetting and plasticity for new knowledge acquisition, particularly under real-world constraints where historical data access is restricted. While pre-trained models (PTMs) have shown promise in CIL for vision and NLP domains, their potential in time series class-incremental learning (TSCIL) remains underexplored due to the scarcity of large-scale time series pre-trained models. Prompted by the recent emergence of large-scale pre-trained models (PTMs) for time series data, we present the first exploration of PTM-based Time Series Class-Incremental Learning (TSCIL). Our approach leverages frozen PTM backbones coupled with incrementally tuning the shared adapter, preserving generalization capabilities while mitigating feature drift through knowledge distillation. Furthermore, we introduce a Feature Drift Compensation Network (DCN), designed with a novel two-stage training strategy to precisely model feature space transformations across incremental tasks. This allows for accurate projection of old class prototypes into the new feature space. By employing DCN-corrected prototypes, we effectively enhance the unified classifier retraining, mitigating model feature drift and alleviating catastrophic forgetting. Extensive experiments on five real-world datasets demonstrate state-of-the-art performance, with our method yielding final accuracy gains of 1.4%-6.1% across all datasets compared to existing PTM-based approaches. Our work establishes a new paradigm for TSCIL, providing insights into stability-plasticity optimization for continual learning systems.

Traditional deep learning methods struggle to simultaneously segment, recognize, and forecast human activities from sensor data. This limits their usefulness in many fields, such as healthcare and assisted living, where real-time understanding of ongoing and upcoming activities is crucial. This article introduces P2LHAP, a novel Patch-to-Label Seq2Seq framework that tackles all three tasks in an efficient single-task model. P2LHAP divides sensor data streams into a sequence of “patches,” served as input tokens, and outputs a sequence of patch-level activity labels, including the predicted future activities. A unique smoothing technique based on surrounding patch labels, is proposed to identify activity boundaries accurately. Additionally, P2LHAP learns patch-level representation by sensor signal channel-independent Transformer encoders and decoders. All channels share embedding and Transformer weights across all sequences. Evaluated on the three public datasets, P2LHAP significantly outperforms the state-of-the-art in all three tasks, demonstrating its effectiveness and potential for real-world applications.

Interest in anti-UAV detection systems has increased due to growing concerns about the security and privacy issues associated with unmanned aerial vehicles (UAVs). Achieving real-time detection with high accuracy, while accommodating the limited resources of edge-computing devices poses a significant challenge for anti-UAV detection. Existing deep learning-based models for anti-UAV detection often cannot balance accuracy, processing speed, model size, and computational efficiency. To address these limitations, a lightweight and efficient anti-UAV detection model, DRBD-YOLOv8, is proposed in this paper. The model integrates several innovations, including the application of a Re-parameterization Cross-Stage Efficient Layered Attention Network (RCELAN) and a Bidirectional Feature Pyramid Network (BiFPN), to enhance feature processing capabilities while maintaining a lightweight design. Furthermore, DN-ShapeIoU, a novel loss function, has been established to enhance detection accuracy, and depthwise separable convolutions have been included to decrease computational complexity. The experimental results showed that the proposed model outperformed YOLOV8n in terms of mAP50, mAP95, precision, and FPS while reducing GFLOPs and parameter count. The DRBD-YOLOv8 model is almost half the size of the YOLOv8n model, measuring 3.25 M. Its small size, fast speed, and high accuracy combine to provide a lightweight, accurate device that is excellent for real-time anti-UAV detection on edge-computing devices.

This paper introduces a novel approach for enhancing human activity recognition through the integration of LoRa wireless RF signal preprocessing and deep learning. We tackle the challenge of extracting features from intricate LoRa signals by scrutinizing the unique propagation process of linearly modulated LoRa signals—a critical aspect for effective feature extraction. Our preprocessing technique involves converting intricate data into real numbers, utilizing Short-Time Fourier Transform (STFT) to generate spectrograms, and incorporating differential signal processing (DSP) techniques to augment activity recognition accuracy. Additionally, we employ frequency-to-image conversion for the purpose of intuitive interpretation. In comprehensive experiments covering activity classification, identity recognition, room identification, and presence detection, our carefully selected deep learning models exhibit outstanding accuracy. Notably, ConvNext attains 96.7% accuracy in activity classification, 97.9% in identity recognition, and 97.3% in room identification. The Vision TF model excels with 98.5% accuracy in presence detection. Through leveraging LoRa signal characteristics and sophisticated preprocessing techniques, our transformative approach significantly enhances feature extraction, ensuring heightened accuracy and reliability in human activity recognition.

The rapid expansion of nuclear technology across various sectors due to global economic growth has led to a substantial rise in the transportation of radioactive materials. The International Atomic Energy Agency (IAEA) estimates that approximately 20 million shipments of radioactive materials occur annually. In this context, ensuring the safety and security of radioactive material transportation is of significant importance. IAEA’s “Security of Radioactive Materials in Transport” (Nuclear Security Series No. 9-G) mandates that an effective transport security system should provide immediate detection of any unauthorized removal of the packages. In the present study, an innovative Adam-optimized BP neural network model is developed for detecting unauthorized movements of radioactive material packages. To analyze the performance of the proposed algorithm, numerous experiments were conducted. The results demonstrate that the proposed method achieves a 99.17% accuracy rate in detecting unauthorized movements of radioactive materials, with a missed alarm rate of 0.72% and a false alarm rate of 0.1%. This method also enables real-time detection of unauthorized removal of radioactive materials and effectively enhances the security of radioactive materials during transport to reduce the risks of theft, loss, diversion, or sabotage.

Human Activity Recognition (HAR) systems have made significant progress in recognizing and classifying human activities using sensor data from a variety of sensors. Nevertheless, they have struggled to automatically discover novel activity classes within massive amounts of unlabeled sensor data without external supervision. This restricts their ability to classify new activities of unlabeled sensor data in real-world deployments where fully supervised settings are not applicable. To address this limitation, this paper presents the Novel Class Discovery (NCD) problem, which aims to classify new class activities of unlabeled sensor data by fully utilizing existing activities of labeled data. To address this problem, we propose a new end-to-end framework called More Reliable Neighborhood Contrastive Learning (MRNCL), which is a variant of the Neighborhood Contrastive Learning (NCL) framework commonly used in visual domain. Compared to NCL, our proposed MRNCL framework is more lightweight and introduces an effective similarity measure that can find more reliable k-nearest neighbors of an unlabeled query sample in the embedding space. These neighbors contribute to contrastive learning to facilitate the model. Extensive experiments on three public sensor datasets demonstrate that the proposed model outperforms existing methods in the NCD task in sensor-based HAR, as indicated by the fact that our model performs better in clustering performance of new activity class instances.

Feedrate plays a crucial role in determining the machining quality, tool life, and machining time. Thus, this research aimed to improve the accuracy of NURBS interpolator systems by minimizing feedrate fluctuations during CNC machining. Previous studies have proposed various methods to minimize these fluctuations. However, these methods often require complex calculations and are not suitable for real-time and high-precision machining applications. Given the sensitivity of the curvature-sensitive region to feedrate variations, this paper proposed a two-level parameter compensation method to eliminate the feedrate fluctuation. First, in order to address federate fluctuations in non-curvature sensitive areas with low computational costs, we employed the first-level parameter compensation (FLPC) using the Taylor series expansion method. This compensation allows us to achieve a chord trajectory for the new interpolation point that matches the original arc trajectory. Second, even in curvature-sensitive areas, feedrate fluctuations can still occur because of truncation errors in the first-level parameter compensation. To address this, we employed the Secant-based method for second-level parameter compensation (SLPC), which does not require derivative calculations and can regulate feedrate fluctuation within the fluctuation tolerance. Finally, we applied the proposed method to the simulation of butterfly-shaped NURBS curves. These simulations demonstrated that our method achieved maximum feedrate fluctuation rates below 0.01% with an average computational time of 360 us, which is sufficient for high-precision and real-time machining. Additionally, our method outperformed four other feedrate fluctuation elimination methods, highlighting its feasibility and effectiveness.