Jul 2026· Italian National Conference on Sensors· Vol 26, pp. 4281· 0 citations· 46 references
Medicine
TL;DR
A DOA estimation method that integrates a dynamic-pruning strategy with an adaptive subspace tracking mechanism that reduces computational complexity while enhancing algorithmic stability in complex and time-varying noise environments is proposed.
Abstract
In outdoor road environments, vehicle acoustic source direction-of-arrival (DOA) estimation is challenged by a low signal-to-noise ratio (SNR), dynamic-noise interference, and stringent real-time requirements. Under such conditions, conventional methods often struggle to achieve an effective balance among estimation accuracy, computational efficiency, and robustness against noise. To address this issue, this paper proposes a DOA estimation method that integrates a dynamic-pruning strategy with an adaptive subspace tracking mechanism. The proposed approach reduces computational complexity while enhancing algorithmic stability in complex and time-varying noise environments. Extensive experiments conducted on simulated data, the LOCATA dataset, and real-world outdoor road measurements demonstrate that the proposed method achieves comparable or superior DOA accuracy relative to conventional approaches, while significantly reducing computational cost. Furthermore, it exhibits stronger stability and robustness in real-world static and dynamic vehicle localization scenarios. Our method achieves a more favorable trade-off among multiple performance metrics. The results show that this method has good engineering application potential in complex outdoor environments, and can provide a practical solution for real-world vehicle monitoring.
This paper presents a computationally efficient deep learning framework for accurate direction-of-arrival (DoA) estimation in portable radar applications. Leveraging a MobileNet architecture, the proposed model directly processes raw in-phase and quadrature-phase (IQ) data, enabling more effective learning of both spatial and temporal features. This direct input approach enhances DoA estimation accuracy, particularly under challenging conditions such as low signal-to-noise ratio (SNR) and limited snapshot scenarios. A unified training strategy is adopted for both single-source and multi-source target detection, ensuring consistency and robustness. Comprehensive simulation experiments demonstrate the proposed model’s competitive and robust performance across various conditions, including different SNR levels, closely spaced targets, and random off-grid angles. It also shows that our method achieves performance comparable to or better than recent deep learning approaches in several challenging scenarios, establishing its potential for resource-constrained environments where only low snapshot data are available. The proposed IQ-MobNet DoA estimation model achieves this competitive performance with substantially lower computational complexity, requiring only 0.24 million parameters and 0.42 million Floating Point Operations (FLOPs), representing a reduction of over 96% compared to the recent neural network models. To ensure practical applicability, the proposed IQ-MobNet framework is validated using real-world measured data, confirming its robustness beyond simulated environments.
Neeraja P. Kovilakam, Bindiya T. Sambasivan, Raghu C. Variyam· Electronics· 0 citations
In Indoor Positioning Systems, continuous high-frequency Wi-Fi fingerprint scanning is a primary cause of battery depletion in resource-constrained IoT mobile nodes. While simply extending the scan interval can reduce energy consumption, it often fails to effectively compensate for the resulting data sparsity and positional uncertainty. This often leads to severe tracking lag and path shrinkage in dynamic environments. To overcome these limitations, we propose a joint optimization framework that combines a tabular Q-learning model utilizing spatiotemporal features with a step-coupled Dynamic Kalman Filter(DKF). The proposed system perceives the user’s dynamic state in real time to determine the optimal sampling interval, while simultaneously adjusting the process noise covariance (Q) of the filter to proactively mitigate tracking errors. Experimental results in a real-world ESP32 testbed demonstrate that our approach achieves an 82.5% reduction in data transmission compared to a high-frequency 30ms baseline, while maintaining a highly accurate Root Mean Square Error (RMSE) of 1.31 m. Consequently, the proposed method successfully approaches a Pareto-optimal tradeoff between energy efficiency and localization accuracy.
Hyeongmin Kim, Young-Bae Ko· International Conference on...· 0 citations
Accurate and stable position and orientation estimation in indoor environments is critical for many real-world applications, yet remains an open challenge without a universally optimal solution. In this work, we propose A-LOAF, a novel acoustic-based system that achieves real-time and precise indoor localization and orientation estimation using commercial speakers and a compact microphone array. The system employs spatially distributed acoustic sources to emit inaudible near-ultrasonic linear frequency-modulated (LFM) signals, which are captured by a microphone array mounted on the mobile device to be localized. An angle-of-arrival (AOA) estimation algorithm is employed to determine source directions in real time, and device pose is jointly estimated by minimizing an AOA-related objective function. To further enhance localization accuracy, frequency-modulated continuous wave (FMCW) ranging is integrated to compute the relative displacement between keyframes as interframe constraints, supported by a dedicated keyframe selection strategy. The FMCW-based constraint is then incorporated as a penalty term into the AOA-based optimization framework, enabling robust and high-precision pose estimation. Experimental results in complex real-world indoor environments demonstrate that A-LOAF achieves a median localization error of 0.1171 m and a median orientation error of 0.60°, with an average computational time of 0.0884 s per frame, confirming its effectiveness and practical applicability. To benefit the community, we make the source code publicly available at https://github.com/zjuersdsd/ALOAF.git
Guangyao Liu, Yuzhang Xi, Weimeng Cui et al.· IEEE Transactions on Instrum...· 0 citations
Urban environments pose significant challenges for real-time target localization and tracking due to reverberation, occlusions, and frequent GPS degradation. Visual-only systems often suffer from delayed target acquisition when line-of-sight is obstructed, while acoustic sensing provides passive, low-latency directional cues but lacks spatial resolution for persistent tracking. This work presents a real-time, edge-deployable multimodal framework that integrates acoustic sound source localization (SSL) with visual object detection and tracking to close this initialization gap. Acoustic cues are employed to actively steer a servo-mounted camera, reducing the visual search space and accelerating time-to-acquisition. Visual detections are performed using YOLOv8, while temporal consistency is maintained via BoT-SORT with camera motion compensation. An extended Kalman Filter (EKF) fuses acoustic DoA estimates and visual bounding box measurements to enable robust tracking under nonlinear motion and temporary visual occlusions.
Chidera Igwebuike, Akemi Vasquez, Wagdy H. Mahmoud et al.· Journal of the Acoustical So...· 0 citations
A feedback-guided DNN-based controller fusion framework for robust fixed-parameter ANC that combines a causal WaveNet controller with a feedback-guided mixture-of-experts (MoE) module, where a gating network estimates the weights of multiple pre-trained FIR experts according to the current acoustic condition.
Lu Bai, Yiming He, Xiaofeng Nan et al.· 0 citations
Ultra-wideband positioning accuracy in indoor spaces is often limited by the combined effects of path loss and transient non-line-of-sight (NLOS) interference. While pre-calibration can mitigate base errors, it remains ineffective against dynamic environmental changes and random occlusions. To bridge this gap, this paper proposes a two-stage adaptive disturbance-aware architecture (TADA). The framework integrates a signal-layer adaptive path-loss compensation module with dynamic forgetting factor (APLC-DFF) and a state-layer NLOS-resilient composite filter (NRCF). By utilizing a cross-layer residual feedback mechanism, TADA enables multi-timescale error separation: APLC-DFF adaptively self-calibrates slow-varying systematic biases, while NRCF actively compensates for transient NLOS disturbances via a disturbance observer. Experimental results demonstrate that TADA achieves a 44.7% improvement in RMSE compared to conventional EKF and significantly outperforms the state-of-the-art (SOTA) GMC-EKF algorithm. This paradigm provides a robust and high-precision solution for indoor positioning.
Zhiran Shi, Le Gao, Ruoyu Chen et al.· Measurement science and tech...· 0 citations