Problem
Research questions and friction points this paper is trying to address.
Signal Instability
Energy Supply Uncertainty
Sound Source Localization
LOCUS: LOcalization with Channel Uncertainty and Sporadic Energy
🤖 AI Summary
In batteryless acoustic source localization, energy harvesting induces stochastic multi-channel data missing, severely degrading conventional direction-of-arrival (DoA) estimation performance under incomplete observations. To address this, we propose an information-entropy-driven framework for quantified missing-data identification and conditional imputation. Our method innovatively integrates three core components: information-weighted beamforming, latent feature synthesis, and guided replacement—enabling high-fidelity reconstruction of missing-channel features. By synergistically combining deep learning, conditional variational imputation, and robust multi-channel signal processing, the approach achieves up to 36.91% reduction in DoA estimation error on DCASE and LargeSet benchmarks. Under realistic intermittent power supply conditions, it delivers consistent improvements of 25.87–59.46%. Furthermore, we release a 50-hour multichannel self-supervised learning (SSL) dataset to support research in energy-constrained acoustic sensing.
📝 Abstract
Accurate sound source localization (SSL) requires consistent multichannel data for reliable degree of arrival (DoA) estimation. However, intermittently powered batteryless systems often suffer from incomplete sensor data due to the stochastic nature of energy harvesting. Existing methods struggle with missing channels, leading to significant performance degradation. In this paper, we propose $ extit{LOCUS}$, a novel deep learning-based system designed to recover corrupted features for SSL in batteryless systems. $ extit{LOCUS}$ addresses missing data by leveraging information entropy estimation and conditional interpolation, combining three modules: (1) Information-Weighted Focus (InFo), which identifies and quantifies corrupted data elements, (2) Latent Feature Synthesizer (LaFS), which synthesizes missing features, and (3) Guided Replacement (GRep), which intelligently replaces missing elements while preserving valid data. We demonstrate significant performance improvements using two datasets: DCASE and LargeSet, where $ extit{LOCUS}$ achieves up to $36.91%$ lower DoA error compared to existing methods. Real-world evaluations across three environments with intermittent power sources show a $25.87-59.46%$ improvement in performance when channels are stochastically missing. Additionally, we release a 50-hour multichannel dataset to support further research in SSL.
Innovation
Methods, ideas, or system contributions that make the work stand out.
LOCUS system
Deep Learning for Acoustic Source Localization
Information Quality and Replacement
🔎 Similar Papers
No similar papers found.
