Unlimited Sampling of Multiband Signals: Single-Channel Acquisition and Recovery

2025-10-30

AI Summary

This work addresses the sub-Nyquist reconstruction of multiband signals under single-channel modulo folding sampling. We propose a low-complexity, single-channel acquisition and reconstruction framework grounded in an unbounded sensing paradigm. Departing from conventional oversampling requirements, our approach tightens the bandpass sampling theorem by deeply integrating modulo folding sampling with sub-Nyquist reconstruction algorithms, while incorporating hardware-aware signal processing. Unlike traditional methods, the proposed scheme overcomes the dynamic range limitation inherent in modulo sampling, enabling high-fidelity recovery of up to six spectral bands using only a single hardware channel. Experimental results demonstrate a 13× improvement in effective dynamic range and a substantial reduction in sampling rate, thereby significantly alleviating hardware resource constraints.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

In this paper, we address the problem of reconstructing multiband signals from modulo-folded, pointwise samples within the Unlimited Sensing Framework (USF). Focusing on a low-complexity, single-channel acquisition setup, we establish recovery guarantees demonstrating that sub-Nyquist sampling is achievable under the USF paradigm. In doing so, we also tighten the previous sampling theorem for bandpass signals. Our recovery algorithm demonstrates up to a 13x dynamic range improvement in hardware experiments with up to 6 spectral bands. These results enable practical high-dynamic-range multiband acquisition in scenarios previously limited by dynamic range and excessive oversampling.

Problem

• Reconstructing multiband signals from modulo-folded samples
• Achieving sub-Nyquist sampling with single-channel acquisition
• Improving dynamic range for practical multiband signal acquisition

Innovation

• Single-channel modulo-folded sampling for multiband signals
• Sub-Nyquist sampling guarantees under Unlimited Sensing Framework
• Hardware experiments show 13x dynamic range improvement