Dywave: Event-Aligned Dynamic Tokenization for Heterogeneous IoT Sensing Signal

2026-05-13

AI Summary

Heterogeneous IoT sensing signals exhibit non-stationarity and multi-scale characteristics, rendering traditional tokenization methods ineffective at capturing their temporal structure and semantic events. To address this challenge, this work proposes Dywave, a dynamic tokenization framework that introduces event-aligned dynamic segmentation into IoT signal processing for the first time. Dywave leverages wavelet-based hierarchical decomposition to adaptively identify semantic event boundaries and compress redundant segments, thereby constructing compact input representations while preserving temporal consistency. The method seamlessly integrates with mainstream sequence models and achieves up to a 12% accuracy improvement and as much as a 75% reduction in input length across five real-world IoT datasets. Furthermore, it significantly enhances robustness under domain shifts and adaptability to varying sequence lengths.

Abstract

Internet of Things (IoT) systems continuously collect heterogeneous sensing signals from ubiquitous sensors to support intelligent applications such as human activity analysis, emotion monitoring, and environmental perception. These signals are inherently non-stationary and multi-scale, posing unique challenges for standard tokenization techniques. This paper proposes Dywave, a dynamic tokenization framework for IoT sensing signals that constructs compact input representations aligned with intrinsic temporal structures and underlying physical events. Dywave leverages wavelet-based hierarchical decomposition, identifies meaningful temporal boundaries corresponding to underlying semantic events, and adaptively compresses redundant intervals while preserving temporal coherence. Extensive evaluations on five real-world IoT sensing datasets across activity recognition, stress assessment, and nearby object detection demonstrate that Dywave outperforms state-of-the-art methods by up to 12% in accuracy, while improving computational efficiency by reducing input token lengths by up to 75% across mainstream sequence models. Moreover, Dywave exhibits improved robustness to domain shifts and varying sequence lengths.

Problem

• IoT sensing signals
• non-stationary
• multi-scale
• tokenization
• temporal structure

Innovation

• dynamic tokenization
• wavelet decomposition
• event-aligned representation
• IoT sensing signals
• temporal coherence