MSCGN: Multiscale complementary gating network for time series classification

doi:10.1016/j.bspc.2025.108563

MSCGN: Multiscale complementary gating network for time series classification

2025-09-06

0

PRE

AI

DOI：10.1016/j.bspc.2025.108563

Original

Original request for help

Save

Abstract

En 中文

The classification of time-series data remains a formidable challenge due to its inherent diversity across various domains and tasks. In this study, we propose a novel Multiscale Complementary Gating Network (MSCGN) that exhibits remarkable generalization capabilities and robustness across diverse fields. The model employs a multi-scale residual subtraction strategy to extract a more comprehensive and robust feature representation by integrating information from multi-granularity, allowing the network to perceive coarse-grained and fine-grained difference information. The model displays excellent accuracy and generalization on six time-series classification tasks, including atrial fibrillation recognition, human activity recognition, sleep monitoring, epilepsy recognition, heart sound detection, and gesture recognition, which surpasses benchmark models across different time-series domains. The proposed MSCGN offers a powerful and flexible tool for time-series classification, addressing the complexity and variability of time-series data.

MSCGN: Multiscale complementary gating network for time series classification

Abstract

Journal

Organization