Spectral-Temporal Fusion Representation for Person-in-Bed Detection

TL;DR

This paper introduces a spectral-temporal fusion feature representation combined with mixup data augmentation and IoU loss, achieving top performance in accelerometer-based person-in-bed detection challenges.

Contribution

It presents a novel spectral-temporal fusion approach with data augmentation and IoU loss for improved bed occupancy detection accuracy.

Findings

Achieved 100% detection score in in-bed track.

Secured third place with 95.55% in streaming detection.

Outperformed existing methods in the challenge.

Abstract

This study is based on the ICASSP 2025 Signal Processing Grand Challenge's Accelerometer-Based Person-in-Bed Detection Challenge, which aims to determine bed occupancy using accelerometer signals. The task is divided into two tracks: "in bed" and "not in bed" segmented detection, and streaming detection, facing challenges such as individual differences, posture variations, and external disturbances. We propose a spectral-temporal fusion-based feature representation method with mixup data augmentation, and adopt Intersection over Union (IoU) loss to optimize detection accuracy. In the two tracks, our method achieved outstanding results of 100.00% and 95.55% in detection scores, securing first place and third place, respectively.