Abstract

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

 

Quick Read (beta)