软件导刊2025，Vol.24Issue(8)：38-42,5.DOI:10.11907/rjdk.241526
基于可分离卷积的人脸视频心率检测方法
Separable Convolutional-based Heart Rate Detection Method for Face Video
摘要
Abstract
Heart rate estimation plays an important role in analyzing human health status such as heart.In order to address the problem of in-adequate face feature modeling in remote contactless heart rate detection,the separable convolutional based heart rate detection for face video(FVSC-HR)is proposed.First,the convolutional gating unit is incorporated to obtain fine-grained features of neighboring frames.Second,a multi-head attention mechanism based on temporal differential convolution is used to refine the local spatio-temporal representation anti-inter-ference capability.Then,the heart rate is output through the spatio-temporal feed-forward network.The model is trained and tested on UBFC-rPPG and UBFC-phys datasets.Tested on the UBFC-rPPG and UBFC-phys datasets,the MAE of the FVSC-HR method on the UBFC-rPPG dataset is 3.4,and the ρ-evaluation index is 0.8;on the UBFC-phys dataset,the MAE of the FVSC-HR method is 7.3;the FVSC-HR outper-forms the six methods compared,and it can be used to provide intelligent health monitoring and telemedicine to provide heart rate detection technology.关键词
心率检测/人脸视频/可分离卷积/差分注意力机制/卷积门控单元Key words
heart rate detection/face video/separable convolution/differential attention mechanism/convolutional gating unit分类
信息技术与安全科学引用本文复制引用
王猛,杨观赐..基于可分离卷积的人脸视频心率检测方法[J].软件导刊,2025,24(8):38-42,5.基金项目
国家自然科学基金项目(62163007,62373116) （62163007,62373116）
贵州省科技计划项目(黔科合平台人才[2020]6007-2,黔科合支撑[2023]一般118) （黔科合平台人才[2020]6007-2,黔科合支撑[2023]一般118）
