基于多可穿戴传感器的用户人体复杂行为主动识别
Active Identification of Complex Behaviors of the Human Body by Users Based on Multi-Wearable Sensors
叶彩仙 1胥立军2
作者信息
- 1. 广州新华学院信息与智能工程学院,广东东莞 523133
- 2. 广州新华学院人工智能与数据科学系,广东东莞 523133
- 折叠
摘要
文章旨在解决人体复杂行为的主动识别问题,通过对多可穿戴式传感器数据的融合,提高识别准确性并降低识别耗时.为此,设计了一种基于多可穿戴式传感器数据融合的方法.首先,根据用户生理、生化、行为指标设计采集流程;其次,采集用户指标传感数据,将其融合,结果作为卷积神经网络(Convolutional Neural Network,CNN)的输入,构建用户虚拟现实场景;在此场景中,利用自相关系数和循环神经网络提取可识别特征点,并通过全连接层和Softmax函数实现用户行为的分类和识别.实验结果表明,该文方法相较于传统方法,识别耗时更低且准确率更高,F1-score达到了0.986.
Abstract
The article aims to solve the problem of active recognition of complex human behavior,improving recognition accuracy and reducing recognition time by fusing data from multiple wearable sensors.A method based on data fusion of multiple wearable sensors was designed for the purpose.Firstly,design a collection process based on user physiological,biochemical,and behavioral indicators.Secondly,collect user indicator sensing data and use the fusion result as input for Convolutional Neural Network(CNN)to construct a user virtual reality scene.In the scenario,identifiable feature points are extracted using autocorrelation coefficients and recurrent neural networks,and user behavior classification and recognition are achieved through fully connected layers and Softmax functions.The experimental results show that compared to traditional methods,the proposed method has lower recognition time and higher accuracy,with an Fl-score of 0.986.
关键词
多可穿戴式/传感器数据融合/用户人体行为/复杂行为/主动识别/识别方法Key words
multi-wearable/sensor data fusion/user human behavior/complex behavior/active identification/identification method引用本文复制引用
出版年
2024
NETL
NSTL
万方数据