基于神经网络架构搜索技术的人体行为识别方法研究

  • 打印
  • 收藏
收藏成功


打开文本图片集

中图分类号:TP312

文献标志码:A

Research on Human Action Recognition Method Based on Neural Architecture Search Technology

Cao Yang (School ofElectronic Information Engineering,Chongqing Technologyand Business Institute, Chongqing 401520, China)

Abstract:Thisresearchusesneuralarchitecturesearch(NAS)technologtodesignaneuralnetworkmodelforhumanbehaviorbasedon sensordata,coveringthedesignof search space,selectionofsearch strategyandperformanceevaluation strategy.Thegeerated CNN-LSTMmodelisverifiedontheOPPORTUNITYdataset.Resultsshowthatwithasignificantreductioninparameters,tisodel performs comparably to DeepConvLSTM model.

Key words:sensor;human action recognition;neural architecture search

0引言

随着人工智能技术、通信技术和传感器技术的发展,人体行为识别已经广泛应用于智能家居、智能看护和医疗康复等领域。(剩余4640字)

monitor
客服机器人