3D CNNs与LSTMs在行为识别中的组合及其应用

Combination of 3D CNNs and LSTMs and Its Application in Activity Recognition

中文关键词:  行为识别  深度学习  神经网络  模式识别

英文关键词:activity recognition  deep learning  neural networks  pattern recognition

中文摘要:

      基于机器视觉的人体运动识别在视频监控、虚拟现实、医疗护理等诸多领域发挥着重要的作用。结合深度学习中的三维卷积神经网络和长短期记忆神经网络，提出一种融合模型，并与另外两种行为识别模型——长效递归卷积网络和时空域卷积网络，进行了对比，利用公开的KTH数据集，进行了实验测试。实验表明，提出的融合模型与长效递归卷积网络和时空域卷积网络相比，对于人体行为图像或视频数据集的学习效果明显，论证了模型的泛化性能和鲁棒性。

英文摘要:

      Human activity recognition based on machine vision plays an important role in many fields such as video monitoring,virtual reality,medical care and so on.A new fusion model is proposed by combining the 3D convolutional neural networks and long short term memory neural network in deep learning,and compared with the other two activity recognition models,long-term recurrent convolution network and spatiotemporal convolutional neural network.Test experiments using public KTH data have been done to prove that the combination model has a better learning effect based on human activity image or video data set,which demonstrates the generalization performance and robustness of the model.