Modeling spatiotemporal patterns of gait anomaly with a CNN-LSTM deep neural network

作者：

Highlights：

• A multi-class skeleton-based gait anomaly recognition framework (SGAR) is presented.

• An ablation study illustrates the effect of different elements of the framework.

• New data augmentation is presented to improve the deep models’ performance for SGAR.

• The presented model acquires state-of-the-art performance on three SGAR datasets.

摘要

•A multi-class skeleton-based gait anomaly recognition framework (SGAR) is presented.•An ablation study illustrates the effect of different elements of the framework.•New data augmentation is presented to improve the deep models’ performance for SGAR.•The presented model acquires state-of-the-art performance on three SGAR datasets.

论文关键词：Gait anomalies,Skeleton-based gait anomaly recognition (SGAR),Multi-class anomaly recognition,Deep learning

论文评审过程：Received 16 January 2021, Revised 19 April 2021, Accepted 7 July 2021, Available online 18 July 2021, Version of Record 24 July 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115582