Human action recognition toward massive-scale sport sceneries based on deep multi-model feature fusion

作者：

Highlights：

• We fuse visual feature, audio signal and skeleton posture into a hybrid feature representation.

• We convert the audio information to spectrogram image.

• We conduct a comparative study to compare the performance of different feature fusion.

摘要

•We fuse visual feature, audio signal and skeleton posture into a hybrid feature representation.•We convert the audio information to spectrogram image.•We conduct a comparative study to compare the performance of different feature fusion.

论文关键词：Human action recognition,Multi-model feature fusion,Sport scenes,Multi-class SVM,Human behavior

论文评审过程：Received 17 September 2019, Revised 10 January 2020, Accepted 20 January 2020, Available online 23 January 2020, Version of Record 3 March 2020.

论文官网地址：https://doi.org/10.1016/j.image.2020.115802