Gesture sequence recognition with one shot learned CRF/HMM hybrid model

作者：

Highlights：

• A hybrid CRF/HMM system for gesture recognition is proposed.

• HMM and CRF advantages combination and disadvantages compensation.

• Gesture Signature: an optical-flow-based gesture characterization model is proposed.

• Evaluation on the Chalearn competition data set under a one-shot learning context.

摘要

•A hybrid CRF/HMM system for gesture recognition is proposed.•HMM and CRF advantages combination and disadvantages compensation.•Gesture Signature: an optical-flow-based gesture characterization model is proposed.•Evaluation on the Chalearn competition data set under a one-shot learning context.

论文关键词：Gesture recognition,One-shot-learning,Hybrid system,Hidden Markov model,Conditional random field,Gesture characterisation

论文评审过程：Received 26 June 2016, Revised 28 January 2017, Accepted 8 February 2017, Available online 24 February 2017, Version of Record 9 March 2017.

论文官网地址：https://doi.org/10.1016/j.imavis.2017.02.003