Fast hand posture classification using depth features extracted from random line segments

作者：

Highlights：

• A set of features used with random forest to classify hand posture in a depth image.

• A very fast algorithm that tests images at 600fps using one core of CPU.

• Accuracy on one of the most challenging dataset very close to state-of-the-art.

• Good potential of the features to work for postures in difficult view angles.

• A pre-trained demo program available to public.

摘要

•A set of features used with random forest to classify hand posture in a depth image.•A very fast algorithm that tests images at 600fps using one core of CPU.•Accuracy on one of the most challenging dataset very close to state-of-the-art.•Good potential of the features to work for postures in difficult view angles.•A pre-trained demo program available to public.

论文关键词：Hand posture,Depth feature,Random forest

论文评审过程：Received 24 February 2016, Revised 23 November 2016, Accepted 27 November 2016, Available online 2 December 2016, Version of Record 14 December 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.11.022