手势识别 [英] Gesture recognition

问题描述

您好！

我正致力于使用机器学习使用Windows sdk v2来识别手势和情绪的项目。我将使用kinect录制视频进行训练，但实际上数据量非常大。  有更好的方法处理小数据量而
达到高准确识别率吗？

谢谢

Daniel

解决方案

您使用的是Kinect Studio还是自己动手？ Kinect Studio录制可控制您捕获的数据源。与往常一样，不要记录您不需要的东西，这样可以保持尺寸不变。颜色将成为最大的消费者
的数据，所以如果你可以避免它，那将是最好的。

如果您的手势基于身体追踪数据，我们已经提供了一个工具，可以使用名为Visual Gesture Builder的Kinect Studio剪辑。在channel9上已经录制了几个深度视频。

http://social.msdn.microsoft.com/Forums/en-US/02e0302a-e3bd -46d3-9146-0dacd11d2a8d /深俯冲视频和 - 白皮书换视觉姿势的助洗剂？论坛= kinectv2sdk

Hello!

I am working on project to recognize a gesture and emotions using windows sdk v2 using machine learning. I am to record videos using kinect for training but actually the data size is quite huge.  Is there a better way to work on small data size while achieving a high accurate recognition rate?

thanks

Daniel

解决方案

Are you using Kinect Studio or did you roll your own? Kinect Studio recording provide control over which data sources you are capturing. As always, don't record things your don't need that should keep the size down. Color is going to be the biggest consumer of data, so if you can avoid it that would be best.

If your gestures are based on Body tracking data, we already provide a tool that would with Kinect Studio clips called Visual Gesture Builder. There are a couple in depth videos already recorded on channel9.

http://social.msdn.microsoft.com/Forums/en-US/02e0302a-e3bd-46d3-9146-0dacd11d2a8d/deep-dive-videos-and-whitepaper-for-visual-gesture-builder?forum=kinectv2sdk

这篇关于手势识别的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！