活跃的骨架和演讲 [英] Active skeleton and speech

问题描述

大家好，

我们在一个或多或少嘈杂或嘈杂的环境中使用Kinect。我们也想使用语音识别 - 但这种识别应该与被跟踪的骨架有关 - 这样我们就可以忽略其他语音信号和噪音。

we are using Kinect in a more or less noisy or crowdy environment. We would like to use also speech recognition - but this recognition should be related to the skeleton that is tracked - so that we can ignore other voice signal and noises.

这是可能的，它是怎么回事？你有什么建议？

Is this possible and it it is how? What is you ADVICE?

你们是如何处理这个功能的？你们最好的做法是什么？

How are you guys handling this functionality - what are you best practices?

tnx

Kristjan

Kristjan http://www.adora-med.com

Kristjan http://www.adora-med.com

推荐答案

我还没有将语音识别与Kinect v2一起使用，但是我在v1上做了并且导致多个扬声器的环境非常差。在我的测试中，当有些人在Kinect用户附近说话时，认可成功率大幅下降。我会说其他
类型的噪音不太重要。

I have not used Speech Recognition with Kinect v2 yet, but I did on v1 and results in environments with multiple speakers were very poor. In my tests, recognition success fell dramatically when some people were talking near the Kinect user. I would say other kind of noises are less critical.

对于这种情况，你有几个跟踪的机构，你想知道哪个是在说话（当只有一个这样做时），有一个

最近的帖子暗示如何使用AudioBeamSubFrame信息。

For the case you have several Bodies tracked and you want to know which is talking (when only one did so), there is a recent post hinting to how to use AudioBeamSubFrame information.

这篇关于活跃的骨架和演讲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！