Microsoft Look to better Kinect Body Parts via Analysis

Posted on Mar 30 2011 - 11:53pm by Thomas Sharp

You would think that any electronics device which had entered the Guinness Book of Records for selling over 10 million units since its launch at the tail end of 2010 must already be pretty well designed.

However, it seems that Microsoft researchers have found some new ways to improve their Kinect device even further. The study was carried out by Microsoft Research in Cambridge and has resulted in a paper which suggests a way of giving a better performance.

The idea put forward is the use of the 3D depth maps together with a server to gather the information and give a more accurate 3D movement predictor. This would work by giving the servers the images of different parts and let it look over a million images a day and learn from this. This would let it track individual parts more easily and the idea led to the research paper’s name of “Real Time Human Pose Recognition in Parts from Single Depth Images”.

The big benefit which the researchers think that we will be able to see is a drop in the lag time which is currently experienced with Kinect. The process of recognizing body parts would then be reduced to as short a time scale as 5 ms on the basic Xbox 360.

It is not yet clear whether this approach is going to be taken up by Microsoft but it is clear that they are looking at ways to make the Kinect work better.

Are you unhappy with Kinect time lags or do you think that it gives a good performance?

Leave A Response