Today, I’m going to show you how to detect a jump using Kinect and Vitruvius. Jumping is one of the most challenging actions to detect.
In this tutorial, you’ll also learn how to measure the vertical distance between the person and the floor. To have a better visualization, we are also going to animate a 3D model while the person is moving. Sounds complex? Don’t worry! Using Vitruvius, we can do it in a few minutes.
Why detecting a jump is difficult?
Detecting and measuring a jump is not trivial. Why? Because jumping is a complex action. Jumping is not about body joints only. We need to consider various parameters from the human body and the environment. As a result, we need two types of data:
- The position of the lower body joints.
- The position & orientation of the floor.
Let’s see how Kinect, Unity and Vitruvius can solve this problem.
Prerequisites
To run the demos, you’ll need the following software and hardware:
- Kinect for XBOX v2 sensor with an adapter (or Kinect for Windows v2 sensor)
- Kinect for Windows v2 SDK
- Unity3D 5.5 or higher
- Windows 8.1 or higher
- Visual Studio 2013 or higher
- A dedicated USB 3 port
Jump Detection in Unity
Since we want to animate a 3D model, we’ll be using Unity3D. Here is a step-by-step guide:
Step #1 – Download Vitruvius
The Jump gesture/action is available in the Academic, Premium, and Platinum packages of Vitruvius. When you download the package, unzip the compressed folder and launch Unity3D.
Step #2 – Open the Sample projects
Vitruvius comes with 11 sample projects, so you can get started immediately. Open LightBuzz.Vitruvius.3.6.0.unitypackage and extract their contents into a new Unity project. Check the official Unity Documentation on how to import a custom Unity package.
Step #3 – Open the Jump Scene
The scene is loaded with a 3D avatar:
Step #4 – Add the JumpFBX script
Select the avatar in the Hierarchy window and check its properties in the Inspector window. Add the script named “JumpFBX.cs”. The JumpFBX script will add jump capabilities to any FBX 3D model.
Step #4 – Move around!
When Kinect is connected, stand in front of the sensor and move 2 times back-and-forth within the field of view. This is an essential step, since Kinect will be scanning the floor while you are moving.
Check this video to see how:
Step #5 – Get the height of the jump
To detect the exact height of a jump, all you have to do is call the JumpHeight property of the model object. It’s measured in meters:
var height = model.JumpHeight;
The sample project will automatically notify you when someone is jumping. Here is a snapshot of Michail (the guy is jumping really high):
This is it! You can now create amazing apps using Kinect and Vitruvius motion analysis tools.
‘Til the next time, keep Kinecting.
Download Vitruvius
Hello VANGOS PTERNEAS, its great to see that you have provided a sample on Jump. I need some help in detecting the walking and 360 rotation in kinect when I am running/walking on a treadmil. that means my legs are just going forward and backward standing at one position. Can you please help me out with this?
thank you.
Hello, Faisal. Unfortunately, this is not possible using the Kinect SDK. It would require a lot of custom algorithms and raw-data processing.
Hi,
Do you have a similar script for detecting jump heights in WPF projects?
Thanks!
Hi Reuben. We are preparing some new jumping samples (WPF/Unity) with floor detection. Just stay tuned 🙂
is this better then the iclone 7 software??
Hi Paul. Vitruvius is software for totally different purposes and use-cases. Vitruvius is a motion analysis tool that simplifies Kinect development. iClone is a 3D animation software.
Do you still support this library? I sent an email two days ago and didn’t get an answer yet.
Hello and thank you for your message. Vitruvius is supported and updated! We reply to our Academic, Premium, and Platinum customers within 24 hours. We reply to our Free customers within 48 hours. In case you have contacted us and received no reply, please check your Spam folder.
Thanks for the reply Vangos.
If I may ask another question; about the jump algorithm of Vitruvius this time: how do you compensate the variations on the 3D points of each joint (mainly on the lower body joints as they seem to vary much more) even when the person is standing still? Do you use any magic constant to get only higher variations when detecting for jumps? Or do you use some more sophisticated algorithm to recognize the jumps?
We check the current frame in terms of previous frames. We are also considering the distance between the ankles (Vector3) and the floor (Vector4).