#kinect notes.

I am using a 64-bit operating system, Windows 10 home OS, connected to the Kinect for Windows sensor device. The Processing Library I am using is Bryan Chung’s Kinect for Processing Library.

01.   Install Kinect for Windows SDK v1.8, Kinect for Windows Runtime v1.8, and the Kinect for Windows Developer Toolkit v1.8; you can download the SDK here, the Runtime here, and the toolkit here. Windows installation instructions warn against having your Kinect device connected while downloading these packages. Make sure your Kinect device is disconnected from your computer before installing. Complete the installation process and close all the windows.

02.   Connect your Kinect for Windows device by the USB port and the power socket. Check that the green light between the cameras is blinking. That will tell you the device is connected to the computer and to power.

03.   Open Device Manager. ‘Kinect for Windows’ should expand to a camera element (and possibly an audio element). If there is a yellow warning sign next to the Kinect for Windows icon, you may have the wrong SDK installed (see previous post for details).

04.   Open Processing and import Bryan Chung’s Kinect for Processing Library (Sketch > Import Library > Add Library > Filter by ‘Kinect’ > Install ‘Kinect4WinSDK’). If the installation seems to be stuck on loading while the library is available, close and open Processing before checking if the library is installed fully in the ‘Import Library’ tab.

Note: this post is friendlier on dash view because of the bullet-point format. Click here to view on the dash.

As per the instructions on the Windows site, I downloaded and ran the v2 SDK and Kinect Studio. However, my Kinect for Windows device failed to connect with my computer, as the Device Manager showed a yellow warning sign alongside the Kinect for Windows icon.

This was solved by uninstalling the v2 SDK + accompany v2 applications, and installing the v1.8 SDK + accompanying v1.8 toolkit. The Device Manager then registered the Kinect for Windows device as working.

Initially, I was also following the tutorial Daniel Shiffman provided, as well as the open source Kinect for Windows library that he recommended in said tutorial. However, I got the error message ‘initDevice() does not exist’, among other things. 

This was solved by instead using Bryan Chung’s Kinect4Win library. By running the provided example program alongside the v.18 SDK, the Kinect sensors were able to provide me with all four views when executed (skeleton + depth + RGB, mask).

Oddly, after a successful connection and execution of the four-view feedback provided by the Kinect for Windows device, it ceased to work again when I got back home and tried running it. I had not altered anything since the successful attempt. When I ran the Kinect4Win example program, it gave me the error message: “(Init) Initialize sensor failed 0” — the 4-view window was also blank (no RGB / depth / infrared / skeleton view).

I uninstalled all the Kinect for Windows programs (v1.8 + v2, SDK, Runtime, Kinect Studio, Toolkit, etc) and removed the Kinect libraries from Processing (Bryan Chung’s Kinect4Win, Daniel Shiffman + Thomas Sanchez Lengeling’s Open Kinect for Processing). Then I reinstalled the v1.8 SDK and Runtime, as well as Bryan Chung’s Kinect4Win library in Processing. 

In Device Manager, I also enabled the hidden devices (View > ‘Show hidden devices’) to double-check that the other elements (Audio Array, Security Control, etc) was available - just not connected.

I then rebooted my computer and ran the Kinect4Win example code. This sometimes gave me the RGB and depth, and sometimes it did not. Neither the skeleton + depth view showed. I also noticed the infrared camera on the Kinect device itself would flare with infrared light for a few seconds at the start before turning off. Note: for this attempt, the Kinect device is plugged directly into the wall socket.

After disconnecting the device and reconnecting again (I had to move to another workspace), it reverted back to only being connected to the Kinect for Windows Camera. Note: for this attempt, I am using an extension cord rather than plugging Kinect device directly into the wall socket - unsure if that is a variable?

Rebooted computer, disconnected + reconnected device, ran the example code despite only the Kinect camera showing. Started working anyways???? Don’t know how Kinect works but I’m too afraid to ask.

Skeleton and mask views weren’t showing up for some reason

Moved at a distance in an attempted to activate the sensor. It worked! Skeleton and mask views became live.

Depth data would revert back to 0 when I got to close to the Kinect device, and would not register my depth even if I moved back to a distance that it detected before.