Microsoft releases the Kinect for Windows SDK beta

Background

A software development kit (SDK) is the package a platform maker hands to outside programmers: libraries that talk to the hardware, sample code, and documentation. Without one, using a device means reverse-engineering its data streams. For Kinect, the interesting data is not an ordinary video picture but a set of measurements about the room and the people in it — how far away each point is, where a body's joints are, and which direction a voice came from.

The announcement

On 16 June 2011, Microsoft announced that a beta version of the Kinect for Windows SDK was available from Microsoft Research ^[1]. Microsoft described the beta as a free release for noncommercial applications ^[1]. Tony Hey of Microsoft Research wrote that the SDK beta was released on 16 June 2011 to enable the next phase of bringing natural user interfaces to the PC ^[2].

What developers got

Microsoft said the SDK was designed to let developers create experiences involving depth sensing, human motion tracking, and voice and object recognition on Windows 7 ^[1]. Depth sensing means producing, for each pixel, an estimate of distance rather than colour; that turns a flat image into a rough three-dimensional map of a room. Human motion tracking went further: skeletal tracking could follow the skeleton image of one or two people moving within the Kinect field of view ^[1]. In practice that means fitting a jointed model — head, shoulders, arms, legs — to the body in the sensor's view, so a program can react to a gesture instead of a keystroke.

The SDK also gave access to raw streams from the depth sensor, the colour camera sensor and the four-element microphone array ^[1]. Raw access matters because it lets developers ignore the platform's own interpretations and build their own. A four-element microphone array, for example, hears the same sound at slightly different moments at each microphone; comparing those arrival times gives a direction to the speaker, which is how a system can tell who is talking in a room.

Alongside the code, Microsoft said the SDK included more than 100 pages of technical documentation ^[1] — the part that often decides whether an outside programmer gets something working in an afternoon or gives up.

Just before the release

Microsoft Research says it hosted a 24-hour coding marathon for selected researchers and enthusiasts in Redmond just before the general release ^[2]. The format is a common one for new hardware and software platforms: hand a small group early access, let them build something in a single sitting, and use what they make as evidence that the tools are usable.

What it signalled

Microsoft said it intended to release a commercial version of the SDK at a later date ^[1], which set out a two-step plan: a free beta for noncommercial work first, a commercial route for products afterwards. The framing Microsoft Research used was broader than games. Bringing natural user interfaces to the PC ^[2] means input based on body movement, gesture and speech rather than the keyboard and mouse — and the SDK was the piece that let people outside Microsoft try that on Windows 7 ^[1].