Toggle menu
Toggle preferences menu
Toggle personal menu
Not logged in
Your IP address will be publicly visible if you make any edits.
Azure Kinect DK[1][2]
Basic Info
VR/AR Augmented reality, Virtual reality
Type Cameras
Subtype Time-of-flight camera
Developer Microsoft
Manufacturer Microsoft
Announcement Date 24 February 2019 (Mobile World Congress 2019); previewed as Project Kinect for Azure on 7 May 2018
Release Date July 2019 (United States and China)
Price US$399 (launch price)
Website https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/
Requires Host PC with a USB 3 port, running Windows 10 version 1803 or later (x64) or Ubuntu 18.04 (x64)
Predecessor Kinect v2
Display
Resolution Depth up to 1024 x 1024; RGB up to 4096 x 3072 (4:3) or 3840 x 2160 (16:9)
Image
Field of View Depth 75 x 65 degrees (narrow modes) or 120 x 120 degrees (wide modes); RGB 90 x 59 degrees (16:9) or 90 x 74.3 degrees (4:3)
Tracking
Tracking Markerless body tracking through the Azure Kinect Body Tracking SDK (32-joint skeleton)
Audio
Microphone Seven-microphone circular array (USB audio class 2.0)
Camera 1-megapixel time-of-flight depth camera; 12-megapixel OV12A10 CMOS rolling-shutter RGB camera
Connectivity
Connectivity USB 3 over USB-C
Ports USB-C (data and power), barrel power connector, 3.5 mm Sync in and Sync out jacks
Power In-box power supply or USB-C; up to 5.9 W
Device
Dimensions 103 x 39 x 126 mm
Weight 440 g
Sensors Time-of-flight depth camera, RGB camera, IMU (LSM6DSMUS accelerometer and gyroscope)


Azure Kinect DK (often shortened to Azure Kinect) is a developer kit from Microsoft that combines a 1-megapixel time-of-flight depth camera, a 12-megapixel RGB camera, a seven-microphone circular array and an inertial measurement unit in one USB peripheral.[1] Microsoft announced it on 24 February 2019 at Mobile World Congress in Barcelona, at the same event as HoloLens 2, and took preorders at US$399 in the United States and China. Julia White, Microsoft's corporate vice president for Azure, wrote that "at its core is the time-of-flight depth sensor we developed for HoloLens 2".[2]

Microsoft described it as the fourth generation of the Microsoft Kinect line, and it was aimed at developers and commercial businesses rather than game players.[1][3][4] It became generally available in the United States and China in July 2019.[5] In VR and AR work it was used as a depth and color capture sensor for volumetric video, 3D telepresence and body tracking. Microsoft announced the end of production in August 2023 and sold further units until the end of October 2023. It pointed customers to the depth camera maker Orbbec, whose Femto Bolt camera uses the same depth camera module.[6][7]

Reviewed 11 October 2026. Dates, price, hardware specifications, SDK details, discontinuation, VR/AR uses and the five cited papers were checked against the cited Microsoft, press, GitHub, Orbbec and journal sources. About review dates.

History

Project Kinect for Azure

Microsoft stopped making the consumer Kinect in October 2017.[8] In February 2018, Cyrus S. Bamji and co-authors presented the depth sensor that the Azure Kinect would later use at the IEEE International Solid-State Circuits Conference (ISSCC), in a paper titled "1Mpixel 65nm BSI 320MHz Demodulated TOF Image Sensor with 3.5μm Global Shutter Pixels and Analog Binning". Microsoft's documentation describes the paper as covering the depth camera to be used in Project Kinect for Azure and the next version of HoloLens.[9][10]

Microsoft unveiled Project Kinect for Azure at its Build developer conference on 7 May 2018. GamesBeat described it as a package of sensors, including the next-generation depth camera, meant for AI developers. The specifications Microsoft gave at the time were a 1024 x 1024 depth resolution, overall system power of 225 to 950 mW, automatic per-pixel gain selection and a global shutter for better performance in sunlight. Microsoft called the project the fourth generation of Kinect.[3]

Announcement and release

Microsoft introduced the finished developer kit as the Azure Kinect DK on 24 February 2019 at Mobile World Congress in Barcelona, where it also unveiled HoloLens 2. Preorders opened that day at US$399, initially in the United States and China.[2] UploadVR reported that Microsoft's site gave 27 June 2019 as the date by which orders would start shipping, when new SDKs would also become available. External sync pins allowed several units to be linked.[4] In his coverage for Engadget, Mat Smith noted that the device could be used alone or paired with other Azure Kinect sensors, and wrote that its "intelligent edge" came from interpreting what it sensed as well as capturing sound and video in detail.[11] Next Reality listed three early-access companies: DataMesh, which used AR to train manufacturing workers; AVA Retail, for autonomous store technology; and Ocuvera, for a computer vision fall prevention system in healthcare facilities.[12]

On 11 July 2019, ahead of its Inspire conference in Las Vegas, Microsoft made the kit generally available in the United States and China and began shipping to customers who had preordered.[5][13] Microsoft's Azure update announcing general availability is dated 15 July 2019.[14] At that point Microsoft had released several SDKs, including a preview of a body tracking SDK. TechCrunch noted that the kit could connect to Azure's machine learning services but did not require Azure, and that Microsoft was targeting health and life sciences, retail, logistics and robotics.[5] Sales later expanded to the United Kingdom, Germany and Japan; Road to VR reported in November 2020 that the kit was sold only in those three countries plus the United States and China.[15]

Licensing of the depth technology

In November 2020 Microsoft said it was working with Analog Devices and the industrial sensor maker SICK AG to build commercial products on the Azure Kinect's time-of-flight depth technology. Analog Devices planned depth sensor silicon and a commercial depth camera module for "consumer electronics, automotive cabins and industrial logistic use-cases", and SICK planned to add the technology to its Visionary-T camera line for intralogistics, robotics and industrial vehicles. Both companies were to manufacture and sell their own products.[15]

Discontinuation

In August 2023 Microsoft said it had "made the decision to end production of Azure Kinect Developer Kit." It directed customers to Orbbec, whose cameras "use the same depth camera module as Azure Kinect Developer Kit." Orbbec had just launched the Femto Bolt, and developers could move existing Azure Kinect applications to Orbbec cameras through an API bridge in Orbbec's SDK.[6] According to The Robot Report, Microsoft sold further units until the end of October 2023 "or until supplies last", with the standard limited hardware warranty. Existing users could keep using the device "without disruption" and the SDK stayed available for download.[7]

Orbbec's documentation says the Femto Bolt has "identical operating modes and performance" to the Azure Kinect DK's depth camera. Its listed differences include a narrower RGB field of view (80 x 51 degrees against 90 x 59 degrees), a depth of 64.9 mm against 125.4 mm, RGB HDR support, an 8-pin sync connector instead of 3.5 mm audio jacks, and a screw-locking USB-C connector. Orbbec also publishes a K4A wrapper so that applications written for the Azure Kinect DK can be switched to the Femto Bolt.[16]

Microsoft published a final Sensor SDK release, version 1.4.2, on 21 June 2024; its release notes say it "contains security fixes". The SDK's GitHub repository was archived on 22 August 2024 and is now read-only.[17][18] Microsoft's Azure Kinect DK documentation now sits in the archived "previous versions" section of Microsoft Learn.[19]

Design and hardware

The Azure Kinect DK measures 103 x 39 x 126 mm and weighs 440 g. Microsoft specifies operation at 10-25 degrees Celsius and 8-90% non-condensing relative humidity; a cooling channel runs between the front section and the rear sleeve and must not be blocked. The device is calibrated at the factory, and the calibration data for the visual and inertial sensors can be read through the Sensor SDK. A camera-streaming indicator on the front can be switched off in software, and a status LED on the back shows power and connection state.[1]

Depth camera

The depth camera works on the amplitude modulated continuous wave (AMCW) time-of-flight principle. It projects modulated near-infrared light onto the scene and indirectly measures how long the light takes to return, producing a depth map in millimeters and a "clean IR" image. Microsoft lists a 1-megapixel imaging chip with 3.5 μm by 3.5 μm pixels (which it calls the world's smallest time-of-flight pixel), two near-infrared laser diodes for the narrow and wide field-of-view modes, automatic per-pixel gain selection, a global shutter and a multi-phase depth calculation. The camera sends raw modulated IR images to the host PC, where GPU-accelerated depth engine software turns them into depth maps. Pixels are marked invalid (depth value 0) when they fall outside the illumination mask, are saturated, receive too little signal, or show multi-path interference, which is common in room corners and at object edges.[20]

Microsoft specifies the following depth modes. Operating ranges assume 15% to 95% reflectivity at 850 nm; Microsoft gives a random error standard deviation of 17 mm or less and a typical systematic error below 11 mm plus 0.1% of distance, excluding multi-path interference.[1]

Mode Resolution Field of interest Frame rates (fps) Operating range Exposure time
Narrow FOV, unbinned 640 x 576 75 x 65 degrees 0, 5, 15, 30 0.5-3.86 m 12.8 ms
Narrow FOV, 2x2 binned (software) 320 x 288 75 x 65 degrees 0, 5, 15, 30 0.5-5.46 m 12.8 ms
Wide FOV, 2x2 binned 512 x 512 120 x 120 degrees 0, 5, 15, 30 0.25-2.88 m 12.8 ms
Wide FOV, unbinned 1024 x 1024 120 x 120 degrees 0, 5, 15 0.25-2.21 m 20.3 ms
Passive IR 1024 x 1024 N/A 0, 5, 15, 30 N/A 1.6 ms

RGB camera

The color camera uses an OV12A10 12-megapixel CMOS rolling-shutter sensor. It is USB Video Class compatible and works without the Sensor SDK. Native modes output MJPEG, with YUY2 and NV12 also available at 1280 x 720; the SDK can deliver BGRA images by converting MJPEG on the host CPU.[1]

Resolution Aspect ratio Frame rates (fps) Nominal field of view
3840 x 2160 16:9 0, 5, 15, 30 90 x 59 degrees
2560 x 1440 16:9 0, 5, 15, 30 90 x 59 degrees
1920 x 1080 16:9 0, 5, 15, 30 90 x 59 degrees
1280 x 720 16:9 0, 5, 15, 30 90 x 59 degrees
4096 x 3072 4:3 0, 5, 15 90 x 74.3 degrees
2048 x 1536 4:3 0, 5, 15, 30 90 x 74.3 degrees

Motion sensor, microphones and connections

The IMU is an LSM6DSMUS with an accelerometer and a gyroscope, sampled together at 1.6 kHz and reported to the host at 208 Hz. The seven-microphone circular array appears as a standard USB audio class 2.0 device with all seven channels accessible; Microsoft gives a sensitivity of -22 dBFS (94 dB SPL, 1 kHz), a signal-to-noise ratio above 65 dB and an acoustic overload point of 116 dB.[1] The kit has no speakers.[21]

The device is a USB 3 composite device built around a USB 3.1 Gen 2 hub, with the depth and color cameras on USB 3.0 and the microphones on high-speed USB. It can be powered by the in-box supply, whose cable runs from USB Type-A to a barrel connector, or by a single USB Type-C to Type-C cable carrying both power and data; a Type-C to Type-C cable is not included. Power draw is up to 5.9 W depending on use.[1]

Multi-device synchronization

Two 3.5 mm jacks, Sync in and Sync out, sit under a removable plastic cover. In a daisy-chain setup one master unit can drive up to eight subordinate units, and a star setup with a headphone splitter supports up to two subordinates; an external trigger source can also drive the master. Because each depth capture fires the lasers nine times for 125 μs each, Microsoft recommends offsetting the captures of overlapping depth cameras by at least 160 μs, which allows up to ten cameras to interleave their exposures without interfering. Microsoft lists filling occlusions, scanning objects in three dimensions, raising the effective frame rate above 30 fps and capturing multiple aligned 4K color images among the reasons to combine units.[22]

Software

Microsoft provided three SDKs for the kit: the Sensor SDK for low-level device access, the Body Tracking SDK for tracking bodies in 3D, and the Azure AI Speech SDK for microphone access and cloud speech services. Azure's cloud vision services could also be used with the RGB camera.[21]

The Sensor SDK (also called K4A) is MIT-licensed and runs on Windows and Linux.[17] It exposes depth mode control, RGB camera control, IMU data, synchronized depth and RGB streaming with configurable delay, external device synchronization, frame metadata and calibration data. It ships with a viewer, a recorder and playback API that use the Matroska container format, and a firmware update tool.[21] Microsoft's minimum host for the Sensor SDK on Windows is a seventh-generation Intel Core i3 (dual-core 2.4 GHz with HD620 graphics), 4 GB of memory, a dedicated USB 3 port and graphics driver support for OpenGL 4.4 or DirectX 11.0. The supported systems are Windows 10 version 1803 (x64) or later and Ubuntu 18.04 (x64); the SDK does not support UWP applications or Windows 10 in S mode.[23]

The Body Tracking SDK tracks several people at once. Each tracked body has an ID that persists between frames and a skeleton of 32 joints, from the pelvis out to the hand tips, feet, eyes and ears. Joint positions are given in millimeters and orientations as normalized quaternions in the depth camera's coordinate system.[24] Version 1.1.0 updated the SDK to ONNX Runtime 1.6 with CPU, CUDA, DirectML (Windows only) and TensorRT execution of the pose estimation model, and added a "lite" model that Microsoft says trades about 5% accuracy for roughly twice the performance. The last listed version is 1.1.2, which added C# wrapper support on Linux. On Linux, Microsoft supports Ubuntu 18.04 and 20.04 for body tracking.[25] Microsoft's recommended minimum host for body tracking is a seventh-generation Intel Core i5 (quad-core 2.4 GHz), 4 GB of memory, an NVIDIA GeForce GTX 1050 or equivalent and a dedicated USB 3 port, based on tracking five people in the narrow unbinned depth mode at 30 fps.[23]

Use in VR and AR

The Azure Kinect is built around the time-of-flight depth sensor that Microsoft developed for HoloLens 2, packaged as a USB peripheral for a host PC.[2][5]

Microsoft Research's project page for Holoportation, its 3D capture and telepresence system, describes a "new Holoportation system" that uses Azure Kinect units for depth and RGB capture. The project page states that "the Azure Kinects provide some of the highest quality depth data of any RGB+D sensor on the market."[26]

The volumetric capture software Depthkit added official Azure Kinect support in version 0.3.12 on 11 July 2019. Its developers said they had worked with Microsoft since December 2018, integrating Depthkit with pre-release sensor hardware.[27]

Researchers have paired the device with mixed reality headsets. A 2025 proof-of-concept study by Bini and colleagues combined HoloLens 2, a Wii Balance Board and an Azure Kinect into a mixed reality framework for assessing balance and posture.[28] Another group used the sensor's 3D body tracking for both standard motor tests and virtual exergames in a remote monitoring and rehabilitation system for people with Parkinson's disease.[29]

Reception and evaluation

At launch, Engadget's Mat Smith argued that the hardware mattered less than the Azure cloud and AI services it was designed to work with, and called the healthcare use cases the most interesting.[11] Next Reality's Tommy Palladino described the US$399 price as in line with similar peripherals such as Occipital's Structure Core.[12]

Independent measurements have broadly confirmed Microsoft's depth figures. Tölgyessy and colleagues (2021) compared all three Kinect generations and tested the Azure Kinect with a robotic arm, 18 materials and indoor and outdoor scenes. They confirmed a standard deviation of 17 mm or less and a distance error below 11 mm up to 3.5 m in all four depth modes. They also found that the device needs a warm-up of at least 40-50 minutes to give stable results and cannot be used reliably in direct sunlight, which makes it mainly an indoor sensor.[30] Kurillo and colleagues (2022) measured a planar target against laser-scanner ground truth. Beyond 2.5 m the Azure Kinect was more accurate than the Kinect v2 in both spatial and temporal terms; below 2.5 m the two were comparable. The authors concluded that it is a suitable substitute for the Kinect v2 in 3D scanning.[31] In a treadmill gait study against a Vicon marker-based system, Albert and colleagues (2020) found significantly higher accuracy for spatial gait parameters with the Azure Kinect than with the Kinect v2, with no significant difference for temporal parameters.[32]

See also

References

  1. ↑ 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 "Azure Kinect DK hardware specifications". Microsoft Learn. Microsoft. 2021-03-18. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/hardware-specification. Retrieved 2026-10-11.
  2. ↑ 2.0 2.1 2.2 2.3 Julia White (2019-02-24). "Microsoft at MWC Barcelona: Introducing Microsoft HoloLens 2". Official Microsoft Blog. Microsoft. https://blogs.microsoft.com/blog/2019/02/24/microsoft-at-mwc-barcelona-introducing-microsoft-hololens-2/. Retrieved 2026-10-11.
  3. ↑ 3.0 3.1 "Microsoft announces Project Kinect for Azure with its next-generation depth camera". GamesBeat. 2018-05-07. https://gamesbeat.com/microsoft-announces-project-kinect-for-azure-with-its-next-generation-depth-camera/. Retrieved 2026-10-11.
  4. ↑ 4.0 4.1 Ian Hamilton (2019-02-24). "Microsoft's $399 Azure Kinect Development Kit Now Available For Pre-Order". UploadVR. https://www.uploadvr.com/microsoft-dk-azure-kinect/. Retrieved 2026-10-11.
  5. ↑ 5.0 5.1 5.2 5.3 Frederic Lardinois (2019-07-11). "Microsoft's $399 Azure Kinect AI camera is now shipping in the US and China". TechCrunch. https://techcrunch.com/2019/07/11/microsofts-399-azure-kinect-ai-camera-is-now-shipping-in-the-u-s-and-china/. Retrieved 2026-10-11.
  6. ↑ 6.0 6.1 Brian Heater (2023-08-22). "Microsoft can't stop discontinuing Kinect". TechCrunch. https://techcrunch.com/2023/08/22/microsoft-cant-stop-discontinuing-kinect/. Retrieved 2026-10-11.
  7. ↑ 7.0 7.1 Brianna Wessling (2023-09-05). "Microsoft ending production of Azure Kinect Developer Kit". The Robot Report. https://www.therobotreport.com/microsoft-ending-production-of-azure-kinect-developer-kit/. Retrieved 2026-10-11.
  8. ↑ "The Kinect is a self-contained 3D sensor again, and it only costs $399". Geo Week News. 2019-02-28. https://www.geoweeknews.com/news/the-kinect-is-a-self-contained-3d-sensor-again-and-it-only-costs-399. Retrieved 2026-10-11.
  9. ↑ "Depth camera whitepaper - ISSCC 2018". Microsoft Learn. Microsoft. 2018-07-05. https://learn.microsoft.com/en-us/windows/mixed-reality/out-of-scope/isscc-2018. Retrieved 2026-10-11.
  10. ↑ Cyrus S. Bamji, Swati Mehta, Barry Thompson, et al. (2018-02). "1Mpixel 65nm BSI 320MHz Demodulated TOF Image Sensor with 3.5μm Global Shutter Pixels and Analog Binning". 2018 IEEE International Solid-State Circuits Conference (ISSCC). pp. 94-95. https://ieeexplore.ieee.org/document/8310200. Retrieved 2026-10-11.
  11. ↑ 11.0 11.1 Mat Smith (2019-02-24). "Microsoft resurrects the Kinect, but for business". Engadget. https://www.engadget.com/2019-02-24-microsoft-resurrects-kinect-azure-kinect.html. Retrieved 2026-10-11.
  12. ↑ 12.0 12.1 Tommy Palladino. "Microsoft's Azure Kinect Standalone Depth Sensor Powers Major Augmented Reality Improvements for $399". Next Reality. https://next.reality.news/news/microsofts-azure-kinect-standalone-depth-sensor-powers-major-augmented-reality-improvements-for-399-0194185/. Retrieved 2026-10-11.
  13. ↑ Kyle Wiggers (2019-07-11). "Microsoft's Azure Kinect Developer Kit begins shipping in the U.S. and China". VentureBeat. https://venturebeat.com/ai/microsofts-azure-kinect-developer-kit-begins-shipping-in-the-u-s-and-china. Retrieved 2026-10-11.
  14. ↑ "Azure Kinect DK is now generally available in the United States and China". Azure updates. Microsoft. 2019-07-15. Archived from the original on 2019-10-14. https://web.archive.org/web/20191014184906/https://azure.microsoft.com/en-us/updates/azure-kinect-dk-is-now-generally-available-in-the-united-states-and-china/. Retrieved 2026-10-11.
  15. ↑ 15.0 15.1 Scott Hayden (2020-11-02). "Microsoft to Expand Azure Kinect's Computer Vision Tech to Commercial Market". Road to VR. https://www.roadtovr.com/microsoft-azure-kinect-commercial/. Retrieved 2026-10-11.
  16. ↑ "Femto Bolt Comparison with Azure Kinect DK". Orbbec Documentation. Orbbec. 2025-09-16. https://doc.orbbec.com/documentation/Orbbec%20Femto%20Bolt%20Documentation/Femto%20Bolt%20Comparison%20with%20Azure%20Kinect%20DK. Retrieved 2026-10-11.
  17. ↑ 17.0 17.1 "Azure Kinect SDK (K4A)". GitHub. Microsoft. https://github.com/microsoft/Azure-Kinect-Sensor-SDK. Retrieved 2026-10-11.
  18. ↑ "Release 1.4.2". GitHub. Microsoft. 2024-06-21. https://github.com/microsoft/Azure-Kinect-Sensor-SDK/releases/tag/v1.4.2. Retrieved 2026-10-11.
  19. ↑ "Azure Kinect DK documentation". Microsoft Learn. Microsoft. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/. Retrieved 2026-10-11.
  20. ↑ "Azure Kinect DK depth camera". Microsoft Learn. Microsoft. 2019-06-26. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/depth-camera. Retrieved 2026-10-11.
  21. ↑ 21.0 21.1 21.2 "About Azure Kinect DK". Microsoft Learn. Microsoft. 2019-06-26. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/about-azure-kinect-dk. Retrieved 2026-10-11.
  22. ↑ "Synchronize multiple Azure Kinect DK devices". Microsoft Learn. Microsoft. 2020-02-20. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/multi-camera-sync. Retrieved 2026-10-11.
  23. ↑ 23.0 23.1 "Azure Kinect Sensor SDK system requirements". Microsoft Learn. Microsoft. 2021-03-05. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/system-requirements. Retrieved 2026-10-11.
  24. ↑ "Azure Kinect body tracking joints". Microsoft Learn. Microsoft. 2019-06-26. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/body-joints. Retrieved 2026-10-11.
  25. ↑ "Azure Kinect Body Tracking SDK download". Microsoft Learn. Microsoft. 2022-03-21. https://learn.microsoft.com/en-us/previous-versions/azure/kinect-dk/body-sdk-download. Retrieved 2026-10-11.
  26. ↑ "Holoportation". Microsoft Research. Microsoft. https://www.microsoft.com/en-us/research/project/holoportation-3/. Retrieved 2026-10-11.
  27. ↑ "Announcing Azure Kinect support in Depthkit". Depthkit. Scatter. 2019-07-11. https://www.depthkit.tv/posts/azure-kinect-depthkit-sensor-kinect-microsoft. Retrieved 2026-10-11.
  28. ↑ F. Bini, M. Franzò, A. Finti, F. Tiberi, V. M. T. Grillo, E. Covelli, M. Barbara, F. Marinozzi (2025). "Extended Reality-Based Proof-of-Concept for Clinical Assessment Balance and Postural Disorders for Personalized Innovative Protocol". Bioengineering, vol. 12, no. 8. https://doi.org/10.3390/bioengineering12080850. Retrieved 2026-10-11.
  29. ↑ G. Amprimo, G. Masi, L. Priano, C. Azzaro, F. Galli, G. Pettiti, A. Mauro, C. Ferraris (2022). "Assessment Tasks and Virtual Exergames for Remote Monitoring of Parkinson's Disease: An Integrated Approach Based on Azure Kinect". Sensors, vol. 22, no. 21. https://doi.org/10.3390/s22218173. Retrieved 2026-10-11.
  30. ↑ M. Tölgyessy, M. Dekan, Ľ. Chovanec, P. Hubinský (2021). "Evaluation of the Azure Kinect and Its Comparison to Kinect V1 and Kinect V2". Sensors, vol. 21, no. 2. https://doi.org/10.3390/s21020413. Retrieved 2026-10-11.
  31. ↑ G. Kurillo, E. Hemingway, M.-L. Cheng, L. Cheng (2022). "Evaluating the Accuracy of the Azure Kinect and Kinect v2". Sensors, vol. 22, no. 7. https://doi.org/10.3390/s22072469. Retrieved 2026-10-11.
  32. ↑ J. A. Albert, V. Owolabi, A. Gebel, C. M. Brahms, U. Granacher, B. Arnrich (2020). "Evaluation of the Pose Tracking Performance of the Azure Kinect and Kinect v2 for Gait Analysis in Comparison with a Gold Standard: A Pilot Study". Sensors, vol. 20, no. 18. https://doi.org/10.3390/s20185104. Retrieved 2026-10-11.