On May 21, Shanghai HiSilicon released its XR chip platform, the first XR chip with 8K decoding capability and GPU and NPU integration. The first AR glasses based on this platform are the Rokid Vision.
HiSilicon says the XR chip platform supports 8K hard decoding capability up to 42.7PPD for a single eye.
In VR, for example, the ideal resolution for the human eye is about 60PPD, and the general support for 4KP60 decoding ability of the chip can support the single-eye resolution of 21.3PPD, corresponding to the TV can only be equivalent to 480P TV display effect, but is far from comfortable and immersive effect.
The XR chips released by HiSilicon have also been upgraded from decoding capabilities to provide a clearer presentation of content.
The platform incorporates HiSilicon's proprietary architecture, NPU, which provides NPU computing power up to 9TOPS.
An important application scenario for AR products is face recognition and license plate recognition, which requires real-time analysis of the video images captured in the AR glasses camera, correct identification of the face and license plate in each frame, uploading the identification results to the server for comparison, and displaying the comparison results on the screen in the glasses.
From image acquisition, analysis, uploading to display, the shorter the time spent, the faster the speed, the better the user experience, and the analysis stage is the longest time-consuming link in the whole chain, in order to complete the analysis and identification of hundreds of faces and license plates every minute in millisecond time, the chip needs to have strong NPU support. It is also with this in mind that HiSilicon has optimized its platform architecture.
As the first product of the HiSiliconXR chip platform, Rokid Vision is a core product of Rokid, a domestic AR company.
The frontal appearance of the product is similar to ordinary sunglasses, with two independent displays on the inside of the lens, with 40° + field of view (FOV), 1080P 3D display effect, the use of optical waveguide display scheme, together with the stereo sound on both sides of the lens legs, can bring users an immersive experience equivalent to a 110-inch 3D HD TV 4 meters away.
Rokid Vision will be used in security, industrial, education, retail, MICE and many other sectors to provide a more powerful, portable and immersive experience.
Rokid has also officially announced its addition to HiSilicon's XR chip platform. Previously, Rokid has worked with HiSilicon to optimize algorithms and schemes specific to this platform.
The two sides have done a lot of optimization work on AI computing capabilities such as voice, vision and environment awareness, and completed the high-resolution, low-latency design for multiple video codecs, providing core basic capabilities such as AI interaction and multi-player collaboration for XR industry from multiple levels of chip specifications, system software and cloud services.