Nweon

Meta的马克·拉布金透露了关于Quest 3S市场成功的细节。Quest 3S是Meta推出的一款价格较低的入门款VR头显，它吸引了大量的新用户。据马克·拉布金表示，Quest 3S的新人比例是自Quest 2以来最高的。Meta此前推出了一系列的VR头显，每款设备都会吸引新的用户，但是Quest 3S由于价格较亲民，因此吸引了更多的新人。拉布金进一步指出，除了吸引新用户，Meta重要的任务之一是留住长期活跃的用户，因为他们对Quest内容生态系统的健康发展非常重要。

Valve最近发布了Steam 2.8，并正式支持Meta Quest的Steam Link手部追踪。然而，需要注意的是，这并不是完整的手部追踪支持，用户无法使用手和手指来控制SteamVR界面。如果在使用手部追踪时遇到问题，用户可以在Meta Quest设置中关闭“手部和身体追踪”的设备权限。另外，Valve强调他们将继续将OpenXR作为新游戏和应用的首选API，并提供了SteamVR的官方OpenXR子论坛。Steam 2.8的更新内容包括了改进了Dashboard Grab Handles的运动模型、重新设计了桌面窗口的控件样式、允许Tundra追踪器绑定到左右脚踝以及修复了一些问题。在OpenXR方面，修复了在stage和local空间切换时的不一致问题和非基础投影层显示在错误空间的问题。此外，Steam Link也更新至支持Quest 3S，并对2D/3D VR行为进行了改进，修复了手部追踪控制器在绑定UI中的图像绑定位置问题。

奥地利眼动追踪解决方案厂商Viewpointsystem宣布加入增强现实联盟（AR Alliance）。AR Alliance是一个专注于协调、推进和发展全球增强现实供应链以及加速创新的机构。该联盟致力于为各种规模的公司提供支持性和中立的环境，帮助推进和加强增强现实硬件开发生态系统，并加快创新的突破。AR Alliance的成员通过交流经验教训和市场经验，相互探讨有效的方法和无效的途径，共同努力创造有意义和积极的用户体验。Viewpointsystem作为AR Alliance的成员，将利用其在眼动追踪技术方面的领先地位来开拓增强现实市场。眼动追踪被认为是释放增强现实潜力的最有前景的技术之一，它通过直观的控制增强了用户体验，并具有检测认知负荷和情绪状态的能力。Viewpointsystem已经将眼动追踪技术应用于实际场景，并为未来的大众市场AR设备做好了准备。

苹果在2024年10月29日发布了visionOS 2.1的正式版本。注册开发者可以通过设备设置进入软件更新，选择“测试版更新”来下载和安装测试版。visionOS 2可以从2D照片中提取深度信息，并运用先进的机器学习技术转换为3D照片，以增加照片的深度感。此外，新版本提供了新的手势来激活主视图和控制中心，并且可以对主视图进行自定义，重新排列应用程序。旅行模式增加了对火车的支持，并且访客用户配置文件可以保存30天，不需要每次都进行设置。另外，值得一提的是Mac虚拟显示屏将支持更高分辨率和更大尺寸的虚拟显示，相当于两台并排放置的4K显示器。

2024年青岛国际虚拟现实大会将于11月12日-13日举行，并举办供需对接专题活动。届时，来自XR产业链上下游各个环节的近百家单位将围绕产业供需面临的机遇与挑战进行分享和沟通。供需对接活动分为A场和B场，A场聚焦XR内容应用子生态，B场聚焦器件终端子生态。活动将邀请来自京东方、咪咕新空、小鸟看看、万有引力、Unity中国、世界超高清视频产业联盟、渲光科技等20家企业进行演讲，并围绕终端产业链、芯屏终端发展趋势、3D空间应用与沉浸视听生态、人工智能与眼镜头显融合创新等议题进行讨论。此外，参会企业可以通过扫描二维码报名参会，并与其他企业进行供需对接。详细信息请访问提供的链接。

《Apple专利|用于头部到躯干方向的空间音频再现》描绘了一种基于头部取向数据来确定头部到源取向和头部到躯干取向的方法，通过这种方法可以确定头部到躯干的方向，进而应用于双耳音频滤波器来产生音频输出信号，从而实现准确的空间音频再现。《Apple专利|用于基于运动的色彩校正电子设备》披露了一种基于照明条件和检测到的运动来自动校正电子设备颜色的方法，其中颜色适应速度可根据照明条件和检测到的运动量进行调整。《Apple专利|用于透视显示内容的动态亮度范围》展示了一种动态调整头戴式设备透视显示内容亮度范围的方法，该方法根据用户对显示器的适应程度来动态调整最大允许亮度，以实现透视显示内容的更好展示效果。《Samsung专利|显示设备》描述了一种具有栅电极和电容器的显示设备，其中栅电极设置在有源图案上，电容器设置在栅电极上形成电容器单元。以上是其中一些AR/VR专利的摘要和要点。三星专利揭示了一种像素电路和显示设备，其中包括发光元件、第一晶体管和第一电容器等组件，以及响应写入栅极信号的第二晶体管。此外，该专利还介绍了一种显示设备和驱动方法，通过使用近红外光、传感器像素和眼睛特征模型实现眼睛追踪和高分辨率显示。另一方面，该专利还介绍了一种亚像素和显示设备，其中包括发光元件、第一晶体管和第二晶体管。此外，索尼专利涵盖了导光板、信息处理设备、点云压缩和头戴式显示器等技术。Snap的专利则涵盖了眼动追踪校准、增强现实内容排名和基于生成模型输出的XR体验等技术。谷歌专利涉及到反射波导、眼内显示、图像校准和基于虚拟现实的购物等领域。最后，微软专利涉及到移动平台追踪、深度图生成和动态环境信息智能放置等技术。这些专利反映了XR行业的创新和发展趋势。1. "Snap Patent | Projecting images in augmented reality environments based on visual markers" describes a method for projecting images in augmented reality based on visual markers. It involves using a digital image sensor on a computing device to detect visual markers and determine their position and boundaries, and then display the images on a digital display, creating the illusion of the images being projected onto the physical space. 2. "Snap Patent | Texture generation using multimodal embeddings" describes a method for generating texture in augmented reality. It involves storing interactive data representing one or more interactions in a multimodal memory, detecting objects depicted in images captured by an interactive client, generating prompts based on the objects and the stored interactive data, generating artificial texture based on the prompts, and modifying the texture of the objects depicted in the images based on the artificial texture. 3. "Snap Patent | Product image generation based on diffusion model" describes a system for generating product images based on a diffusion model. It involves receiving images depicting real-world objects, generating prompts that include textual descriptions of fashion items, analyzing the images and textual descriptions using a generative machine learning model, generating artificial images depicting fashion items that match the textual descriptions, and replacing the artificial fashion items with real-world product images that match the visual attributes of the artificial fashion items. 4. "Snap Patent | Scannable codes as landmarks for augmented-reality content" describes a method for generating and displaying augmented-reality content based on scannable codes. It involves generating a scan request on a client device, determining the position and orientation of the encoded image in 3D space, defining a reference point on the client device based on the encoded image, accessing media content from a media library based on the reference point, and displaying the media content on the client device based on the reference point. 5. "Snap Patent | Ingestion pipeline for generating augmented reality content generators" describes a pipeline for generating augmented reality content generators. It involves generating a 3D model file of a product, converting the 3D model file to a 3D object file, associating the 3D object file with a product catalog service, and publishing the augmented reality content generator associated with the product. 6. "Snap Patent | 3D content display using head-wearable apparatuses" describes a method for displaying 3D content using head-wearable apparatuses. It involves determining the position of content items that are closest to the curves of the head-wearable apparatus, adjusting the shape of the content items based on the position of the content items and the user's viewpoint, and displaying the adjusted content items on the display of the head-wearable apparatus. 7. "Snap Patent | Alignment of augmented reality components with the physical world" describes a system for aligning augmented reality components with the physical world. It involves obtaining surface plane information, detecting edges in an image, and presenting virtual content with orientation and position based on the edges and the surface plane information. 8. "Snap Patent | Avatar call platform" describes a platform for generating and displaying animated avatars of callers during a call. The facial expressions of the animated avatars match the expressions of the callers during the call. 9. "Qualcomm Patent | Receiver side prediction of encoding selection data for video encoding" describes a method for encoding and decoding video data. It involves receiving encoding selection data indicating the encoding scheme for a third picture based on the encoding scheme used for a second picture, and encoding the second picture based on the encoding selection data. 10. "Qualcomm Patent | Receiver selected decimation scheme for video coding" describes a method for encoding and decoding video data. It involves receiving a decimation mode indicator indicating a decimation mode for a second group of pictures based on the decimation mode used for a first group of pictures, and encoding the second group of pictures based on the decimation mode. 11. "Qualcomm Patent | V-dmc displacement vector integer quantization" describes a method for decoding encoded grid data. It involves determining a set of coefficients based on the encoded grid data, applying inverse scaling to the coefficients using integer arithmetic to obtain a set of quantized coefficients, determining a displacement vector based on the quantized coefficients, deforming a base grid based on the displacement vector, and outputting the decoded grid. 12. "Qualcomm Patent | Distributed video coding using reliability data" describes a method for encoding and decoding video data. It involves generating prediction data for pictures based on previous reconstructed pictures, generating encoded video data based on the prediction data, scaling the coefficients of transform blocks of the encoded video data based on reliability values, generating error correction data based on the encoded video data, and reconstructing pictures based on the error correction data. 13. "Qualcomm Patent | Entropy continuation and dependent frame entropy coding in point cloud compression" describes a method for compressing point cloud data. It involves sending or parsing a slice level flag indicating the entropy coding status of the first slice of a current frame, and encoding or decoding the first slice based on the entropy coding status indicated by the slice level flag. 14. "Qualcomm Patent | Companion device assisted multi-view video coding" describes a method for encoding and decoding multi-view video data. It involves generating a first group of pictures based on a first set of view pictures, sending the first encoded video data to a receiving device, receiving a multi-view encoding hint from the receiving device, generating a second group of pictures based on a second set of view pictures, encoding the second group of pictures based on the multi-view encoding hint, and sending the second encoded video data to the receiving device. 15. "Qualcomm Patent | Channel estimation based on samples of downlink transmission" describes a method for estimating the channel state based on downlink transmission samples. It involves sending a first downlink transmission and receiving uplink transmission including channel state information for the first downlink transmission. Based on the channel state information, a second downlink transmission is sent. 16. "Qualcomm Patent | Pose optimization for object tracking" describes a method for optimizing the pose of an object. It involves obtaining the pose of the object in a world coordinate system, obtaining an image of the object from a camera position, obtaining a transformation from the world to the camera position, determining reprojection error based on the pose, image, and transformation, and adjusting the pose based on the reprojection error. 17. "Magic Leap Patent | Systems and methods for sign language recognition" describes a method for recognizing and interpreting sign language using a mixed reality device. The system can recognize and translate sign language gestures and present the translated information to the user of the mixed reality device. 18. "Magic Leap Patent | Projector architecture incorporating artifact mitigation" describes a projector architecture for a head-mounted display. The architecture includes a substrate with a set of filters, each capable of transmitting a different wavelength range, and waveguides to guide light to the user's eyes. 19. "Magic Leap Patent | Near-field audio rendering" describes a method for rendering audio with near-field effects using a mixed reality device. The method involves determining the source position of an audio signal, determining the virtual speaker positions, determining the head-related transfer function (HRTF) for each virtual speaker position and ear, and applying the HRTF to the audio signal for each ear. 20. "Magic Leap Patent | Dynamic browser stage" describes a method for displaying 3D content in a spatial environment. The method involves receiving a request to display 3D content at a specific position in the spatial environment, determining if the position is within an authorized area, expanding the authorized area based on user adjustments, and displaying the 3D content within the expanded authorized area. 21. "Magic Leap Patent | Gaze timer based augmentation of functionality of a user input device" describes a method for temporarily modifying the functionality of a handheld controller or other user input device based on the user's gaze. For example, the functionality of specific buttons on the controller may be modified when the user looks at the controller. 22. "Magic Leap Patent | Matching meshes for virtual avatars" describes a method for matching the base mesh of a virtual avatar with a target mesh. The method automatically matches the base mesh with the target mesh by using rigid transformations in high-error areas and non-rigid deformations in low-error areas. 23. "HTC Patent | Head-mounted display and method for image processing based on diopter adjustment" describes a head-mounted display and a method for processing images based on diopter adjustment. The method includes receiving a command indicating a diopter setting and rendering an image based on the diopter setting. These summaries are a brief overview of the patents mentioned, providing a general understanding of their subject matter.

苹果公司在一份专利申请中提出了一种使用图像数据和超声数据的混合注视追踪方法。该方法利用各种组合方式使用超声数据和成像数据来实现精准注视追踪，并以较低的功耗率运作。相比单独使用其中任一种方法，同时使用图像数据和超声数据的注视追踪更稳健，因为成像数据受温度、压力和气流的影响较小，而超声数据受到杂散光的影响较小。 该专利申请中描述了实施例，其中使用传感器识别眼球的位置和眼球运动，并利用眼球的深度信息来确定眼睛的位置/方向。通过同时使用超声数据和光测量数据的不同组合，可以有效对抗环境噪声，获得所需的注视追踪精度和功耗率。 此外，该专利申请还介绍了一个头显系统，其中包括注视追踪系统，可以检测眼球位置和眼球运动，并进行注视追踪。头显系统通过组合成像系统和超声系统的数据来提供更稳健的注视追踪功能。 总的来说，苹果公司的这项专利申请提出了一种创新的混合注视追踪方法，有望提高AR/VR头显的眼动追踪精度和稳定性。

研究人员在一项研究中展示了神经辐射场（NeRF）在移动系统中的应用瓶颈，并引入了名为CICERO的解决方案。他们通过两种算法和一种片上数据布局策略解决了NeRF的性能问题。CICERO的纯软件实现在移动Volta GPU上提供了8倍的加速和7.9倍的节能，与专用DNN加速器基线相比，加速和能量减少增加到28.2倍和37.8倍。研究人员还提出了稀疏辐射扭曲（SPARW）方法，通过重复使用前一帧计算的射线辐射来减少NeRF中的计算工作量。他们还优化了内存访问和数据布局，通过以内存为中心的顺序访问和通道主布局来解决不规则的DRAM访问和SRAM组冲突。最终，他们将CICERO与三种先进的NeRF模型集成，实现了数量级的加速和节能。 以上是一项关于神经辐射场在移动系统中应用瓶颈及其解决方案的研究。通过引入算法和架构的改进，并优化内存访问和数据布局，研究人员成功提高了NeRF的性能，并在移动系统上实现了较大的加速和节能。

索尼提出了一种机器人快速眼动模拟人眼扫视的技术，通过调整图像捕获时间和生成的图像序列时间压缩，使得机器人的眼睛能以人眼的速度进行扫视。这种技术解决了机器人眼球运动速度不足的问题，可以用于头显测试和实验。该技术利用机器学习训练的人工神经网络和人工智能模型来处理数据以进行推断，模型包括输入层、输出层和中间的多个隐藏层。通过与传感器相关联的致动器控制机器人眼在插槽中移动，并通过摄像头和照明器进行图像捕获和显示。在机器人测试或校准眼动追踪系统时，可以通过拍摄机器人眼球运动作为慢动作现象来模拟快速眼球运动，并以更快的动作播放所产生的视频。这种方法可以虚拟地再现扫视，同时节省时间。索尼的专利申请名为“Reproducing fast eye movement using imaging of robot with limited actuator speed”，最初于2023年4月提交，并已公布。

美国加州大学团队提出了一种可穿戴触觉人造肌肉皮肤（HAMS），可以集成到XR系统中，增强沉浸感并在娱乐、教育和辅助技术中应用。HAMS基于全软毫米级多层介电弹性体致动器（DEA），能产生明显的面外变形和压痕，实现复杂的触觉反馈，并具有高感知精度。传统的触觉致动器在形状参数和特定身体部位的限制方面缺乏通用性，限制了用户的沉浸式体验。HAMS提供了一个有前景的替代方案，可以灵活适应不同身体部位，减少侵入性，并实现与虚拟环境的自然交互。HAMS使用全软毫米级多层介电弹性体致动器，可以产生大的面外驱动，实现高位移和高力，并保持高性能。HAMS可以无缝地附着在皮肤上，提供复杂的触觉反馈，具有超过90%的感知精度。HAMS的柔软结构和可穿戴性使其具有扩展性和定制性，适用于不同的皮肤表面，并可以覆盖整个身体。团队表示，将HAMS应用于用户身上并与VR头显相结合，可以提供身临其境的XR体验。