Singapore Researchers Develop 'MaskFi' for Real-Time Human Tracking in Metaverse
Summary:
Researchers from Singapore's Nanyang Technological University have developed a novel technique for tracking human activity in real time for the metaverse using Wi-Fi sensing and artificial intelligence. They have introduced an AI-based system, "MaskFi," which overcomes the limitations of existing methods based on physical sensors or cameras. Trained through unsupervised learning, MaskFi uses unlabeled video and Wi-Fi activity data, reaching an accuracy rate of 97%. This innovative approach could pave the way for a real-time, exact metaverse replication of the real world.
Scientific investigators at Nanyang Technological University based in Singapore have recently revealed a novel approach for tracking real-time human activity for the metaverse. A significant aspect of the metaverse is the power to mirror real-world objects and individuals within the online space instantly. Within the realm of digital reality, users have the capability to turn their heads to change their perspective or use real-world controllers to impact the online environment. The existing way of capturing human movements in the metaverse involves device-centric sensors, cameras, or a blend of both. However, the researchers noted serious limitations with these options in their preprint study. Related: Sam Altman, OpenAI facing lawsuit filed by Elon Musk over breached agreement. A device-dependent sensing network, like a motion sensor in a handheld controller, can only gather information from a single point on the human body, thereby limiting complex activity capture, as the researchers state. Furthermore, camera-focused tracking systems suffer from issues related to dim lighting conditions and physical obstructions. This is where WiFi sensing comes into play. For several years, scientists have used WiFi sensors to monitor human actions. Similar to RADAR, the radio signals used in the transmission and reception of WiFi data can be used to detect objects in an environment. They can be adjusted to monitor heart rates, track sleeping and breathing patterns, and even detect people through walls. Researchers in the field of the metaverse have previously tried merging WiFi sensing with conventional tracking methods with mixed results. This also leads us to the realm of artificial intelligence. WiFi tracking calls for the application of artificial intelligence models. Unfortunately, the process of training these models has often posed stiff challenges for researchers. As per the paper by the team from Singapore: "Existing techniques using WiFi and visual modalities rely on a large volume of labeled data, which is difficult to collect. ...We propose a unique unsupervised multimodal HAR solution, behind the name MaskFi, that needs only unlabeled video and WiFi activity data for model training.” To train the necessary AI models to experiment with WiFi sensing for Human Activity Recognition (HAR), scientists need to create a vast database of training data. These datasets used for training AI could represent thousands or even millions of data points, based on the specific aims of each model. Labeling these datasets often becomes the lengthiest part of these investigations. That's where MaskFi comes in. The team from Nanyang Technological University developed the "MaskFi" system to tackle this predicament. This system applies AI models created through a method known as "unsupervised learning". In unsupervised learning, an AI model is initially trained on a smaller dataset and then repeatedly revised until it can project output states with a reasonable degree of accuracy. This process enables scientists to direct their focus on the models, rather than investing time in creating robust training datasets. Source: Yang, et. al., 2024 As per the investigators, the MaskFi system accomplished almost 97% precision across two associated benchmarks. This suggests that, following more development, this mechanism could act as a springboard for a radically new metaverse modality: a metaverse offering a real-time and perfect replication of the real world.
Published At
3/1/2024 8:23:40 PM
Disclaimer: Algoine does not endorse any content or product on this page. Readers should conduct their own research before taking any actions related to the asset, company, or any information in this article and assume full responsibility for their decisions. This article should not be considered as investment advice. Our news is prepared with AI support.
Do you suspect this content may be misleading, incomplete, or inappropriate in any way, requiring modification or removal?
We appreciate your report.