Fujitsu advances AI in radio access networks for real-time video analytics

Fujitsu, in collaboration with NVIDIA, is advancing the integration of AI into radio access networks (RAN) to enhance real-time video analytics capabilities. This integration enables applications such as predicting criminal behavior and detecting falls among vulnerable individuals.

The company showcased a number of applications powered by the technology at the Mobile World Congress in Barcelona this week.

Radio access networks are essential components of mobile telecommunications systems, connecting individual devices to the broader network via radio connections. By incorporating AI, these networks can achieve reduced latency, which is crucial for applications requiring immediate responses. ​

Fujitsu has been at the forefront of developing virtualized RAN (vRAN) solutions that leverage AI to optimize network performance and support advanced applications. By integrating NVIDIA’s GPU processing engines and AI frameworks, Fujitsu’s vRAN technology enables the parallel processing of base station communications and AI-driven applications on a unified platform. This architecture facilitates real-time video analytics which can be utilized for various purposes, including public safety and health monitoring. ​

In November 2024, Fujitsu and SoftBank announced a strengthened partnership to accelerate the commercialization of AI-RAN technologies by 2026. This collaboration involves joint research and development of vRAN software that integrates AI to maximize RAN performance and enhance communication quality.

A significant aspect of this partnership is the establishment of a verification lab in Dallas, Texas that is dedicated to validating hardware, software, and applications constituting the AI-RAN framework. ​

Central to this technological advancement is NVIDIA’s AI Aerial platform, which provides a software-defined, accelerated computing infrastructure capable of handling both RAN and AI workloads. This platform supports dynamic allocation of resources, allowing for efficient processing of AI applications such as video analytics. By deploying AI applications on the same infrastructure as 5G networks, service providers can offer innovative services like real-time video monitoring to detect suspicious behavior or monitor the safety of individuals in various environments. ​

The integration of AI into RANs opens new possibilities for public safety and health monitoring. For instance, real-time video analytics can detect unusual behaviors that may indicate criminal intent, enabling proactive responses from law enforcement. Similarly, the technology can monitor environments such as assisted living facilities to promptly identify if a resident has fallen, ensuring timely assistance. These applications demonstrate the potential of AI-RAN to enhance community safety and individual well-being.​

One notable application of this technology is in closed-circuit television (CCTV) monitoring. The AI system can analyze human skeletal movements from camera footage to interpret actions and predict potential criminal intent or emotional states based on body language. For instance, it can identify aggressive behavior or anticipate a fight before it occurs, allowing security personnel to respond proactively.

Additionally, the system can detect fraudulent activities in retail environments, such as when an individual attempts to scan a cheaper item’s label over a more expensive product at a self-checkout. By recognizing this specific motion, the AI can alert security to intervene promptly. ​

“We can even get to the point where we can interpret emotions, so you can tell where there’s angry body language and not angry body language,” Rob Hughes, head of Fujitsu wireless marketing, told IOT World Today. “One of the use cases here is security cameras. There are CCTV cameras everywhere, but you can’t have human eyes watching everything. This could identify threatening behavior or a fight about to start so you can immediately dispatch security.”

Fujitsu’s partnership with NVIDIA has been pivotal in developing these AI capabilities within RAN. Their joint efforts have led to the creation of AI-native RAN solutions, which utilize NVIDIA’s AI Aerial software stack and accelerated computing platforms. This collaboration has resulted in more efficient and scalable virtualized RAN deployments, capable of handling real-world traffic demands.

Fujitsu has demonstrated the potential of AI-powered video analytics in various sectors and has showcased applications where AI assists in assessing the width of trees in logging operations, aiding workers in making informed decisions about which trees to cut.

The continued evolution of AI-powered RAN technology promises even greater capabilities. As machine learning models become more sophisticated and datasets expand, AI-driven video analytics will improve in accuracy and adaptability. Future iterations of this technology could incorporate multimodal data sources, such as integrating audio analysis alongside video monitoring to provide even more context for decision-making. An AI system capable of recognizing distress calls or unusual auditory cues in conjunction with visual movement analysis would further enhance the accuracy of threat detection and emergency response.

Fujitsu’s pioneering work in AI-native RAN solutions, powered by NVIDIA’s cutting-edge technology, marks a significant leap forward in real-time video analytics and intelligent monitoring. By harnessing the capabilities of AI to detect potential criminal acts, identify vulnerable individuals, and enhance safety in various sectors, the company is setting new standards for what is possible in wireless communication and surveillance.

As this technology matures and finds broader adoption, it has the potential to reshape how societies approach security, healthcare, and public safety in the digital age. The fusion of AI and RAN is not just an incremental improvement, it also represents a fundamental shift toward a smarter, more proactive intervention capability to prevent harm before it occurs.

Article Topics

biometrics  |  cctv  |  criminal ID  |  edge AI  |  Fujitsu  |  MWC  |  Nvidia  |  video analytics  |  video surveillance