Advancing toward practical sensing in the 6G era — anywhere covered by base station signals can become a sensing area, using only communication devices

NTT Corporation (Headquarters: Chiyoda, Tokyo; President and CEO: Akira Shimada; hereinafter “NTT”) and Sophia University have achieved a world-first demonstration of human detection by analyzing variations in radio propagation information extracted from synchronization signals transmitted by commercially operated base stations.

This research was conducted to evaluate the effectiveness of ISAC (Integrated Sensing and Communications), which is expected to be adopted in sixth-generation mobile communication systems (6G). By applying AI analysis to propagation data obtained from base station signals, the team successfully detected human presence without requiring any additional sensors.

These results suggest that integrating sensing capabilities into existing mobile communication systems—originally designed solely for communication—can significantly expand their range of applications.

This technology will be exhibited at Wireless Japan × Wireless Technology Park¹, to be held from May 28 to 30, 2025.

Professor Masakatsu Ogawa from the Department of Information and Communication Sciences at Sophia University is participating in the project.

ISAC enables sensing capabilities using the same radio wave propagation information used for communication, eliminating the need for installing additional dedicated sensing devices. Moreover, since ISAC relies on radio signals rather than optical imaging, it allows for privacy-conscious sensing without capturing images of the target. ISAC is also expected to perform well in environments where conventional cameras struggle—such as in darkness or where line-of-sight is obstructed.

Recognizing these advantages, the 3rd Generation Partnership Project (3GPP⁵) has defined a wide range of ISAC use cases (Table 2), envisioning the deployment of radio-based sensing across diverse indoor and outdoor environments.

However, most existing experimental validations have been limited to indoor use cases using wireless LAN systems. While theoretical studies on ISAC-based wireless sensing in mobile communication systems are increasing, experimental demonstrations remain scarce. Even when such reports exist, they often rely on experimental base stations or controlled test environments, leaving the practical feasibility of 3GPP-defined ISAC use cases unverified.

Against this backdrop, NTT and Sophia University have been conducting R&D on wireless sensing technologies using commercial base stations, based on the envisioned use cases of ISAC (Figure 3). The wireless propagation information acquisition system developed for this demonstration is capable of collecting data from commercially deployed 4G and 5G networks. Accordingly, this demonstration specifically validated ISAC technology using the most widely deployed 4G commercial base stations in real-life environments.

NTT and Sophia University have developed a system that utilizes synchronization signals periodically transmitted from commercial base stations to extract radio propagation information. Using this system, a field demonstration was conducted to sense the number of pedestrians on an outdoor pathway at the Yotsuya Campus of Sophia University.

(1) Experimental Environment

In this demonstration, pedestrian detection in an outdoor setting—one of the ISAC use cases—was conducted on Sophia University’s Yotsuya Campus.

Radio signals transmitted from commercial base stations installed and operating on campus were received using two antennas positioned along a campus walkway. These signals were then analyzed using the developed wireless sensing system (Figure 4).

The radio signals used for analysis came from the 2 GHz Band 1 of a 4G system. From the signals, Received Signal Strength Indicator (RSSI) and Channel State Information (CSI) were extracted and used for sensing.

The measurement period spanned 3.5 hours, from 10:30 AM to 2:00 PM, including both low-traffic lecture hours and high-traffic lunch breaks.

In addition to the wireless sensing system, a camera installed on campus was used to obtain ground truth pedestrian counts. Specifically, the number of people passing through the green-highlighted area shown in Figure 4 was counted from the camera footage⁶, enabling simultaneous collection of radio signal data and pedestrian data.

(2) Experimental Results

■Time Series of Moving Variance of RSSI

Figure 5 shows the number of pedestrians observed (blue line) and the moving variance of RSSI extracted from 4G signals using the wireless sensing system (pink line), measured between 10:30 AM and 2:00 PM.

The graph reveals a clear pattern: fewer pedestrians during lecture hours (when students are in classrooms), and more during lunch breaks (when students move between buildings).

The variance of RSSI showed a trend closely resembling the pedestrian count.

 ■AI-Based Pedestrian Count Estimation

Figure 6 presents results of pedestrian count estimation using deep learning-based AI analysis.

RSSI, representing the strength of received signals, is lightweight in terms of processing and can be measured 100 times per second.

CSI, which captures amplitude and phase information across multiple subcarriers and antennas, provides spatial information but is high-dimensional and computationally intensive—thus limited to about one measurement per second when using synchronization signals from commercial base stations.

Since both temporal and spatial information are important for estimating crowd density, this demonstration used both RSSI and CSI in deep learning.

For preprocessing:

RSSI: Moving standard deviation was calculated over multiple time windows to capture various pedestrian movement patterns.

CSI: Variations across antennas and subcarriers were computed to extract spatial features.

Deep learning models were trained to extract temporal features from RSSI and spatial features from CSI, which were then combined to estimate pedestrian counts.

Outdoor environments are subject to noise such as wind-induced movement, which can degrade model generalization.

To address this, data augmentation techniques were applied by adding noise to the observed values and ground truth labels⁷, helping to prevent overfitting⁸.

Figure 6 shows that using moving standard deviation of RSSI, spatial information from CSI, and data augmentation significantly reduced estimation error—cutting it by half.

NTT:

Developed the wireless sensing system to measure synchronization signals transmitted from operational commercial base stations. Verified the effectiveness of wireless sensing technology using deep learning.

Sophia University:

Provided the experimental environment (Yotsuya Campus, Sophia University), performed pedestrian counting using cameras, and created the pedestrian count dataset.　Evaluated useful RSSI features for estimating pedestrian numbers.

This demonstration successfully showcased the practicality of ISAC (Integrated Sensing and Communications), where sensing can be achieved solely through communication devices within the coverage area of commercial base station signals. Since the wireless sensing system developed in this study can utilize both 4G and 5G signals, ISAC can be demonstrated in real-world environments even before the rollout and widespread adoption of 6G.

Moving forward, we will continue to conduct ISAC demonstration experiments for various real-life use cases using commercial 4G and 5G signals, further advancing ISAC technology research and development. The results obtained will be submitted to 3GPP to promote the standardization and practical implementation of ISAC in 6G. In addition, we plan to explore technologies for integrating and analyzing synchronization signals from base stations of multiple mobile operators, aiming to further enhance sensing accuracy.

Our goal is to verify and explore service applications using ISAC technology toward its realization around 2030, when 6G is expected to be widely deployed.

Research-related inquiries

Department of Information and Communication Sciences, Sophia University

Professor Masakatsu Ogawa　(E-mail：m-ogawa@sophia.ac.jp)

Press Inquiries

Office of Public Relations（sophiapr-co@sophia.ac.jp)