Page 338 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 10: AI-Driven RF-Based Object Detection and

Classification for 5G-Advanced and Beyond

Organization: HFCL Limited, Department/Division/Group: 5G BU

Country: India

Contact Person(s):

Dr. Sonali, sonali@ hfcl �com
Subhas Mondal, subhas�mondal@ hfcl �com

1 Use Case Summary Table

Item Details
Category 5G

Problem Real-time monitoring using AI-powered RF sensing
Addressed

AI enabled Integrated Sensing and communication (ISAC): AI enabled
Key Aspects of
Solution Environmental sensing to create a demography map, AI enabled Object
detection and identification

Technology ISAC, 6G, AI-native RAN
Keywords

Public – mmWave DISC dataset for integrated sensing and communication.
Data Availability Models must be trained across sub-6 GHz and mmWave bands to ensure
robustness across propagation characteristics [1].
Numerical. Metadata Descriptions:
Each RF sample in the dataset should be tagged with relevant metadata
to support model training and validation. This includes:
• Environment type: Urban, rural, indoor, outdoor
Metadata (Type of • Scenario: LoS/NLoS, static/dynamic objects
Data) • Object properties: Material, shape, motion, size
• Signal parameters: Frequency band, SNR, timestamp
• Label info: Object class, location, reflectivity

Metadata enables supervised learning and helps AI models generalize
better across scenarios.

RF-based object detection is trained using raw reflected signal data across
Model Training and varied environments. CNN-based models are evaluated for classification,
Fine-Tuning with emphasis on differentiating LoS/NLoS scenarios and tracking move-
ment patterns.

302