Passive RF Positioning With Zero-Shot Multi-Target Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing wireless communication systems, particularly 5G NR, face challenges in accurately identifying and counting multiple moving objects within a coverage area, especially in indoor environments, due to the difficulty in obtaining on-site multi-person training data for passive positioning systems.

Innovation Solution

A passive positioning system is trained using single-object data to identify multiple moving objects, employing deep learning methods and source separation techniques to process Wi-Fi/Bluetooth/UWB communication channels, enabling unsupervised people counting and precise positioning on 2D/3D maps.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multi-object training data is used for passive positioning systems, then the system can identify multiple moving objects, but obtaining such training data becomes impractical and difficult

Engineering Contradiction:

Improvemulti-object identification capabilityVSAvoidtraining data acquisition

Core Design Contradiction:

Adaptability or versatilityVSEase of manufacture

Solution Approach 1:

The patent segments the complex multi-object training problem into simpler single-object training instances. By processing RF channel data for each object individually and using source separation techniques, the system breaks down the intractable multi-object training task into manageable single-object tasks that can be executed independently and then combined.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces source separation techniques as an intermediary mechanism between single-object training data and multi-object identification. This intermediary process enables the system to derive multi-object capabilities from single-object training data by separating and identifying individual object signatures in the RF channel data.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If deep learning methods are employed to process RF channel data, then the system achieves practical and accurate identification of moving objects, but the complexity of the system increases

Engineering Contradiction:

Improveobject identification accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-processing RF channel data through source separation techniques before applying deep learning methods. This preliminary step organizes and structures the data in a way that simplifies subsequent deep learning processing, reducing the overall system complexity while maintaining high identification accuracy.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12495383B2Zero-shot deep learning for multi-target RF positioning

Publication Date: 2025.12.09 QUALCOMM INC

AI summary

Aspects presented herein may enable a passive positioning system, which may be a network entity or node, to be trained to identify multiple moving objects based on using training data for a single object. In one aspect, a network entity receives first RF channel data recorded by a set of devices for a coverage area during a first time period. The network entity trains an ML model based on the set of devices and the first RF channel data. The network entity receives second RF channel data recorded by the set of devices at a second time instance that is outside of the first time period. The network entity computes a number of moving objects in the coverage area at the second time instance based on the second RF channel data using the ML model.