Split Inference Media Transfer Across Local and Remote Networks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing communication systems face challenges in efficiently transmitting and processing input media data and inference result data across various wireless networks, particularly in scenarios involving local and remote data capture and consumption, leading to inefficiencies and suboptimal performance.

Innovation Solution

The implementation of methods, apparatuses, and systems that facilitate the transmission and processing of input media data and inference result data across wired and wireless networks, utilizing technologies such as CDMA, TDMA, FDMA, OFDMA, and SC-FDMA, along with advanced radio access technologies like LTE, LTE-A, NR, and IEEE 802.11, to optimize data transfer and consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If input media data is transmitted across wireless networks for remote inference, then processing capabilities are improved, but transmission latency and network dependency increase

Engineering Contradiction:
Improveprocessing capabilitiesVSAvoidtransmission latency
Core Design Contradiction:
Extent of automationVSLoss of time

Solution Approach 1:

The patent segments the inference processing into local and remote components. Local inference processing handles time-sensitive operations at the edge device, while remote inference processing handles complex computations over the network. This segmentation allows the system to balance processing capabilities with transmission latency by keeping critical path operations local.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary inference processing architecture that mediates between local and remote processing. The intermediary can buffer, pre-process, or queue data locally before transmission, reducing the impact of network latency on overall processing performance.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If data is transmitted between local and remote endpoints, then system versatility is improved, but communication complexity increases

Engineering Contradiction:
Improvesystem versatilityVSAvoidcommunication complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a universal communication interface that handles multiple transmission modes (local-only, remote-only, hybrid) through a single standardized protocol. This multi-functional interface simplifies the communication complexity by providing a unified approach to diverse transmission scenarios.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent employs dynamic communication routing that adapts the data transmission path based on current system conditions, data type, and performance requirements. The system can dynamically switch between local and remote processing modes, optimizing versatility while managing communication complexity through adaptive decision-making.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4648386A1Methods, architectures, apparatuses and systems for communicating input media data for inference and inference output result data
Publication Date: 2025.11.12 INTERDIGITAL CE PATENT HOLDINGS SAS
  • EP4648386A1 patent drawingFigure 1A
  • EP4648386A1 patent drawingFigure 1B
  • EP4648386A1 patent drawingFigure 1C

AI summary

Procedures, methods, architectures, apparatuses, systems, devices, and computer program products for communicating input media for inference by an inference engine and post-processed output result data that is output from the inference engine, for media consumption. In a split inference model for processing media data, topologies may include local inference, e.g., on a user equipment, with remote media consumption, e.g., in a network, or remote inference with local media consumption, which may be negotiated between the user equipment and the network. Then, input media for inference and output result data are communicated between the user equipment and the network accordingly.