Multimodal Data Sending With Modality Identifiers for Accurate Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing agent interactions in the 6G era are passive and unimodal, leading to inaccurate data processing due to the inability to determine the modality of multimodal data, resulting in errors in processing results.
Innovation Solution
Determine and send an identifier corresponding to the modality of each piece of data to a centralized processing device, enabling accurate processing by identifying the correct processing mode for each modality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multimodal data is sent to centralized processing device without modality identifiers, then data transmission is simplified, but the processing accuracy deteriorates due to inability to determine data modality
Solution Approach 1:
The patent introduces modality identifiers as intermediary elements that mediate between the data and the processing device. These identifiers serve as a bridge that enables the centralized processing device to accurately determine the modality of each data piece without fundamentally changing the data structure or transmission protocol, thus resolving the contradiction between processing accuracy and transmission complexity.
Solution Approach 2:
The patent segments the data transmission process into distinct components: data pieces and their corresponding modality identifiers. By separating the modality information from the data itself and transmitting them as distinct but associated elements, the system achieves accurate modality determination while maintaining relatively simple transmission mechanisms.
2Measurement precision
If unimodal data processing is used, then processing mode is simple, but processing result accuracy deteriorates when handling multimodal data
Solution Approach 1:
The patent implements a dynamic processing mode selection mechanism where the centralized processing device adjusts its processing behavior based on the modality identifiers received with the data. This allows the system to transition between different processing modes (unimodal or multimodal) dynamically, achieving high accuracy for multimodal data while maintaining simplicity for unimodal data processing.
Solution Approach 2:
The patent changes the processing parameters of the centralized processing device based on the modality identifiers. By adjusting processing parameters according to the identified modality (audio, video, text, etc.), the system achieves accurate processing results for diverse data types without requiring fundamentally different processing architectures.
Data Source
Figure 1~3
Figure 4~5
Figure 6~7
AI summary
A multimodal data sending method, comprising: acquiring at least one piece of data, wherein each of the at least one piece of data has a different modality; determining the identifier corresponding to the modality of each piece of data; and sending at least one piece of data and at least one identifier to a centralized processing device. The identifier corresponding to the modality of each piece of data can be determined, such that not only can at least one piece of data be sent to the centralized processing device, but the identifier corresponding to the modality of each piece of data can be sent to the centralized processing device, so as to enable the centralized processing device to determine the modality of each piece of data according to the identifier, thereby accurately processing data according to a processing method corresponding to the modality to obtain an accurate processing result.