Conference Audio Time-Contiguous Containers for Real-Time Transcription

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current conferencing systems process audio data after the conference, preventing real-time transcription and AI/ML feedback, and fail to accurately associate audio payloads with speakers.

Innovation Solution

A conferencing server generates time-contiguous containers with speaker identification, transmitting them to a consumer server for real-time processing, enabling immediate transcription and AI/ML inference during the conference.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If audio data is processed after the conference, then processing complexity is reduced, but real-time transcription and AI/ML feedback cannot be provided

Engineering Contradiction:

Improvereal-time processing capabilityVSAvoidprocessing system complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the audio data processing into multiple time-contiguous containers, each associated with a specific time frame and speaker identifier. This segmentation enables the system to process audio data in real-time chunks rather than handling the entire recording at once, thereby achieving real-time transcription and AI/ML feedback while managing processing complexity through structured data organization.

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If audio data is not associated with speakers, then processing is simpler, but accurate transcription and feedback cannot be provided

Engineering Contradiction:

Improvespeaker identification accuracyVSAvoiddata structure complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges the audio payload with metadata identifying the speaker and time frame within the time-contiguous container structure. This combination allows the system to maintain accurate speaker identification and attribution while processing the audio data, enabling precise transcription and feedback without requiring separate complex tracking mechanisms.

Inventive Principle:

Principle #5Merging (Combining)

3Loss of time

If processing occurs after conference, then system resources are saved, but real-time engagement and feedback are lost

Engineering Contradiction:

Improveprocessing delayVSAvoidcomputational resource consumption

Core Design Contradiction:

Loss of timeVSUse of energy by moving object

Solution Approach 1:

The patent implements preliminary processing actions by pre-segmenting and preparing audio data into time-contiguous containers with associated metadata before the actual transcription and AI/ML processing occurs. This preparation enables the system to process audio data more efficiently in real-time, reducing processing delay while managing computational resource consumption through optimized data structures.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250106269A1Processing Conference Audio Data

Publication Date: 2025.03.27 ZOOM COMMUNICATIONS INC

AI summary

A conferencing server receives audio data from devices connected to a conference. The conferencing server generates multiple time-contiguous containers. Each time-contiguous container includes an identifier of an associated device of the devices and one or more payloads of the audio data from the associated device. Each payload has a predefined time length. The conferencing server transmits the multiple time-contiguous containers to a consumer server for processing. Based on the identifier and the payloads, the consumer server performs at least one of generating a transcript or obtaining intelligence.