Conference Call Transcript Surfacing for Overlapping Speech
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conference calls with overlapping audio from multiple participants lead to chaotic and unintelligible audio streams, making it difficult for participants to understand who is speaking and what is being said, causing confusion, frustration, and hindering effective communication and collaboration.
Innovation Solution
An electronic device detects overlapping audio and visually identifies the contributing participants on a conference call dashboard, rendering text transcriptions proximal to each participant's region, and provides an overlap warning to promote respectful communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple participants speak simultaneously in a conference call, then the quantity of communication increases, but the audio stream becomes chaotic and unintelligible
Solution Approach 1:
The patent segments the audio stream by detecting overlapping speech from multiple participants and separating their contributions. The system identifies individual speakers during overlap periods and routes their audio to separate output channels, allowing participants to hear distinct speech streams rather than a chaotic mixture, thereby maintaining audio intelligibility while supporting simultaneous communication.
Solution Approach 2:
The patent introduces an intermediary speech detection and analysis system that processes the combined audio stream from multiple participants. This intermediary system identifies overlapping speech, determines which participants are speaking simultaneously, and manages the separation and routing of their audio contributions, resolving the conflict between allowing simultaneous communication and maintaining audio clarity.
2Loss of information
If audio overlap detection and transcription features are added to identify speakers during overlaps, then communication clarity improves, but device complexity increases
Solution Approach 1:
The patent implements self-service by having each participant's device independently detect and analyze its own received audio stream for overlapping speech. Each device performs speech detection, speaker identification, and transcription generation locally, reducing the need for complex centralized processing and distributing the computational burden across multiple participants' devices.
Solution Approach 2:
The patent replaces complex mechanical audio processing systems with software-based speech detection and analysis algorithms. By using automated speech recognition and machine learning models running on standard computing devices, the system achieves sophisticated speaker identification and overlap detection without requiring specialized hardware, thereby managing device complexity.
Data Source
AI summary
A method provides techniques for detecting, by a processor of an electronic device, a connection to a conference call comprising multiple of call participants. The method includes rendering a conference call dashboard on a display of the electronic device. The conference call dashboard includes a participant region for each participant of the plurality of call participants. The method further includes detecting overlapping audio originating from at least two call participants. The method further includes, in response to detecting the overlapping audio, presenting, on the display, a text transcription of a respective audio of each of the at least two participants identified.