Multi-Speaker Voice Recognition via Segmented Transcription Units
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice recognition technologies fail to accurately transcribe multiple speakers during simultaneous conversations, particularly in telecommunication conferences, due to limitations in vocabulary and training requirements, making machine transcription of conferences ineffective.
Innovation Solution
The implementation of individual voice recognition units for each speaker in a conference, merging their outputs on a time basis to produce a textual transcription of the entire conference call, with features like accelerated playback and topic-based alerts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If untrained voice recognition algorithms are used, then they can recognize words from a broad spectrum of population, but they require more time to recognize words and have limited vocabulary
Solution Approach 1:
The system segments the voice recognition task by assigning individual voice recognition units to each speaker. Each unit is trained specifically on that speaker's voice characteristics, allowing for faster and more accurate recognition without sacrificing vocabulary coverage. This segmentation resolves the contradiction by combining the adaptability of untrained algorithms with the speed of trained algorithms through modular speaker-specific processing.
2Loss of time
If trained voice recognition algorithms are used, then they recognize words faster with better accuracy, but they can only recognize words spoken by one individual
Solution Approach 1:
The system divides the multi-speaker recognition problem into separate speaker-specific recognition units. Each unit is trained on a single speaker's voice, achieving fast and accurate recognition for that individual. The system then combines multiple segmented outputs to handle multi-speaker scenarios, resolving the contradiction between training specificity and multi-speaker adaptability.
Solution Approach 2:
The system creates a universal transcription service that works across multiple speakers by combining individual speaker recognition units. Each unit is specialized for one speaker, but the overall system achieves multi-speaker capability through coordinated operation of multiple specialized units, enabling both speed and versatility.
3Measurement precision
If individual voice recognition units are assigned to each speaker, then accurate transcription of multi-speaker conversations is achieved, but device complexity increases
Solution Approach 1:
The system segments the transcription task into speaker-specific recognition units, which improves accuracy by reducing confusion between speakers. While this creates multiple processing units, the modular architecture allows for efficient resource utilization and manageable complexity through standardized interfaces and coordinated timing mechanisms.
4Productivity
If multiple speakers talk simultaneously, then conference productivity is maintained, but prior art voice recognition fails to transcribe accurately
Solution Approach 1:
The system assigns dedicated voice recognition units to each speaker, allowing simultaneous processing of multiple speakers without interference. This segmentation enables accurate transcription during concurrent speech, maintaining conference productivity while achieving precision that prior art systems could not deliver in multi-speaker scenarios.
Data Source
AI summary
Utilizing individual voice recognition units for each speaker in a conference to perform automatic transcription of that speaker's contribution to the conference. The output of each of the voice recognition units is then merged on a time basis to produce a textual transcription of the entire telecommunication conference call.