Acoustic Voiceprint Diarization for Faster Speaker Labeling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing diarization systems are time-intensive and inefficient due to the need for manual separation of speech from identified speakers and limitations in live interactions, which affect the accuracy of speaker separation and transcription.

Innovation Solution

A method and system that utilize acoustic voiceprints to identify and label speakers in audio data by selecting subsets of audio files that maximize acoustic differences, applying these voiceprints to new audio files for accurate speaker identification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual separation of speech from identified speakers is used, then speaker identification can be achieved, but the process becomes time-intensive and inefficient

Engineering Contradiction:

Improvespeaker identification accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual speech separation with acoustic modeling and voiceprint matching. Acoustic models automatically separate and identify speakers by comparing audio features against trained voiceprints, eliminating the need for manual audio segmentation while maintaining identification accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system creates acoustic voiceprints (copies of speaker characteristics) from training speech and uses these copies to identify speakers in new audio. This allows automatic matching without manually separating speech, as the voiceprint copies enable direct comparison and identification.

Inventive Principle:

Principle #26Copying

2Measurement precision

If recorded training speech is used to create acoustic models, then speaker identification is possible, but the speaker speaks differently than in live interactions

Engineering Contradiction:

Improvespeaker identification accuracyVSAvoidlive interaction adaptability

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent implements dynamic adaptation where acoustic models are initially trained on recorded speech but then adapted during live interactions. The system continuously refines voiceprints by incorporating live speech data, allowing the models to adjust to the speaker's actual speaking style in real-time interactions while retaining the foundation from recorded training.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system performs preliminary training on recorded speech to create initial voiceprints, then uses these pre-trained models as a starting point for live interaction. This preliminary action provides a head start, and the models are subsequently fine-tuned during actual use to match live speaking patterns.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12518761B2Diarization using acoustic labeling

Publication Date: 2026.01.06 VERINT SYST INC

AI summary

Systems and method of diarization of audio files use an acoustic voiceprint model. A plurality of audio files are analyzed to arrive at an acoustic voiceprint model associated to an identified speaker. Metadata associate with an audio file is used to select an acoustic voiceprint model. The selected acoustic voiceprint model is applied in a diarization to identify audio data of the identified speaker.