Speaker Feature Mapping for Registration-Free Voice Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing voice processing systems require pre-registration of user voices and identification information for personalized services, limiting their ability to predict user relationships and states accurately.

Innovation Solution

A method and apparatus that extracts speaker feature vectors from voice data, generates a speaker feature map, and classifies speakers based on positional relationships without prior registration, using deep neural networks to form clusters and predict user relationships and states.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If pre-registration of user voices and identification information is required, then user recognition accuracy is improved, but system complexity and user burden increase

Engineering Contradiction:
Improveuser recognition accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs self-service by automatically extracting speaker features and forming clusters without requiring user registration. The speaker feature map is generated autonomously from input speeches, and the system self-organizes speaker groups based on feature similarity, eliminating the need for manual user setup while maintaining recognition capability

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary action by pre-forming speaker clusters and generating the speaker feature map in advance, before actual user recognition is needed. This allows the system to have recognition capability ready without requiring individual user registration, as the clustering structure is prepared beforehand from available speech data

Inventive Principle:
Principle #10Preliminary action

2Adaptability or versatility

If pre-registration processes are implemented, then personalized service capability is improved, but ease of operation deteriorates

Engineering Contradiction:
Improvepersonalized service capabilityVSAvoidease of operation
Core Design Contradiction:
Adaptability or versatilityVSEase of operation

Solution Approach 1:

The system provides personalized service through self-service mechanisms where the speaker feature map and clusters are automatically generated from input speeches without user intervention. The system autonomously identifies speaker characteristics and organizes them into personalized groups, delivering adaptive service while maintaining operational simplicity

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary clustering and feature map generation before personalized service delivery is actually requested. This advance preparation enables immediate personalized service response without requiring users to go through registration processes at the moment of service need

Inventive Principle:
Principle #10Preliminary action

3Ease of operation

If speaker clustering is performed without registration, then ease of operation is improved, but measurement precision may deteriorate

Engineering Contradiction:
Improveease of operationVSAvoidspeaker classification accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The system changes parameters by using continuous speaker feature vectors (such as MFCC, pitch, energy) extracted from speech signals to represent speakers, rather than discrete registered user IDs. The clustering algorithm processes these continuous parameters to form groups, maintaining accuracy while enabling operation without registration

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system replaces the mechanical registration process with an automated feature extraction and clustering mechanism. Instead of manually registering users with identification information, the system automatically extracts acoustic features, computes speaker representations, and performs unsupervised clustering to achieve accurate speaker classification without registration

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3857546B1Method and apparatus for processing voice data of speech
Publication Date: 2025.10.22 SAMSUNG ELECTRONICS CO LTD
  • EP3857546B1 patent drawingFigure 1
  • EP3857546B1 patent drawingFigure 2A~2B
  • EP3857546B1 patent drawingFigure 3A

AI summary

A method and apparatus for processing voice data of a speech received from a speaker are provided. The method includes extracting a speaker feature vector from the voice data of the speech received from a speaker, generating a speaker feature map by positioning the extracted speaker feature vector at a specific position on a multi-dimensional vector space, forming a plurality of clusters indicating features of voices of a plurality of speakers by grouping at least one speaker feature vector positioned on the speaker feature map, and classifying the plurality of speakers according to the plurality of clusters.