Voice Registration via Clustering Historical Voice Data

Solution Overview

Problem

Current voice registration processes are inefficient and prone to failure due to the need for repeated recitation of fixed sentences, leading to low registration efficiency and success rates.

Innovation Solution

A method that involves acquiring historical voice data, clustering voice feature vectors, and training user voice models based on high-frequency user conditions, allowing for automatic identity association and registration upon matching voice features.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the speaker actively performs voice registration by repeatedly saying fixed sentences, then the voice interaction device can extract voice features and provide personalized service, but the registration process takes a long time and has low efficiency

Engineering Contradiction:

Improvevoice registration success rateVSAvoidregistration time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary voice feature extraction and clustering on historical voice data before actual registration is needed. By pre-processing and organizing voice features into clusters, the system prepares the data structure in advance, so that when registration is needed, the matching process is already optimized and ready, significantly reducing the time required during actual registration.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system automatically performs voice feature clustering and model training without requiring active user participation. The voice interaction device autonomously processes historical voice data, identifies high-frequency users, and trains user-specific voice models in the background, eliminating the need for users to manually repeat fixed sentences multiple times.

Inventive Principle:

Principle #25Self-service

2Adaptability or versatility

If the speaker actively performs voice registration with fixed sentences, then the system can recognize the speaker and provide personalized service, but the registration process is complex and requires multiple repetitions

Engineering Contradiction:

Improvepersonalized service capabilityVSAvoidregistration process complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system automatically performs voice feature clustering and model training without requiring active user participation. The voice interaction device autonomously processes historical voice data, identifies high-frequency users, and trains user-specific voice models in the background, eliminating the need for users to manually repeat fixed sentences multiple times.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system changes the approach from requiring users to repeat fixed sentences to automatically clustering voice features based on acoustic parameters. By transforming the registration process into an automatic parameter-based clustering task, the system reduces procedural complexity while maintaining the ability to provide personalized services.

Inventive Principle:

Principle #35Parameter changes

3Quantity of substance

If the system uses traditional voice registration methods, then it can collect voice features, but the registration success rate is low due to user carelessness and deviation from fixed sentences

Engineering Contradiction:

Improvevoice feature dataVSAvoidregistration success rate

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The system automatically performs voice feature clustering and model training without requiring active user participation. The voice interaction device autonomously processes historical voice data, identifies high-frequency users, and trains user-specific voice models in the background, eliminating the need for users to manually repeat fixed sentences multiple times.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system uses historical voice data as feedback to continuously improve voice models. By analyzing past voice interactions and clustering similar features, the system creates a feedback loop that refines voice recognition accuracy over time, thereby increasing registration success rates without requiring perfect user performance during registration.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11189263B2Voice data processing method, voice interaction device, and storage medium for binding user identity with user voice model

Publication Date: 2021.11.30 TENCENT TECHNOLOGY (SHENZHEN) CO LTD

AI summary

A voice data processing method includes acquiring historical voice data, acquiring historical voice feature vectors corresponding to the historical voice data, and performing clustering on the historical voice feature vectors to obtain a voice feature cluster, the voice feature cluster comprising at least one historical voice feature vector with a similar feature. The method also includes, when the voice feature cluster matches a high-frequency user condition, training a corresponding user voice model according to the historical voice feature vectors contained in the voice feature cluster; after a current voice feature vector of the current voice data matches the user voice model, initiating a user identity association request associated with the current voice data; and, after a response message corresponding to the user identity association request is received, binding user identity information in the response message to the user voice model.