Personalized Speech Recognition Model Adaptation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing speech recognition technologies struggle to achieve high accuracy for individual users, as they are typically trained on general data and do not adapt effectively to personal speech patterns.

Innovation Solution

A processor-implemented method that obtains feedback data from recognizing a user's speech input, determines whether to update a speech recognition model based on this feedback, and selectively updates the model to personalize it for the user.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a speech recognition model is trained on general data, then it can recognize speech from multiple users, but it cannot achieve high accuracy for individual users with personal speech patterns

Engineering Contradiction:

Improveability to recognize speech from multiple usersVSAvoidrecognition accuracy for individual users

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The speech recognition system is segmented into a general speech recognition model trained on diverse data for multi-user capability, and a personalized model component that adapts to individual users through feedback. This segmentation allows the system to maintain versatility while improving individual user accuracy through specialized personalization modules.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The speech recognition model is made dynamic by enabling continuous updates through feedback mechanisms. The model evolves over time by incorporating user-specific feedback data, allowing it to adapt its parameters and improve personalization while maintaining its base functionality for general speech recognition.

Inventive Principle:

Principle #15Dynamics

2Measurement precision

If feedback data is collected and used to update the speech recognition model, then recognition accuracy improves, but system complexity and computational requirements increase

Engineering Contradiction:

Improvespeech recognition accuracyVSAvoidmodel update mechanism complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

A feedback mechanism is implemented where recognition results are collected and used to continuously update the speech recognition model. This feedback loop enables the system to learn from actual usage patterns and improve accuracy over time without requiring complex manual retraining procedures.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The speech recognition model performs self-updates by automatically incorporating feedback data from its own operation. The system collects feedback during normal use and autonomously retrains or adjusts its parameters, reducing the need for external intervention and simplifying the overall system architecture despite the continuous learning capability.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS12236941B2Method and apparatus with a personalized speech recognition model

Publication Date: 2025.02.25 SAMSUNG ELECTRONICS CO LTD

AI summary

A method and apparatus for personalizing a speech recognition model is disclosed. The apparatus may obtain feedback data that is a result of recognizing a first speech input of a user using a trained speech recognition model, determine whether to update the speech recognition model based on the obtained feedback data, and selectively update, dependent on the determining, the speech recognition model based on the feedback data.