Voice Recognition Start Command Registration with Reference Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for registering a user's voluntary start command for speech recognition functions are inefficient in resource usage, slow in verification speed, and inaccurate due to misrecognition, affecting the practical use and convenience of the speech recognition function.
Innovation Solution
Optimizing the data used in the registration process by utilizing reference audio specific to the user's characteristics, such as utterance history and surroundings, to improve suitability verification, thereby reducing misrecognition and enhancing verification speed and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large amount of data is used in verifying the suitability of the start command, then the verification accuracy is improved, but the resource efficiency and verification speed deteriorate
Solution Approach 1:
The patent segments the verification process into two distinct stages: preliminary verification using a small amount of data to quickly filter out clearly unsuitable start commands, and detailed verification using a larger amount of data only for commands that pass the preliminary stage. This segmentation resolves the contradiction by applying different data quantities at different verification stages, thereby maintaining high accuracy while improving overall verification speed and resource efficiency.
2Measurement precision
If a large amount of data is used in verifying the suitability of the start command, then the verification accuracy is improved, but the resource efficiency deteriorates
Solution Approach 1:
The patent divides the verification process into preliminary verification (using minimal data) and detailed verification (using extensive data). This segmentation ensures that the large amount of data is processed only when necessary, significantly reducing overall resource consumption while maintaining verification accuracy. The preliminary stage acts as a filter to avoid unnecessary processing of clearly unsuitable commands.
3Productivity
If less-relevant data is used in verifying the suitability of the start command, then the resource efficiency is improved, but the verification accuracy deteriorates
Solution Approach 1:
The patent extracts and separates relevant data from less-relevant data in the verification process. The preliminary verification stage uses only essential, highly relevant data to quickly assess basic suitability, while the detailed verification stage incorporates additional relevant data for commands that pass the preliminary stage. This extraction approach ensures that resource efficiency is maintained while verification accuracy is preserved for commands that require thorough checking.
4Ease of operation
If the start command is voluntarily registered by the user, then the practical use and convenience are improved, but the misrecognition rate increases
Solution Approach 1:
The patent implements preliminary verification as a preliminary action before final registration of the start command. During this preliminary stage, the system quickly assesses whether the user's chosen start command is likely to cause misrecognition by comparing it against a small set of critical reference data. Commands that fail this preliminary check are flagged for further detailed verification or rejected, thereby preventing misrecognition while preserving user convenience for suitable commands.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed is an electronic apparatus which may receive a user voice of a user to be registered as a start command for activating a speech recognition function of the electronic apparatus; identify whether the user voice is misrecognized as a reference audio, wherein the reference audio includes a voice uttered by a user with high frequency from among a plurality of voices previously uttered by the user; and based on the user voice being not misrecognized as the reference audio, register the user voice as the start command.