Voice Recognition Buffering Wake-Up Word Position
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition technologies fail to recognize spoken sentences when the wake-up word is positioned after, within, or before the intended voice command in a user's utterance, leading to incomplete or missed voice command recognition.
Innovation Solution
A speech recognition method that buffers the spoken utterance and activates the speech recognition function upon detecting a wake-up word, allowing for the isolation and processing of spoken sentences regardless of their position, including varying buffering parameters based on environmental conditions to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If the speech recognition function is initiated only after the wake-up word is recognized, then the computational load is reduced and the device can operate efficiently, but the spoken sentence that precedes the wake-up word cannot be recognized
Solution Approach 1:
The system performs preliminary buffering of the spoken utterance before wake-up word recognition. By storing the audio data in advance in a buffer, the system ensures that the spoken sentence is captured even before the wake-up word is detected, allowing complete recognition while maintaining efficient processing
2Speed
If the wake-up word recognition is always running in the background, then the device can respond quickly to wake-up commands, but the spoken sentence position flexibility is limited
Solution Approach 1:
The system segments the spoken utterance into multiple parts by identifying the wake-up word position within the buffered audio. This segmentation allows the system to handle various scenarios: spoken sentences before the wake-up word, after the wake-up word, or the wake-up word itself as the command, providing position flexibility while maintaining quick response
3Reliability
If the system buffers the entire spoken utterance to enable recognition regardless of wake-up word position, then recognition completeness is improved, but the memory usage and processing complexity increase
Solution Approach 1:
The system extracts only the relevant spoken sentence portions from the buffered utterance based on wake-up word position. By isolating and processing only the necessary segments (spoken sentences before, after, or the wake-up word itself), the system achieves complete recognition while minimizing memory usage and processing complexity
Data Source
AI summary
Disclosed are a speech recognition method and a speech recognition device, in which speech recognition is performed by executing an artificial intelligence (AI) algorithm and/or a machine learning algorithm provided therein. According to an embodiment of the present disclosure, the speech recognition method includes buffering an inputted spoken utterance, determining whether a preset wake-up word is present in the spoken utterance by comparing the buffered spoken utterance to the preset wake-up word, and in response to the preset wake-up word in the spoken utterance, activating a speech recognition function and isolating, from the spoken utterance, a spoken sentence as a voice command without the wake-up word, and processing the spoken sentence and outputting a processing result.