Voice Recognition Buffering for Short Pause Wake-Up Word Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition technologies face challenges in accurately recognizing wake-up words and processing spoken sentences when the pause between the wake-up word and the spoken sentence is shorter than a preset duration, leading to unclear boundaries and incomplete recognition of user intent.
Innovation Solution
A speech recognition method that buffers spoken utterances with pauses shorter than a preset duration, extracts standby wake-up words, analyzes their role using natural-language processing algorithms, and determines the speech intent to accurately execute voice commands, regardless of the wake-up word's position within the utterance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech recognition starts only after wake-up word recognition, then wake-up word recognition accuracy is improved, but spoken sentences preceding the wake-up word are not recognized
Solution Approach 1:
The system performs preliminary buffering of the entire spoken utterance before wake-up word recognition is completed. This allows the system to have access to all spoken content (including sentences before the wake-up word) while still maintaining accurate wake-up word detection through subsequent analysis of the buffered data.
2Productivity
If speech recognition initiates only after wake-up word recognition, then processing efficiency is improved, but spoken sentences containing the wake-up word in the middle or end are not fully recognized
Solution Approach 1:
The complete spoken utterance is buffered in advance before wake-up word recognition completes, ensuring all spoken content is captured. Then the system analyzes the buffered data to identify the wake-up word position and extracts the relevant spoken sentence for processing, maintaining both efficiency and completeness.
Solution Approach 2:
The system extracts the specific spoken sentence containing the wake-up word from the buffered utterance after identifying the wake-up word's position. This extraction allows the system to process only the relevant portion while having access to the complete original utterance for accurate identification.
3Device complexity
If pause duration threshold is used to separate wake-up word and spoken sentence, then boundary detection is simplified, but recognition accuracy deteriorates when pause is shorter than threshold
Solution Approach 1:
The system replaces the simple mechanical pause-duration-threshold method with an AI-based natural language processing approach. The AI model analyzes the buffered utterance to intelligently identify the wake-up word and its position, accurately determining boundaries regardless of pause duration, thus maintaining low complexity while achieving high precision.
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 one embodiment, the speech recognition method includes buffering a spoken utterance, extracting a standby wake-up word corresponding to a preset wake-up word from the spoken utterance by comparing the buffered spoken utterance to the preset wake-up word, analyzing the role of the standby wake-up word in the spoken utterance, determining the speech intent in uttering the standby wake-up word by using results of analyzing the role of the standby wake-up word, and determining whether to execute a spoken sentence as a voice command in the spoken utterance and processing the spoken sentence accordingly.