Voice Frame Compensation for Wake-Up Keyword Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice wake-up systems in electronic devices face challenges in accurately recognizing wake-up keywords due to voice activity detection (VAD) processing, leading to incomplete or truncated voice frame segments, which results in unsuccessful wake-ups and reduced user experience.
Innovation Solution
A voice processing method that collects voice signals, processes them into initial frame segments, performs recognition, and if the result does not match the target content, applies frame compensation by extending the segment based on historical data or set thresholds to improve matching accuracy, thereby enhancing the recognition of wake-up keywords.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If voice activity detection (VAD) processing is applied to voice signals, then power consumption is reduced and processing efficiency is improved, but voice frame segments become incomplete or truncated, leading to reduced recognition accuracy
Solution Approach 1:
The system performs preliminary voice recognition on the initial voice frame segment obtained from VAD processing. Based on the recognition result and token analysis, it proactively determines whether frame compensation is needed before final recognition, preventing potential recognition failures due to truncated frames.
Solution Approach 2:
The system implements a feedback mechanism where the initial voice recognition result is analyzed to determine if frame compensation is required. If the recognition result indicates incomplete content (through token analysis matching preset conditions), the system triggers frame compensation to extend the voice frame segment, and performs recognition again to verify improvement in matching degree.
2Measurement precision
If frame compensation is performed to extend voice frame segments, then recognition accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The system performs preliminary voice recognition on the initial voice frame segment before final recognition. This preliminary step allows early identification of cases requiring frame compensation, avoiding unnecessary compensation processing for already complete frames and reducing overall processing time.
Solution Approach 2:
The system applies frame compensation selectively rather than to all voice segments. By analyzing tokens in the preliminary recognition result and comparing them against preset conditions, the system performs compensation only when the recognition degree indicates incomplete content, thus balancing accuracy improvement with processing efficiency.
3Reliability
If multiple recognition passes are performed with frame compensation, then wake-up rate is improved, but device complexity and processing load increase
Solution Approach 1:
The system performs preliminary voice recognition and token analysis before final wake-up determination. This preliminary pass identifies candidates for frame compensation through preset condition matching, simplifying the decision logic for when to perform additional recognition passes and reducing overall system complexity.
Solution Approach 2:
The system uses feedback from preliminary recognition results to control the frame compensation process. By analyzing whether tokens in the preliminary result meet preset conditions, the system intelligently triggers or skips frame compensation and subsequent recognition passes, optimizing the balance between wake-up rate and processing complexity.
Data Source
AI summary
A voice processing method includes: collecting a voice signal by a microphone of an electronic device, and signal-processing the collected voice signal to obtain a first voice frame segment; performing voice recognition on the first voice frame segment to obtain a first recognition result; in response to the first recognition result not matching a target content and a plurality of tokens in the first recognition result meeting a preset condition, performing frame compensation on the first voice frame segment to obtain a second voice frame segment; and performing voice recognition on the second voice frame segment to obtain a second recognition result. A matching degree between the second recognition result and the target content is greater than a matching degree between the first recognition result and the target content.