Adaptive Voice Activity Detection Threshold Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional voice activity detection (VAD) algorithms are not adaptive to background noise variations, leading to inaccurate judgments and wastage of bandwidth.
Innovation Solution
A VAD device and method that includes a background analyzing unit to identify noise features, a VAD threshold adjusting unit to calculate a bias for the VAD threshold based on noise variation parameters, and a VAD judging unit to modify the threshold for accurate noise judgment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a fixed VAD threshold is used based on long-term background noise average, then the algorithm is simple to implement, but it cannot adapt to background noise variations leading to inaccurate judgment
Solution Approach 1:
The patent applies dynamics by making the VAD threshold adjustable and adaptive to changing background noise conditions. The threshold is no longer fixed but dynamically modified based on short-term noise analysis, allowing the system to adapt to noise variations while maintaining reasonable complexity through structured adjustment mechanisms.
Solution Approach 2:
The patent changes the threshold parameter based on background noise characteristics. By analyzing short-term noise energy and comparing it with long-term averages, the system modifies the VAD threshold parameter to match current noise conditions, thereby improving adaptability without requiring complete redesign of the algorithm.
2Measurement precision
If the VAD threshold is adjusted to be adaptive to background noise, then judgment accuracy improves, but the algorithm complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-calculating and storing long-term background noise averages before actual voice activity detection. This preparation allows the adaptive threshold mechanism to work more efficiently during real-time operation, as the baseline noise level is already established, reducing the computational burden during critical detection phases.
Solution Approach 2:
The patent implements feedback by continuously monitoring short-term noise energy and using this information to adjust the VAD threshold. The system feeds back the noise analysis results to modify the threshold, creating a closed-loop adaptive mechanism that improves judgment accuracy while managing complexity through iterative refinement.
3Loss of energy
If background noise is not accurately distinguished from voice signals, then bandwidth is wasted by encoding noise frames, but increasing VAD stringency may cause false negative detections
Solution Approach 1:
The patent changes the decision parameter (VAD threshold) based on background noise characteristics. By adjusting the threshold according to short-term noise analysis, the system can dynamically adapt its sensitivity, allowing it to be more lenient in high-noise conditions to avoid false negatives while being more stringent in low-noise conditions to save bandwidth.
Data Source
AI summary
A voice activity detection (VAD) device and method provide for a VAD threshold that is adaptive to background noise variation.