Two-Stage Audio AGC for Stable Signal Levels in Video Conferencing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video conferencing systems face challenges in maintaining a stable and consistent audio signal level, leading to unpredictable audio quality due to varying speech levels, which can result in unpleasant auditory experiences for recipients.
Innovation Solution
A two-stage automatic gain control (AGC) module is implemented, comprising a long-term level estimator and a short-term level estimator, which work together to gradually and instantaneously adjust the audio signal to a target level, ensuring consistent audio quality by applying appropriate digital gains based on estimated signal levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single-stage AGC is used to adjust audio signal level, then the device complexity is reduced, but the audio signal stability and consistency deteriorate
Solution Approach 1:
The AGC module is divided into two independent stages: a long-term level estimator that provides gradual gain adjustment for overall stability, and a short-term level estimator that provides instantaneous gain adjustment for rapid response. This segmentation allows each stage to specialize in different time scales, resolving the contradiction between simplicity and stability by distributing functions across multiple components rather than using a single complex stage.
2Stability of the object's composition
If long-term signal estimation is used for AGC, then the audio signal stability is improved, but the response speed to sudden speech level changes deteriorates
Solution Approach 1:
The system dynamically switches between long-term and short-term estimation based on the time scale of the input signal variations. For gradual speech level changes, the long-term estimator provides stable control. For sudden speech level changes, the short-term estimator takes over to provide rapid response. This dynamic adaptation resolves the contradiction by adjusting the estimation time constant according to the nature of the input signal variations.
Solution Approach 2:
The AGC module periodically updates gain values based on both long-term and short-term signal estimates, combining the benefits of slow gradual adjustment with fast instantaneous correction. The periodic application of combined gains ensures that stability and response speed requirements are both met through rhythmic adjustment cycles that incorporate both time-scale perspectives.
3Stability of the object's composition
If aggressive gain adjustment is applied to maintain target level, then the audio signal consistency is improved, but noise and interference amplification increases
Solution Approach 1:
The AGC module applies gain adjustment in a controlled, partial manner through two stages rather than using a single aggressive adjustment. The long-term stage applies gradual gain changes that avoid sudden amplification of noise, while the short-term stage provides only instantaneous corrections. This partial action approach maintains signal consistency without excessively amplifying noise and interference that would result from aggressive single-stage adjustment.
Data Source
AI summary
Disclosed are systems and methods for generating a long-term signal level estimate during automatic gain control. In an example method, a computing device receives an input signal comprising feed data, the input signal derived from an upstream input signal. The computing device selects a first subset of the feed data and divides the first subset of the feed data into a plurality of analysis intervals. The computing device calculates a statistical average of each analysis interval and generates a first Gaussian distribution based on the statistical average of each analysis interval of the plurality of analysis intervals. Based on one or more characteristics of the first Gaussian distribution, the computing device determines the long-term signal level estimate, generates a first stage gain based on the long-term signal level estimate, and applies the first stage gain to the input signal.