Two-Stage Audio Noise Suppression for Low-CPU Communication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication platforms struggle to efficiently suppress non-stationary noises in real-time using AI-based noise suppression due to high CPU resource consumption, particularly on client devices.
Innovation Solution
A two-stage noise suppression system that combines digital signal processing (DSP) and AI techniques, where DSP is used initially to filter stationary noise, followed by a machine learning algorithm to classify noise levels, and AI-based noise suppression is applied only when necessary, minimizing CPU usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI-based noise suppression is applied to suppress non-stationary noises, then noise suppression effectiveness is improved, but CPU resource consumption increases
Solution Approach 1:
The noise suppression process is divided into two distinct stages: first, DSP-based suppression handles stationary noises; second, AI-based suppression handles non-stationary noises. This segmentation allows each technique to be applied only where it is most effective, avoiding the unnecessary computational overhead of AI processing for all noise types.
Solution Approach 2:
AI-based noise suppression is applied partially rather than continuously. The system uses a classifier to detect when non-stationary noises are present, and only then activates the computationally intensive AI processing. This partial application significantly reduces CPU resource consumption while maintaining effective noise suppression when needed.
2Manufacturing precision
If AI-based noise suppression is applied in real-time, then noise reduction quality is improved, but processing complexity increases
Solution Approach 1:
The system performs preliminary classification of noise types using a lightweight machine learning classifier before applying the full AI-based noise suppression. This preliminary action identifies when non-stationary noises are present, allowing the complex AI processing to be triggered only when necessary, thereby reducing overall processing complexity while maintaining high noise reduction quality.
Solution Approach 2:
A noise classification module serves as an intermediary between the audio input and the AI-based noise suppression. This intermediary analyzes the audio signal to determine the presence of non-stationary noises and controls when the complex AI processing should be activated, simplifying the overall system architecture and reducing unnecessary computational complexity.
3Productivity
If DSP techniques are used to filter stationary noises, then CPU efficiency is improved, but non-stationary noises are not effectively filtered
Solution Approach 1:
The system creates a multi-functional noise suppression architecture where DSP techniques handle stationary noises and AI-based techniques handle non-stationary noises. Each component is optimized for its specific function, allowing the system to maintain high CPU efficiency for common stationary noises while also effectively filtering non-stationary noises when they occur.
Solution Approach 2:
The system dynamically changes processing parameters based on the type of noise detected. When stationary noises are present, efficient DSP parameters are used; when non-stationary noises are detected by the classifier, the system switches to AI-based processing parameters. This dynamic parameter adjustment maintains CPU efficiency while ensuring effective filtering across different noise types.
Data Source
AI summary
Methods and systems provide users of a communication platform with intelligent, real-time noise suppression for audio signals broadcasted in a communication session. The system receives an input audio signal from an audio capture device; processes the input audio signal to provide a second version of the audio signal with noise suppression based on DSP techniques; transmits the second version of the audio signal to a communication platform for real-time streaming; classifies, via a machine learning algorithm, whether the second version of the audio signal contains noise beyond a noise threshold; based on a classification that the second version of the audio signal contains noise beyond the noise threshold, processes the second version of the audio signal to provide a third version of the audio signal with noise suppression based on AI techniques; and transmits the third version of the audio signal to the communication platform.