Audio Signal Processing Howling Suppression Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio signal processing methods for suppressing howling in voice calls, such as those using frequency shifters or phase shifters, are inefficient for real-time processing and often damage the sound quality of effective audio signals, especially in short-range voice processes where loop gains result in howling.
Innovation Solution
An audio signal processing method utilizing a machine learning model to identify and suppress howling points by determining gain values for both effective and ineffective audio signals, allowing for rapid and thorough howling suppression without affecting the effective audio signal, using a system or program in a terminal device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If frequency shifter or phase shifter is used to suppress howling, then howling suppression is achieved, but audio signal processing efficiency is low and sound quality of effective audio signals is damaged
Solution Approach 1:
The patent transforms the audio signal from time domain to frequency domain through Fourier transform, enabling parameter-based processing in the frequency domain. By identifying howling points through spectral analysis and applying targeted gain adjustments to specific frequency components, the system achieves efficient howling suppression while preserving the quality of non-howling audio signals. This parameter change approach allows selective modification of only the problematic frequency components rather than processing the entire audio signal uniformly.
2Reliability
If frequency shifter or phase shifter is used to suppress howling, then howling suppression is achieved, but sound quality of effective audio signals is damaged
Solution Approach 1:
The patent applies local quality by treating different frequency components differently. Through spectral analysis, the system identifies specific frequency points where howling occurs and applies gain reduction only to those particular frequency components. The majority of the audio spectrum that does not contain howling remains unchanged, preserving the original sound quality of effective audio signals. This localized processing approach ensures that only the problematic howling frequencies are suppressed while the rest of the audio signal maintains its original characteristics.
3Device complexity
If traditional methods are used for howling suppression, then processing is simpler, but howling suppression is not thorough and loop gains continue to occur
Solution Approach 1:
The patent implements a feedback mechanism by continuously monitoring the audio signal in the frequency domain, detecting howling points through spectral analysis, applying gain suppression, and then transforming the processed signal back to time domain. This closed-loop approach ensures that howling is detected and suppressed in real-time, preventing loop gains from occurring. The system dynamically adjusts the gain of frequency components based on continuous feedback from the audio signal analysis, ensuring thorough howling suppression while maintaining system stability.
Data Source
AI summary
This application discloses an audio signal processing method, a model training method, and related apparatus. The audio signal processing method includes inputting a first audio input signal into a machine learning model to obtain a first howling point, and obtaining a first gain value according to the first howling point, the first howling point corresponding to an effective audio signal in the first audio input signal, and the first gain value indicating a suppression parameter of the first howling point; processing the first audio input signal according to the first gain value to obtain a second audio input signal; detecting a second howling point, and obtaining a second gain value according to the second howling point, the second howling point corresponding to a frequency band of an ineffective audio signal in the second audio input signal; and processing the second audio input signal to obtain an audio output signal.