Howling Detector Unit for Conference Audio Feedback Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conferences often experience howling noise due to high gain audio feedback between microphones and speakers, which existing echo-cancelers and noise removal methods struggle to completely eliminate, especially in diverse and uncontrolled environments, leading to CPU-intensive processing and incomplete recovery of voice signals.
Innovation Solution
A method and system using a howling detector unit within a Multipoint Control Unit (MCU) that analyzes audio streams with skewness, flatness, crest, and rolloff analyses to detect howling noise, preventing its propagation to other participants and automatically muting the source, without relying on client-side implementation or CPU-intensive filtering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If echo-cancelers and noise removal methods are used to eliminate howling noise, then some howling noise can be reduced, but the processing becomes CPU-intensive and voice signal recovery is incomplete
Solution Approach 1:
The patent extracts and removes the howling component from the audio signal using spectral analysis. The howling detector identifies howling frequencies in the frequency spectrum and selectively removes only those specific frequency components while preserving the rest of the audio signal, including voice content. This targeted extraction approach avoids CPU-intensive processing of the entire signal while achieving effective howling noise reduction.
Solution Approach 2:
The patent changes the parameter of audio signal representation from time domain to frequency domain through FFT transformation. By analyzing the signal in the frequency spectrum, the system can efficiently identify and remove howling frequencies without intensive time-domain processing. This parameter transformation enables simpler, more CPU-efficient noise removal while maintaining voice signal integrity.
2Loss of time
If echo-cancelers are used to delay howling effect, then howling presence is delayed, but the system becomes unstable when ERL is less or equal than 0 dB and howling cannot be completely removed
Solution Approach 1:
The patent applies preliminary action by detecting and removing howling frequencies before they can cause system instability or propagate to other participants. The howling detector continuously monitors the frequency spectrum and preemptively eliminates howling components as soon as they are detected, preventing the feedback loop from becoming unstable. This proactive approach maintains system reliability regardless of ERL values.
Solution Approach 2:
The patent converts the harmful howling feedback into a detectable spectral pattern. By transforming the audio signal to the frequency domain, the system identifies howling as distinct frequency peaks in the spectrum. This transformation turns the harmful acoustic feedback into a detectable and removable spectral feature, allowing the system to eliminate howling effectively even in unstable acoustic environments where traditional echo-cancelers fail.
3Measurement precision
If double-talk detector halts AEC adaptation when near-end speech is detected, then echo cancellation works better, but howling detection is slowed down or halted
Solution Approach 1:
The patent segments the audio processing into two independent parts: echo cancellation during speech periods and howling detection in the frequency spectrum. The howling detector operates independently on the frequency-transformed signal rather than on the time-domain speech signal. This segmentation allows both functions to operate simultaneously without interference, maintaining both echo cancellation accuracy and howling detection speed.
Solution Approach 2:
The patent substitutes the mechanical speech-based detection mechanism with a spectral analysis mechanism. Instead of relying on time-domain speech detection that conflicts with double-talk scenarios, the system uses frequency-domain spectral analysis to detect howling. This substitution allows howling detection to proceed independently of speech activity, maintaining detection speed while preserving echo cancellation performance during double-talk situations.
Data Source
AI summary
Method and System for avoiding howling disturbance especially on conferences, wherein the method comprising the steps of using a howling detector unit implemented inside a multipoint control unit to receive an audio stream input from a client, analyzing the audio input with the howling detector in order to verify if howling noise is present, using at least two of a skewness analysis, a flatness analysis, a crest analysis, a rolloff analysis and preventing the audio stream input to be forwarded as an output to an audio mixer, if howling noise is present.