Acoustic Feedback Cancellation Using Quadratic Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for eliminating acoustic howling feedback in audio systems are inefficient, unreliable, and impractical, especially when the microphone and speaker positions are fixed or EQ filters are not available, as they fail to accurately estimate howling frequencies and often distort signals.
Innovation Solution
A method involving a conversion module to transform audio signals into the frequency domain, a detection module to estimate howling frequencies using a howling-to-noise ratio (HNR) and notch filters to remove the estimated frequencies, improving frequency estimation precision and implementing a notch filter in real-time to eliminate howling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional FFT methods are used for howling frequency estimation, then the frequency can be detected, but the estimation precision is insufficient (error exceeds +/- 1 Hz with 1024-point FFT)
Solution Approach 1:
The patent introduces an intermediary quadratic interpolation process between the discrete FFT bins and the actual howling frequency. By using the magnitudes of the peak bin and its adjacent bins as intermediaries, the method calculates a fractional bin offset that precisely locates the howling frequency, achieving high precision without increasing FFT size.
Solution Approach 2:
The patent changes the parameter representation from discrete bin indices to continuous frequency values by calculating the fractional offset delta. This parameter transformation allows the system to achieve fine frequency resolution (±1 Hz) while maintaining a fixed 1024-point FFT configuration, avoiding the complexity of larger FFT sizes.
2Object-generated harmful factors
If manual repositioning of microphone and speaker is used to eliminate howling, then the howling can be removed, but the method is not convenient and impractical when positions are fixed
Solution Approach 1:
The system performs self-service by automatically detecting and estimating the howling frequency through signal processing, then automatically generating and applying the appropriate notch filter. This eliminates the need for manual intervention such as repositioning microphones or speakers, making the solution practical for fixed-position installations.
Solution Approach 2:
The patent replaces the mechanical approach of physically repositioning microphones and speakers with an electronic signal processing approach. By substituting mechanical adjustment with digital frequency estimation and filtering, the system achieves howling suppression without requiring any physical movement or manual operation.
3Object-generated harmful factors
If EQ filter is used to adjust gain in different audio bands to eliminate howling, then the howling can be eliminated, but the method is impractical when EQ filter is not available
Solution Approach 1:
The patent extracts only the essential function of EQ filtering by implementing a dedicated notch filter that targets specifically the detected howling frequency. This extracted function can be applied independently without requiring a full EQ filter system, making the solution versatile and applicable in situations where only basic filtering capability is available.
Solution Approach 2:
The system dynamically adapts to different situations by automatically detecting the presence of howling, estimating its frequency, and activating the notch filter only when needed. This dynamic behavior allows the solution to be versatile across various scenarios - working when EQ filters are unavailable, and maintaining audio quality when they are available, without requiring manual configuration.
4Object-generated harmful factors
If prior art digital methods are used for howling elimination, then the howling can be removed, but the processed signals are distorted or too complicated to be implemented in embedded systems
Solution Approach 1:
The patent applies partial action by implementing only the specific notch filtering needed to remove the detected howling frequency, rather than applying complex signal processing to the entire audio spectrum. This selective approach maintains signal integrity for non-howling components while effectively suppressing howling, and reduces computational complexity for embedded system implementation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach provides an efficient, reliable, and robust method for quickly removing howling sounds with improved frequency estimation precision, achieving +/- 1 Hz error using a 1024-point FFT, comparable to an 8192-point FFT in conventional methods, while maintaining intact speech and music signals.
Implementation Method 1
The incoming time domain audio signals from the microphone are transformed into frequency domain by a Fast Fourier Transform (FFT).
Implementation Method 2
a filter design module configured to provide filter coefficients for the adaptive design for removal of the estimated howling frequency; and a filter placed in line with the audio input signal to remove the estimated howling frequency
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
In embodiments of this invention howling sounds are removed without disturbing speech and music. Microphone signals are converted to the frequency domain where howling frequencies are detected and estimated using howling to noise ratios. A notch filter adaptively designed for the howling frequency is applied to remove the howling.