Frequency-Domain Noise Estimation Filter for Audio Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise reduction systems in audio applications, such as vehicle audio systems, face inefficiencies when minimizing noise as they often minimize error across all frequencies, including the phase of the target signal, leading to sub-optimal solutions, especially when noise estimation is performed in the time domain.
Innovation Solution
An audio system that includes a noise-estimation filter configured to receive a magnitude-squared frequency-domain noise-reference signal and generate a magnitude-squared frequency-domain noise-estimation signal, and a noise-reduction filter that suppresses noise components of a microphone signal based on this estimation, allowing for frequency-by-frequency noise reduction by operating in the frequency domain, thereby avoiding phase-related constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If noise reduction is performed by minimizing error across all frequencies in the time domain, then noise suppression is achieved, but the phase of the target signal is affected leading to sub-optimal solutions
Solution Approach 1:
The patent segments the noise reduction process into frequency-specific operations by transforming the time-domain signal into frequency-domain representation. This allows independent processing of magnitude and phase components, where noise estimation is performed on magnitude-squared values while phase information is preserved separately and applied back to the enhanced signal.
Solution Approach 2:
The patent transitions from time-domain processing to frequency-domain processing, adding a frequency dimension to the analysis. This dimensional change enables frequency-by-frequency noise estimation and suppression, allowing selective noise reduction in specific frequency bands while preserving phase relationships across all frequencies.
2Object-affected harmful factors
If frequency-by-frequency noise reduction is implemented, then noise suppression performance is improved, but system complexity increases due to frequency transformation and magnitude-squared operations
Solution Approach 1:
The patent introduces magnitude-squared frequency-domain values as an intermediary representation that captures noise power information without requiring full complex spectral analysis. This intermediary form simplifies the noise estimation process by working with real-valued magnitude-squared data rather than complex frequency components, reducing computational complexity while maintaining frequency-selective noise suppression capability.
3Measurement precision
If noise estimation is performed using magnitude-squared frequency-domain signals, then frequency-specific noise characterization is achieved, but computational requirements increase compared to time-domain estimation
Solution Approach 1:
The patent applies local quality by estimating noise characteristics independently for each frequency bin using magnitude-squared values. This localized approach allows precise noise modeling in each frequency band where it is actually needed, rather than applying a single global noise estimate, thereby improving accuracy while avoiding the computational overhead of full spectral analysis.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
An audio system includes a noise-estimation filter, configured to receive a magnitude-squared frequency-domain noise-reference signal and to generate a magnitude-squared frequency-domain noise-estimation signal; and a noise-reduction filter configured to receive a microphone signal from a microphone, the microphone signal including a noise component correlated to an acoustic noise signal, and to suppress the noise component of the microphone signal, based, at least in part, on the magnitude-squared frequency-domain noise-estimation signal, to generate a noise-suppressed signal.