Multi-Band Audio Noise Reduction for High-Frequency Speech Recovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio noise reduction methods suffer from inaccurate noise reduction processing results due to the use of a single sampling rate, leading to unnecessary computational load or loss of high-frequency information.
Innovation Solution
An audio noise reduction method that divides a noisy frequency domain representation into multiple frequency bands and processes each band using separate noise reduction branches with a consistent structure, followed by modulation and time domain transformation to obtain a noise-free speech signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single sampling rate is used for audio signal processing, then the processing system is simple, but the noise reduction accuracy deteriorates due to loss of high-frequency information or unnecessary computational load
Solution Approach 1:
The patent divides the audio signal processing into multiple parallel branches, each handling a specific frequency band (wide-band branch for 0-8kHz, full-band branch for 0-16kHz, narrow-band branch for 8-16kHz). This segmentation allows each branch to process its designated frequency range with appropriate sampling rates, improving noise reduction accuracy for different frequency components while managing system complexity through modular architecture.
Solution Approach 2:
The patent introduces a frequency band dimension by processing the audio signal through multiple parallel branches operating at different sampling rates simultaneously. Instead of using a single sampling rate, the system creates multiple processing paths (wide-band, full-band, narrow-band) that operate in parallel, then combines their outputs to achieve accurate noise reduction across the entire frequency spectrum.
2Adaptability or versatility
If up-sampling is performed to process wide-band signal with full-band method, then the processing is unified, but computational load increases unnecessarily
Solution Approach 1:
The patent applies different processing qualities to different frequency bands by creating specialized branches: the wide-band branch processes 0-8kHz at a lower sampling rate, the full-band branch processes 0-16kHz at a higher sampling rate, and the narrow-band branch handles specific frequency ranges. Each branch uses processing parameters optimized for its local frequency characteristics, avoiding unnecessary up-sampling and computational overhead.
3Use of energy by moving object
If down-sampling is performed to process full-band signal with wide-band method, then the computational load is reduced, but high-frequency information is lost
Solution Approach 1:
The patent segments the frequency spectrum into multiple bands handled by separate processing branches. The full-band branch processes the complete 0-16kHz range at appropriate sampling rates, while the narrow-band branch specifically handles the 8-16kHz high-frequency region. This segmentation ensures that high-frequency information is preserved through dedicated processing paths without requiring down-sampling that would cause information loss.
Data Source
AI summary
An audio noise reduction processing method and apparatus, a storage medium and an electronic device are disclosed. The method includes: obtaining an audio signal (S202); performing frequency domain transformation on the audio signal to obtain a noisy frequency domain representation (S204); and dividing the noisy frequency domain representation into N noisy frequency bands, and inputting the N noisy frequency bands respectively into N noise reduction branches in an audio processing network (S206); modulating the N branch mask estimation results by employing the noisy frequency domain representation, to obtain N speech frequency domain representations (S208); and performing time domain transformation on the N speech frequency domain representations, to obtain a speech signal free from noise signal interference in the audio signal (S210).