Audio Noise Detection via Time-Frequency Magnitude Differences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting noise in audio signals are inaccurate in real-time and computationally inefficient, especially when noise levels change dramatically, and are influenced by signal-to-noise ratio and entropy statistics.
Innovation Solution
A method that converts audio signals into audio frames, calculates spectral magnitudes, and determines differences in the time-frequency domain to identify noise segments based on the maximum degree of difference, using simple computations and orthogonal directions to accurately detect noise even with rapid changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If moving average calculation is used to track signal strength, then noise estimation can be performed, but real-time noise detection accuracy deteriorates when noise varies dramatically
Solution Approach 1:
The audio signal is divided into multiple audio frames arranged in chronological order with a target frame as center. Each frame's spectral components are analyzed separately to calculate magnitudes, allowing localized noise detection that adapts to rapid changes while maintaining computational efficiency.
Solution Approach 2:
The patent transforms the noise detection problem from simple time-domain moving average to a time-frequency domain analysis by calculating spectral magnitudes and their differences in orthogonal directions. This dimensional transformation enables accurate tracking of dramatic noise variations through gradient calculation in multiple directions.
2Measurement precision
If entropy statistics are used for noise estimation, then comprehensive noise analysis is achieved, but computational complexity increases significantly
Solution Approach 1:
The patent extracts only the essential feature needed for noise detection - the magnitude differences of spectral components in orthogonal directions - rather than performing comprehensive entropy statistics. This extraction approach maintains detection accuracy while dramatically reducing computational complexity by focusing only on the most relevant signal characteristics.
Solution Approach 2:
Instead of performing complete entropy statistics calculation, the patent applies a partial action by calculating only the maximum degree of difference in orthogonal directions. This partial approach provides sufficient noise detection capability without the excessive computational burden of full entropy analysis.
3Measurement precision
If model comparison is used for noise estimation, then theoretical accuracy can be achieved, but reliability depends heavily on voice training material availability
Solution Approach 1:
The patent enables the system to detect noise autonomously by calculating spectral magnitude differences directly from the audio signal itself, without requiring external voice training materials or pre-trained models. This self-service approach improves reliability by making the noise estimation independent of external data sources while maintaining accuracy through mathematical analysis of the signal's inherent characteristics.
Data Source
AI summary
A method and an apparatus for detecting noise of audio signals are provided. The method includes steps of converting an audio signal into a plurality of audio frames, where the audio frames are arranged in chronological order while taking a target frame as a center, calculating a plurality of magnitudes respectively corresponding to a plurality of spectral components of each of the audio frames, calculating differences between the adjacent magnitudes in a time-frequency domain to obtain a plurality of difference values in at least two directions orthogonal to each other in the time-frequency domain, where the time-frequency domain is defined by the audio frames, determining a maximum degree of difference of the magnitudes in the time-frequency domain according to the difference values, and determining whether a part of the audio signal corresponding to the target frame is a noise according to the maximum degree of difference.