Audio Noise Floor Estimation via Cost Function Minimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise reduction methods fail to effectively estimate and remove background noise in user-generated audio content, especially when a quiet fragment is not available, and often discard narrow-band tonal components or make assumptions about the signal type, leading to inefficiencies and inaccuracies.
Innovation Solution
A method that divides the audio signal into buffers, determines time-frequency samples, and uses a cost function combining median and energy variation to estimate the noise floor, allowing for robust noise reduction across all types of audio signals without discarding narrow-band components or making signal-type assumptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing noise reduction methods are used, then noise can be reduced in some cases, but they fail when no quiet fragment is available and discard narrow-band tonal components
Solution Approach 1:
The patent changes the estimation parameters from requiring quiet fragments to using cost function minimization based on median and energy variation. This allows reliable noise floor estimation without discarding tonal components, as the cost function approach identifies noise floors through statistical properties rather than requiring silent periods.
Solution Approach 2:
Instead of finding quiet fragments to estimate noise (traditional approach), the patent inverts the approach by finding the noise floor through cost function minimization that works even when signal is present. The method seeks the minimum cost function value rather than relying on absence of signal.
2Measurement precision
If methods based on fitting smooth curves through minima are used, then noise floor can be estimated, but narrow-band tonal components are discarded
Solution Approach 1:
The patent replaces smooth curve fitting with a cost function based on median and energy variation metrics. This parameter change allows precise noise floor measurement while preserving tonal components, as the cost function operates on statistical properties of the signal rather than requiring smooth interpolation that eliminates peaks.
3Productivity
If percentile-based noise selection is used, then noise floor can be determined, but the method is not robust to fade in and fade out of signal
Solution Approach 1:
The patent uses a cost function that incorporates feedback from both median and energy variation measures. This feedback mechanism allows the system to adapt to signal variations like fade in and fade out, maintaining reliable noise floor determination by continuously evaluating the cost function across different time points and selecting the minimum.
4Ease of manufacture
If methods relying on signal assumptions are used, then noise reduction can be applied, but they do not generalize to all audio signal types
Solution Approach 1:
The patent creates a universal noise floor estimation method based on cost function minimization that works for all audio signal types. The approach uses general statistical properties (median and energy variation) rather than signal-specific assumptions, making it adaptable to speech, music, and other audio content while maintaining implementation simplicity.
Data Source
AI summary
Embodiments are disclosed for noise floor estimation and noise reduction, In an embodiment, a method comprises: obtaining an audio signal; dividing the audio signal into a plurality of buffers; determining time-frequency samples for each buffer of the audio signal; for each buffer and for each frequency, determining a median (or mean) and a measure of an amount of variation of energy based on the samples in the buffer and samples in neighboring buffers that together span a specified time range of the audio signal; combining the median (or mean) and the measure of the amount of variation of energy into a cost function; for each frequency: determining a signal energy of a particular buffer of the audio signal that corresponds to a minimum value of the cost function; selecting the signal energy as the estimated noise floor of the audio signal; and reducing, using the estimated noise floor, noise in the audio signal.