Nonstationary Noise Estimation via Local Energy Minima Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current noise estimation methods in mobile voice communications are inadequate for non-stationary and speech-like noises, leading to inaccurate noise suppression and intelligibility issues due to their reliance on outdated algorithms and slow response times.
Innovation Solution
The Nonstationary Noise Estimation (NNSE) method employs a composite noise energy estimate calculated through local energy minima tracking, probabilistic distribution analysis, and adaptive weighting, allowing for real-time noise estimation even in the presence of speech.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional noise estimation methods are used, then the system is simple to implement, but the noise estimation accuracy deteriorates for non-stationary and speech-like noises
Solution Approach 1:
The patent segments the noisy signal into multiple candidate noise frames by detecting local energy minima, then processes each segment independently to estimate noise characteristics. This segmentation allows the system to handle non-stationary noise by focusing on localized regions where noise dominates, improving estimation accuracy without requiring overly complex global processing
Solution Approach 2:
The patent introduces an intermediary probabilistic model that bridges the observed signal energy and the underlying noise parameters. By using probability distribution functions to represent noise energy at different time instances, the system mediates between uncertain observations and reliable estimates, achieving accurate noise characterization even in complex non-stationary environments
2Speed
If conventional noise estimation methods are used, then the processing is computationally efficient, but the tracking speed of non-stationary noise deteriorates
Solution Approach 1:
The patent performs preliminary action by proactively identifying and flagging candidate noise frames through local energy minima detection before full noise estimation is performed. This preliminary filtering step prepares the data structure and identifies relevant segments in advance, enabling faster subsequent processing of non-stationary noise without requiring exhaustive analysis of the entire signal
Solution Approach 2:
The patent implements periodic updates of noise estimates by systematically examining successive signal frames and updating noise parameters whenever new local energy minima are detected. This periodic examination rhythm allows the system to track non-stationary noise at appropriate intervals, balancing tracking speed with computational energy consumption by avoiding continuous processing
3Reliability
If noise estimate is updated frequently, then the responsiveness to non-stationary noise improves, but the reliability of noise estimate deteriorates due to insufficient averaging
Solution Approach 1:
The patent implements dynamic noise estimation where the averaging window and update frequency adapt based on the detected noise characteristics. For stationary noise, longer averaging provides reliable estimates, while for non-stationary noise detected through local energy minima, the system dynamically shortens the effective averaging period. This dynamic adjustment resolves the contradiction by matching the estimation timescale to the actual noise variability
Solution Approach 2:
The patent uses feedback from the signal analysis process itself to control noise estimate updating. By monitoring local energy minima and signal frame characteristics, the system receives feedback about when noise estimates should be updated versus when to maintain current estimates. This feedback mechanism ensures updates occur reliably when noise changes are detected while avoiding unnecessary updates that would waste time and reduce reliability
4Measurement precision
If speech is present in the signal, then the useful information is preserved, but the noise estimation accuracy deteriorates due to contamination
Solution Approach 1:
The patent extracts pure noise information from the noisy signal by specifically targeting local energy minima regions where speech is absent or minimal. By taking out and isolating these noise-dominated segments, the system obtains accurate noise estimates without contamination from speech content, while the speech signal itself remains intact in the original waveform for subsequent processing
Solution Approach 2:
The patent applies local quality analysis by examining different portions of the signal with different expectations. In regions identified as local energy minima, the system applies noise-focused analysis with the assumption that noise dominates. In other regions, speech-aware processing is applied. This local differentiation allows accurate noise estimation in noise-dominated regions while preserving speech integrity in speech-dominated regions
Data Source
AI summary
A method for estimating acoustic noise in an environment where a mobile communication device is operating and where the acoustic noise includes nonstationary noise or speech-like noises, and wherein the environment also includes speech signals. The method includes searching for a local minimum energy over a plurality of frames using at least two reference signals including a first signal comprised of a time-sensitive current local minimum energy estimate, emin, and a second signal comprised of a time-weighted average of previous detected local energy minima, eminmean; and deciding whether the detected local energy minima of the second reference signal is a noise signal. Also, binning the detected input signal energy minima values within a plurality of histograms; and calculating a composite noise energy estimate comprised of a weighted sum of a maximum probability noise energy estimate and an expected value noise energy estimate. As such a nonstationary noise estimator is formed.