Environment-Aware Acoustic Noise Reduction for Non-Stationary Speech
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing acoustic noise reduction techniques, such as single channel acoustic noise reduction (SCANR), are inadequate for real-time environments with non-stationary noise that varies significantly over time, as they are primarily suited for stationary or slowly varying noise.
Innovation Solution
A user environment-aware acoustic noise reduction technique that transforms noisy signals, determines feature vectors, and compares them with classification data to identify the current user environment, estimating noise levels and speech presence probability, allowing for adaptive noise reduction based on pre-stored thresholds and parameters specific to each environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If spectral subtraction techniques are used for noise reduction, then noise can be reduced from audio signal, but the technique is only suitable for stationary or very slow varying noise and cannot handle non-stationary noise that varies significantly over time
Solution Approach 1:
The patent applies dynamics by making the noise reduction system adaptive to changing environmental conditions. It uses environment classification to identify current acoustic conditions (e.g., office, car, pub) and dynamically adjusts noise reduction parameters accordingly. The system continuously monitors audio signals and reclassifies environments in real-time, allowing the noise reduction algorithm to adapt its behavior based on the detected environment type, thus handling non-stationary noise effectively.
Solution Approach 2:
The patent implements parameter changes by storing multiple sets of noise reduction parameters corresponding to different user environments. When an environment is classified, the system retrieves and applies the appropriate parameter set (e.g., different spectral subtraction coefficients, noise profiles, or processing gains) optimized for that specific environment. This allows the system to switch between different noise reduction strategies depending on whether the environment is stationary or non-stationary, improving both reliability and adaptability.
2Adaptability or versatility
If environment classification is added to identify current user environment, then adaptive noise reduction can be achieved, but the device complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the noise reduction task into distinct modules: environment classification module, parameter selection module, and noise reduction module. Each module handles a specific function independently. The classification module identifies environment type, the parameter module selects appropriate settings, and the reduction module applies noise reduction. This modular segmentation reduces overall system complexity by making each component simpler and more specialized.
Solution Approach 2:
The patent implements preliminary action by pre-storing classification data and noise reduction parameters for multiple user environments before actual noise reduction is needed. The system pre-processes and stores environment-specific parameters (e.g., noise profiles, spectral characteristics) in lookup tables or databases. When noise reduction is required, the system simply retrieves pre-computed parameters based on environment classification, avoiding complex real-time calculations and reducing computational complexity during operation.
Data Source
AI summary
Examples of the disclosure describe user environment aware single channel acoustic noise reduction. A noisy signal received by a computing device is transformed and feature vectors of the received noisy signal are determined. The computing device accesses classification data corresponding to a plurality of user environments. The classification data for each user environment has associated therewith a noise model. A comparison is performed between the determined feature vectors and the accessed classification data to identify a current user environment. A noise level, a speech level, and a speech presence probability from the transformed noisy signal are estimated and the noise signal is reduced based on the estimates. The resulting signal is outputted as an enhanced signal with a reduced or eliminated noise signal.