Noise Attenuation via Segmented Codebook Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise attenuation methods for audio signals, particularly speech, face challenges in reverberant and diffuse noise environments, requiring resource-intensive algorithms and large codebooks, leading to suboptimal performance and high computational demands.
Innovation Solution
A noise attenuation apparatus that segments signals into time segments, generates estimated signal candidates by combining codebook entries from desired and noise signal codebooks, and attenuates noise using a subset of codebook entries selected based on a sensor signal, reducing computational resources and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large noise codebook is used to cover variety of noise types, then noise attenuation performance is improved, but computational resource requirements increase significantly
Solution Approach 1:
The patent segments the large noise codebook into multiple sub-codebooks organized by noise type categories (e.g., stationary noise, non-stationary noise, reverberant noise). This segmentation allows the system to select and process only relevant sub-codebooks based on the current noise environment, reducing computational load while maintaining comprehensive noise coverage.
Solution Approach 2:
The patent performs preliminary classification of the input signal to identify the noise type before proceeding to the matching stage. By pre-categorizing the noise environment and selecting corresponding sub-codebooks in advance, the system avoids searching through the entire large codebook, thereby reducing computational resources while preserving noise attenuation effectiveness.
2Measurement precision
If a search is performed over all possible combinations of speech codebook entries and noise codebook entries, then estimation accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent divides the exhaustive search space into segmented search stages. First, speech codebook entries are selected based on speech activity detection. Then, noise codebook entries are selected from pre-identified relevant sub-codebooks. Finally, matching is performed only on the selected subsets rather than all possible combinations, maintaining estimation accuracy while reducing computational complexity.
Solution Approach 2:
The patent performs partial searching by selectively processing only the most relevant codebook subsets identified through preliminary classification and speech activity detection. Instead of exhaustively searching all possible combinations, the system focuses computational effort on the most promising candidates, achieving sufficient estimation accuracy with reduced complexity.
3Reliability
If beam-forming algorithms are used for spatial filtering, then noise attenuation is achieved, but resource demands and algorithm complexity increase
Solution Approach 1:
The patent extracts and processes different noise components (stationary noise, non-stationary noise, reverberant noise) separately through dedicated codebook matching processes rather than using a unified complex beam-forming algorithm. This extraction approach simplifies the overall algorithm while maintaining effective noise attenuation for each noise type.
Solution Approach 2:
The patent transforms the spatial filtering problem into a spectral parameter matching problem. Instead of performing complex spatial calculations required by beam-forming, the system matches spectral characteristics of the input signal against codebook entries representing different noise types, achieving noise attenuation through parameter-based comparison rather than spatial filtering.
4Reliability
If codebook based algorithms are used for non-stationary noise conditions, then performance is improved, but the large number of noise candidates increases risk of erroneous estimates
Solution Approach 1:
The patent segments the noise codebook into specialized sub-codebooks for different noise conditions including non-stationary noise, stationary noise, and reverberant noise. By organizing noise candidates into distinct categories, the system reduces the search space for each condition and minimizes the risk of selecting erroneous candidates from unrelated noise types.
Solution Approach 2:
The patent performs preliminary classification to identify the current noise condition type before selecting candidates from the appropriate sub-codebook. This preliminary identification step ensures that candidate selection is context-aware and reduces the probability of erroneous estimates by restricting the search to relevant noise types rather than all possible candidates.
Data Source
AI summary
A noise attenuation apparatus receives a first signal comprising a desired and a noise signal component. Two codebooks (109, 111) comprise respectively desired signal candidates and noise signal candidates representing possible desired and noise signal components respectively. A noise attenuator (105) generates estimated signal candidates by for each pair of desired and noise signal candidates generating an estimated signal candidate as a combination of the desired signal candidate and the noise signal candidate. A signal candidate is then determined from the estimated signal candidates and the first signal is noise compensated based on this signal candidate. A sensor signal representing a measurement of the desired source or the noise in the environment is used to reduce the number of candidates searched thereby substantially reducing complexity and computational resource usage. The noise attenuation may specifically be audio noise attenuation.