Signal Processing Device Noise Suppression Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional noise suppression methods, such as the LPSD method, struggle with accurate estimation of power spectrum densities in environments with both coherent and incoherent noise, leading to reduced noise suppressing performance.

Innovation Solution

A signal processing apparatus and method that includes a local PSD estimation unit, a target area/noise area PSD estimation unit, a first component extraction unit, a second component extraction unit, and a various noise responding gain calculation unit, which estimate and separate non-stationary and stationary noise components to calculate a post-filter that effectively suppresses noise.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional LPSD method is used to estimate power spectrum density, then the processing is simple, but the noise suppressing performance deteriorates in environments with both coherent and incoherent noise

Engineering Contradiction:
Improvenoise suppressing performanceVSAvoidestimation accuracy of power spectrum density
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent segments the noise estimation process into two distinct parts: stationary noise estimation and non-stationary noise estimation. By separating these estimation tasks and applying different methods to each type of noise, the system achieves more accurate overall noise suppression compared to conventional unified approaches.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the estimation parameters by introducing separate power spectrum density estimation parameters for stationary and non-stationary noise. This allows the system to adapt different estimation strategies to different noise characteristics, improving measurement precision in complex acoustic environments.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If separate estimation of stationary and non-stationary noise components is performed, then the noise suppressing performance improves, but the device complexity increases

Engineering Contradiction:
Improvenoise suppressing performanceVSAvoidcomplexity of signal processing apparatus
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The signal processing apparatus is segmented into specialized units: a stationary noise estimation unit, a non-stationary noise estimation unit, and a power spectrum density estimation unit. This modular segmentation allows each unit to handle specific tasks efficiently, improving noise suppression while keeping the overall system complexity manageable through functional decomposition.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent designs the signal processing apparatus with multi-functional units that can handle different types of noise estimation within a unified framework. The power spectrum density estimation unit, for example, serves both stationary and non-stationary noise estimation needs, reducing overall device complexity while maintaining improved noise suppression performance.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP3113508B1Signal-processing device, method, and program
Publication Date: 2020.11.11 NIPPON TELEGRAPH & TELEPHONE CORP
  • EP3113508B1 patent drawingFigure 1
  • EP3113508B1 patent drawingFigure 2
  • EP3113508B1 patent drawingFigure 3

AI summary

A signal processing technique the noise suppressing performance of which is more improved than conventional one is provided. A first component extraction unit 14 extracts a non-stationary component ^φS(A)(ω, τ) derived from a sound coming from a target area and a stationary component ^φS(B)(ω, τ) derived from an incoherent noise from a power spectrum density ^φS(ω, τ) of the target area through processing of time average. A second component extraction unit 15 extracts a non-stationary component ^φN(A)(ω, τ) derived from an interference noise and a stationary component ^φN(B)(ω, τ) derived from an incoherent noise from a power spectrum density ^φN(ω, τ) of a noise area.