Sound Source Separation Matrix Updates for Faster Convergence

Solution Overview

Problem

Conventional sound source separation systems face issues with convergence rate and precision due to a fixed step-size parameter, leading to improper update amounts that can result in insufficient or excessive adjustments, affecting the accuracy and efficiency of sound source separation.

Innovation Solution

A sound source separation system that includes a first processing element to recognize a cost function for evaluating separation degrees and a second processing element to iteratively update the separation matrix, adjusting the update amount based on the current cost function value and gradient, using a multi-dimensional Newton's method to optimize convergence rate and precision.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a fixed step-size parameter is used for updating the separation matrix, then the algorithm is simple to implement, but the convergence rate and precision are insufficient due to improper update amounts

Engineering Contradiction:

Improveimplementation simplicityVSAvoidconvergence precision

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent applies dynamics by transitioning from a fixed step-size parameter to a dynamic step-size parameter that automatically adjusts during the iterative update process. The step-size parameter is modified based on the current separation performance and gradient information, allowing the algorithm to take larger steps when far from convergence and smaller steps when approaching the optimal solution, thereby improving both convergence rate and precision without sacrificing implementation feasibility

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements parameter changes by modifying the step-size parameter according to specific criteria during iteration. The step-size is adjusted based on the cost function value and gradient magnitude, transforming a static parameter into a adaptive one that responds to the separation process state, thus resolving the contradiction between simplicity and precision

Inventive Principle:

Principle #35Parameter changes

2Ease of operation

If a fixed step-size parameter is used for updating the separation matrix, then the algorithm is simple to implement, but the convergence speed is slow due to insufficient update amounts

Engineering Contradiction:

Improveimplementation simplicityVSAvoidconvergence speed

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The patent applies dynamics by transitioning from a fixed step-size parameter to a dynamic step-size parameter that automatically adjusts during the iterative update process. The step-size parameter is modified based on the current separation performance and gradient information, allowing the algorithm to take larger steps when far from convergence and smaller steps when approaching the optimal solution, thereby improving both convergence rate and precision without sacrificing implementation feasibility

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements parameter changes by modifying the step-size parameter according to specific criteria during iteration. The step-size is adjusted based on the cost function value and gradient magnitude, transforming a static parameter into a adaptive one that responds to the separation process state, thus resolving the contradiction between simplicity and precision

Inventive Principle:

Principle #35Parameter changes

3Productivity

If the update amount is increased to improve convergence rate, then the convergence speed improves, but the convergence precision deteriorates due to excessive adjustments

Engineering Contradiction:

Improveconvergence speedVSAvoidconvergence precision

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies dynamics by transitioning from a fixed step-size parameter to a dynamic step-size parameter that automatically adjusts during the iterative update process. The step-size parameter is modified based on the current separation performance and gradient information, allowing the algorithm to take larger steps when far from convergence and smaller steps when approaching the optimal solution, thereby improving both convergence rate and precision without sacrificing implementation feasibility

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements parameter changes by modifying the step-size parameter according to specific criteria during iteration. The step-size is adjusted based on the cost function value and gradient magnitude, transforming a static parameter into a adaptive one that responds to the separation process state, thus resolving the contradiction between simplicity and precision

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS8131542B2Sound source separation system which converges a separation matrix using a dynamic update amount based on a cost function

Publication Date: 2012.03.06 HONDA MOTOR CO LTD

AI summary

A system capable of separating sound source signals with high precision while improving a convergence rate and convergence precision. A process of updating a current separation matrix Wk to a next separation matrix Wk+1 such that a next value J(Wk+1) of a cost function is closer to a minimum value J(W0) than a current value J(Wk) is iteratively performed. An update amount ΔWk of the separation matrix is increased as the current value J(Wk) of the cost function is increased and is decreased as a current gradient ∂J(Wk)/∂W of the cost function is rapid. On the basis of input signals x from a plurality of microphones Mi and an optimal separation matrix W0, it is possible to separate sound source signals y(=W0·x) with high precision while improving a convergence rate and convergence precision.