FIR Filter Interpolation Using Coherence for Accurate Sound Fields

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional audio processing systems fail to accurately generate sound fields for all locations due to insufficient spatial sampling of impulse responses, leading to errors in audio reproduction and degraded user experience, particularly in complex environments with highly-directive sources.

Innovation Solution

A computer-implemented method using sub-band impulse responses and non-linear interpolation techniques to estimate impulse responses at target locations, selecting pairs based on coherence values and applying appropriate interpolation methods to generate accurate sound fields with reduced memory usage and measurement time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If linear interpolation is used to estimate impulse responses for all locations, then the system complexity is reduced, but the accuracy of sound field generation deteriorates

Engineering Contradiction:
Improvesystem complexityVSAvoidaccuracy of sound field generation
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent changes the interpolation parameter selection strategy by computing coherence values between impulse responses at different locations and dynamically selecting pairs with coherence below a threshold for non-linear interpolation, while using linear interpolation for pairs above the threshold. This adaptive parameter selection resolves the contradiction by optimizing accuracy for complex sound fields while maintaining simplicity for easier cases.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system dynamically adjusts the interpolation method based on the coherence characteristics of the sound field at each location. By computing coherence values and adaptively selecting between linear and non-linear interpolation techniques, the system optimizes accuracy for each specific location rather than applying a fixed method throughout, resolving the accuracy-complexity tradeoff.

Inventive Principle:
Principle #15Dynamics

2Quantity of substance

If a sparse set of FIR filters is used for a limited number of locations, then the memory usage is reduced, but the accuracy of impulse response estimation for all locations deteriorates

Engineering Contradiction:
Improvememory usageVSAvoidaccuracy of impulse response estimation
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent changes how impulse responses are selected and combined by introducing coherence-based filtering. Instead of uniformly sampling locations, the system identifies pairs of impulse responses with low coherence (below threshold) and applies non-linear interpolation to capture complex sound field variations, thereby achieving higher accuracy with fewer sampled locations and reduced memory requirements.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If higher spatial sampling of impulse responses is performed, then the accuracy of impulse response estimation is improved, but the measurement time and data collection requirements increase

Engineering Contradiction:
Improveaccuracy of impulse response estimationVSAvoidmeasurement time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent changes the sampling strategy by using coherence values as a selection criterion. Instead of uniformly increasing spatial sampling density, the system identifies critical locations where coherence between existing impulse responses is low, indicating complex sound field variations that require measurement. This targeted approach achieves high accuracy while minimizing the number of measurement locations and associated time requirements.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4354904B1Interpolation of finite impulse response filters for generating sound fields
Publication Date: 2025.08.13 HARMAN INT IND INC
  • EP4354904B1 patent drawingFigure 1
  • EP4354904B1 patent drawingFigure 2
  • EP4354904B1 patent drawingFigure 3

AI summary

Various embodiments disclose a computer-implemented method comprising determining a target location in an environment, determining a set of sub-band impulse responses for a first frequency sub-band, each in the set of sub-band impulse responses being associated with a location proximate to the target location, selecting a first pair of sub-band impulse responses for the first frequency sub-band among pairs within the set of sub-band impulse responses, computing a first coherence value indicating a coherence level between sub-band impulse responses in the first pair, determining that the first coherence value is below a coherence threshold, in response, combining the sub-band impulse responses using a non-linear interpolation technique to generate an estimated impulse response for the first frequency sub-band for the target location, generating, based on the estimated impulse response, a filter for a speaker, filtering an audio signal, and causing the speaker to output the filtered audio signal.