Binaural Audio Interpolation via Spectral Extrapolation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for processing audio signals to simulate head rotation using headphones suffer from audible artifacts due to the separation of low and high frequencies during interpolation, which affects sound quality.

Innovation Solution

The system utilizes the known location of loudspeakers to enable continuous, full-bandwidth interpolation of audio signals, aligning signals based on time of arrival and accounting for room acoustics, without separating low and high frequencies, thereby eliminating aliasing artifacts and improving sound quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If low-pass filtering is applied to remove spectral content above 1.5 kHz before interpolation, then the number of microphones is reduced and interpolation is simplified, but high-frequency content is lost and audible artifacts are introduced

Engineering Contradiction:

Improvenumber of microphonesVSAvoidsound quality

Core Design Contradiction:

Device complexityVSManufacturing precision

Solution Approach 1:

The patent segments the frequency spectrum into low-frequency components (below 1.5 kHz) and high-frequency components (above 1.5 kHz). Low-pass filtering is applied to the low-frequency components to enable simplified interpolation with fewer microphones, while the high-frequency components are handled separately through spectral extrapolation to restore them after interpolation, thus resolving the contradiction between reducing microphone number and maintaining sound quality

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary low-pass filtering on the low-frequency components before interpolation to enable simplified processing with fewer microphones. The high-frequency components are then restored through spectral extrapolation after interpolation, allowing the system to achieve both reduced complexity and maintained quality

Inventive Principle:

Principle #10Preliminary action

2Device complexity

If approximate methods are used to restore high-frequency content after low-pass filtering, then the number of microphones is reduced, but audible artifacts are introduced in the sound reproduction

Engineering Contradiction:

Improvenumber of microphonesVSAvoidaudible artifacts

Core Design Contradiction:

Device complexityVSObject-generated harmful factors

Solution Approach 1:

The patent replaces approximate mechanical filtering methods with spectral extrapolation techniques. Instead of using simple low-pass filters followed by approximate high-frequency restoration, the system employs spectral extrapolation to accurately reconstruct high-frequency content from the low-frequency components, eliminating audible artifacts while maintaining reduced microphone count

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the parameter approach from using fixed low-pass filter cutoff frequencies to dynamic spectral extrapolation. By analyzing the spectral content and using extrapolation based on the microphone array geometry and signal characteristics, the system accurately restores high-frequency content without introducing artifacts, improving both quality and accuracy

Inventive Principle:

Principle #35Parameter changes

3Manufacturing precision

If continuous interpolation is performed without frequency separation, then sound quality is improved and artifacts are eliminated, but the number of microphones required increases significantly

Engineering Contradiction:

Improvesound qualityVSAvoidnumber of microphones

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the frequency spectrum to apply different processing strategies: low-pass filtering for low-frequency components enables simplified interpolation with fewer microphones, while spectral extrapolation handles high-frequency components. This segmentation allows continuous interpolation without requiring a large number of microphones for the entire bandwidth

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by using low-pass filtering on only the low-frequency components (below 1.5 kHz) rather than the entire spectrum. This partial processing enables reduced microphone count while maintaining quality through selective frequency handling and spectral extrapolation for high-frequency content

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS9237398B1Motion tracked binaural sound conversion of legacy recordings

Publication Date: 2016.01.12 GOOGLE LLC

AI summary

Systems and methods are disclosed for a sound reproduction apparatus configured for receiving signals representative of the output of a plurality of microphones positioned to sample a sound field at points representing possible locations of a listener's left and right ears when positioned in said sound field at the location of the microphones, receiving a location of at least one sound source relative to said plurality of microphones, receiving orientation data of the listener's head, and calculating a binaural output using the sound source location, microphone output signals and orientation data. The binaural output includes the full-bandwidth of the microphone output signals.