Parallel Audio Analysis and Time-Domain Filtering for Low Latency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional audio signal processing systems face a tradeoff between latency, computational complexity, and signal distortion, particularly introducing long latency and harmonic distortions due to time-to-frequency transforming and critically sampled filterbanks, which negatively impact real-time processing and user experience.

Innovation Solution

The proposed solution involves performing audio signal analysis and processing in parallel using separate modules, filtering in the time domain without time-to-frequency transforming, which reduces latency and supports linear convolution, allowing for real-time processing and flexible filter design to avoid harmonic distortion.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If time-to-frequency transforming and filterbanks are used for audio signal analysis, then signal analysis capability is improved, but processing latency increases

Engineering Contradiction:
Improvesignal analysis capabilityVSAvoidprocessing latency
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system separates audio signal processing into two independent parallel paths: an analysis path that uses time-to-frequency transforming and filterbanks for comprehensive signal analysis, and a processing path that operates directly in the time domain for low-latency real-time processing. This segmentation allows each path to be optimized for its specific function without compromising the other.

Inventive Principle:
Principle #1Segmentation

2Device complexity

If critically sampled filterbanks are used for time-to-frequency transforming, then computational complexity is reduced, but harmonic distortions are introduced

Engineering Contradiction:
Improvecomputational complexityVSAvoidharmonic distortions
Core Design Contradiction:
Device complexityVSObject-generated harmful factors

Solution Approach 1:

The harmful harmonic distortion generation is extracted and isolated to the analysis path where it does not affect the output audio quality. The processing path uses simple time-domain filtering that does not introduce harmonic distortions, while the analysis path can use computationally efficient critically sampled filterbanks without compromising the final audio output.

Inventive Principle:
Principle #2Taking out (Extraction)

3Manufacturing precision

If filterbanks are used for perfect reconstruction of audio signal, then signal fidelity is improved, but band isolation issues are introduced

Engineering Contradiction:
Improvesignal fidelityVSAvoidband isolation
Core Design Contradiction:
Manufacturing precisionVSReliability

Solution Approach 1:

Different quality requirements are applied to different parts of the system: the analysis path uses filterbanks designed for good band isolation and frequency selectivity, while the processing path uses simple time-domain filters that maintain signal fidelity without requiring complex filterbank designs. Each path is optimized for its specific local requirements.

Inventive Principle:
Principle #3Local quality

4Measurement precision

If audio signal is divided into frames by windowing for frequency domain analysis, then spectral analysis accuracy is improved, but processing latency increases

Engineering Contradiction:
Improvespectral analysis accuracyVSAvoidprocessing latency
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system segments the processing functions by performing spectral analysis on framed and windowed audio signals in the analysis path, while the processing path operates on continuous time-domain signals without framing or windowing. This allows spectral accuracy to be achieved where needed without introducing latency to the real-time processing path.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3304547B1Separated audio analysis and processing
Publication Date: 2023.10.11 DOLBY LABORATORIES LICENSING CORP
  • EP3304547B1 patent drawingFigure 1~2
  • EP3304547B1 patent drawingFigure 3~4
  • EP3304547B1 patent drawingFigure 5

AI summary

Example embodiments disclosed herein relate to separated audio analysis and processing. A system for processing an audio signal is disclosed. The system includes an audio analysis module configured to analyze an input audio signal to determine a processing parameter for the input audio signal, the input audio signal being represented in time domain. The system also includes an audio processing module configured to process the input audio signal in parallel with the audio analysis module. The audio processing module includes a time domain filter configured to filter the input audio signal to obtain an output audio signal in the time domain, and a filter controller configured to control a filter coefficient of the time domain filter based on the processing parameter determined by the audio analysis module. Corresponding method and computer program product of processing an audio signal are also disclosed.