Audio Decoder Cross-Over Filtering for Artifact-Free Bandwidth Extension
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audio codecs face challenges in maintaining high compression efficiency and perceptual quality at low bitrates, particularly in bandwidth extension techniques that fail to accurately reconstruct high-frequency content without losing timbre or introducing artifacts like dissonance and warbling.
Innovation Solution
The proposed solution involves a decoder-side signal analysis and adaptive frequency regeneration using a cross-over filter and tile generator to adjust frequency borders and eliminate tonal portions, ensuring accurate spectral reconstruction and minimizing artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If bandwidth extension techniques are used to maintain high compression efficiency at low bitrates, then compression efficiency is improved, but perceptual quality deteriorates due to artifacts like dissonance and warbling
Solution Approach 1:
The patent applies different processing strategies to different spectral regions. The cross-over filter selectively processes frequencies around the transition point between low-frequency and high-frequency bands, applying spectral shaping only where needed to prevent artifacts while preserving natural sound in other regions. This localized approach maintains compression efficiency without degrading overall perceptual quality.
Solution Approach 2:
The patent dynamically adjusts filter parameters including cross-over frequency, filter order, and spectral shaping coefficients based on the input signal characteristics. By adapting these parameters to match the specific audio content, the system maintains high compression ratios while preventing artifacts such as dissonance and warbling that would otherwise degrade perceptual quality.
2Productivity
If spectral patching is used to reconstruct high-frequency content, then bandwidth extension is achieved, but timbre accuracy deteriorates due to loss of original spectral characteristics
Solution Approach 1:
The patent introduces a cross-over filter as an intermediary between the low-frequency input signal and the reconstructed high-frequency output. This filter acts as a bridge that spectrally shapes the transition region, preserving the timbral characteristics of the original signal while enabling bandwidth extension. The intermediary filter prevents direct spectral copying that would otherwise cause timbre loss.
Solution Approach 2:
The patent employs dynamic spectral shaping where the cross-over filter parameters are continuously adjusted based on the instantaneous spectral content of the input signal. This dynamic adaptation allows the system to maintain accurate timbre representation across varying signal conditions while achieving effective bandwidth extension.
3Productivity
If filterbank representation is used for efficient patching, then processing efficiency is improved, but artifact generation increases due to spectral discontinuities at transition frequencies
Solution Approach 1:
The patent converts the potential harm of spectral discontinuities at filterbank transition frequencies into a benefit by using the cross-over filter to deliberately shape the spectral transition. The filter transforms what would be abrupt discontinuities into smooth, natural-sounding transitions, turning a source of artifacts into a mechanism for artifact prevention while maintaining processing efficiency.
Solution Approach 2:
The patent applies preliminary spectral shaping through the cross-over filter before the final spectral reconstruction process. By pre-processing the signal to smooth out potential discontinuities at transition frequencies, the system prevents artifact generation in subsequent processing stages while maintaining the efficiency benefits of filterbank representation.
Data Source
Figure 1A~1B
Figure 2A
Figure 2B
AI summary
Apparatus for decoding an encoded audio signal comprising an encoded core signal (1), comprising: a core decoder (1400) for decoding the encoded core signal (1401) to obtain a decoded core signal; a tile generator (1404) for generating one or more spectral tiles having frequencies not included in the decoded core signal using a spectral portion of the decoded core signal; and a cross-over filter (1406) for spectrally cross-over filtering the decoded core signal and a first frequency tile having frequencies extending from a gap filling frequency (309) to an upper border frequency or for spectrally cross-over filtering a first frequency tile and a second frequency tile.