Audio Encoding Sub-band Normalization for Music Signal Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing coding techniques for super-wide band signals, particularly CELP coding, face challenges in music signal encoding, leading to sound quality deterioration due to excessive peaking in the high band spectrum when normalizing the low band spectrum using incorrect energy estimates.
Innovation Solution
A coding apparatus that normalizes the low band spectrum by finding the largest amplitude value in each sub-band and using it to flatten the spectrum, preventing excessive peaking in the high band by copying a low band part with a sufficiently lowered peaking property to the high band, thereby generating a high-quality extension band spectrum.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If CELP coding is used in the core coding section for super-wide band signals, then speech signal encoding efficiency is improved, but music signal encoding quality deteriorates due to excessive peaking in the high band spectrum
Solution Approach 1:
The patent segments the encoding process into core coding section (low band) and extension band coding section (high band), allowing different coding strategies to be applied to different frequency ranges. This segmentation enables CELP coding to be used for speech signals in the low band while using transform coding for music signals in the extension band, thus resolving the contradiction between speech encoding efficiency and music encoding quality.
2Reliability
If the low band spectrum is normalized using spectrum power envelope, then energy bias is removed, but incorrect energy estimates cause excessive peaking in the high band spectrum
Solution Approach 1:
The patent introduces an intermediary process between low band decoding and high band encoding: the decoded low band signal is transformed into MDCT coefficients and used as an intermediary representation. This intermediary form allows for more accurate energy estimation and normalization, preventing the excessive peaking that occurs when directly normalizing the spectrum power envelope with incorrect energy estimates.
3Productivity
If a low band spectrum with high peaking property is copied to the high band part, then encoding efficiency is maintained, but sound quality deteriorates due to excessive peaking in the extension band
Solution Approach 1:
The patent applies preliminary normalization to the low band spectrum before copying it to the extension band. By pre-flattening the peaking properties through proper energy estimation and normalization using MDCT coefficients, the copied spectrum maintains encoding efficiency while avoiding the sound quality deterioration that would result from copying a high-peaking spectrum to the high band.
Data Source
AI summary
A coding apparatus includes a processor and a memory that stores instructions, which when executed causes the processor to perform operations, including encoding a first band of an input audio signal to be a first spectrum and dividing the first spectrum into a plurality of sub-bands. The operations also include searching a largest amplitude value of the divided first spectrum in each of the plurality of sub-bands, and normalizing the divided first spectrum in each of the plurality of sub-bands. The operations further include emphasizing a harmonic structure in the normalized first spectrum, and searching a best band that has a largest correlation value between each divided band of a second band spectrum and the emphasized first spectrum in which the harmonic structure is emphasized, and encoding the second band spectrum using lag information identifying the best band and transmitting the lag information to a decoder side.