Adaptive Audio Coding for Spectral Envelope and High-Frequency Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sound signal coding methods are inefficient depending on the characteristics of the input sound signal, such as the steepness of the spectral envelope and the degree of spectral concentration, leading to either inefficient coding or artificial-sounding decoded signals.
Innovation Solution
A coding method that selects between different coding processing types frame by frame based on the energy of high-frequency components and the sparsity of the spectral envelope, using either spectral envelope-based coding or differential variable-length coding, to adapt to the specific characteristics of the input sound signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If spectral envelope-based coding is used, then coding efficiency is improved for signals with steep spectral envelope ascents and descents, but code amount increases for signals with gentle spectral envelope changes
Solution Approach 1:
The patent dynamically switches between two different coding methods (spectral envelope-based coding and differential variable-length coding) based on the characteristics of the input sound signal. A selection unit determines which coding method to use by analyzing spectral features, allowing the system to adapt its coding approach to match the signal characteristics, thereby optimizing both coding efficiency and code amount.
Solution Approach 2:
The patent changes the coding parameter set based on signal characteristics. When the spectral envelope has steep ascents and descents, spectral envelope-based coding parameters are used; when the spectral envelope has gentle changes, differential variable-length coding parameters are used. This parameter switching resolves the contradiction by matching the coding parameters to the signal's spectral properties.
2Device complexity
If a single coding method is used for all signals, then device complexity is reduced, but coding efficiency decreases for certain signal types
Solution Approach 1:
The patent segments the coding process into two distinct coding paths: spectral envelope-based coding and differential variable-length coding. A selection unit divides the input signal processing based on spectral characteristics, routing signals to the appropriate coding path. This segmentation allows each coding method to be optimized for specific signal types without requiring a completely separate system for each.
Solution Approach 2:
The patent creates a universal coding system that can handle different types of sound signals (speech, music, etc.) with different spectral characteristics using a single integrated framework. The selection unit and two coding methods work together as a multi-functional system that adapts to various signal types, resolving the contradiction between simplicity and efficiency.
3Quantity of substance
If variable-length coding on average energy differential is used, then code amount is reduced for signals with gentle spectral envelope changes, but coding efficiency decreases for signals with high spectral concentration
Solution Approach 1:
The system dynamically selects between coding methods based on real-time analysis of spectral envelope characteristics. When gentle spectral envelope changes are detected, differential variable-length coding is activated to reduce code amount; when high spectral concentration is detected, spectral envelope-based coding is activated to maintain coding efficiency.
Data Source
AI summary
A coding technology that efficiently codes an input sound signal irrespective of the characteristics thereof and can obtain a decoded sound signal that sounds less artificial to a listener. A coding method codes an input sound signal frame by frame of a predetermined time segment by a selected coding processing from a plurality of types of coding processing in the frequency domain, the coding method makes it possible for a selection unit to select coding processing which is different from the coding processing of the preceding frame as coding processing of the present frame if at least one of the magnitude of the energy of high frequency components of the input sound signal of the preceding frame and the magnitude of the energy of high frequency components of the input sound signal of the present frame is smaller than or equal to a predetermined threshold value.