Bandwidth Extension High-Band Energy Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bandwidth extension techniques for audio signals often result in speech that sounds muffled and has reduced intelligibility due to inaccuracies in estimating the high-band spectral envelope, leading to artifacts and discontinuities in the wideband spectral envelope.
Innovation Solution
A method that estimates high-band energy using a subset of the narrow-band signal, specifically the transition-band, to generate a more accurate high-band spectral envelope, which is then used to synthesize a pseudo wide-band signal, thereby minimizing artifacts and improving audio quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If narrow-band speech techniques are used to limit the quantity of information, then the overhead resources are reduced and wireless communication limitations are accommodated, but the speech sounds muffled and intelligibility is reduced
Solution Approach 1:
The patent segments the frequency spectrum into narrow-band (300-3400 Hz) and high-band (3400-8000 Hz) portions. The narrow-band portion is transmitted directly while the high-band portion is synthesized using spectral envelope estimation, allowing selective transmission of critical information while generating missing components to maintain speech quality.
Solution Approach 2:
The patent uses spectral envelope estimation as an intermediary to bridge the gap between narrow-band and high-band frequencies. By estimating the high-band spectral envelope from the narrow-band spectral envelope through codebook mapping, the system can reconstruct high-frequency components without transmitting them directly, thus maintaining intelligibility while reducing information quantity.
2Quantity of substance
If bandwidth extension techniques are used to synthesize pseudo wide-band signal, then the high-band spectral envelope is generated from narrow-band information, but artifacts and discontinuities occur in the wideband spectral envelope
Solution Approach 1:
The patent changes the parameter used for energy estimation from the entire narrow-band signal to specifically the transition-band portion (2500-3400 Hz). This parameter change reduces over-estimation of high-band energy and minimizes artifacts in the synthesized spectral envelope, improving the precision of bandwidth extension.
Solution Approach 2:
The patent uses only the transition-band portion of the narrow-band signal to estimate high-band energy, rather than using the entire narrow-band signal. This partial action approach focuses the estimation on the frequency region most relevant to high-band characteristics, reducing excessive energy estimation and resulting artifacts.
3Measurement precision
If the high-band spectral envelope shape is estimated through codebook mapping, then the high-band spectral envelope can be extracted, but mistakes in estimating the shape affect the estimates of high-band energy
Solution Approach 1:
The patent separates the spectral envelope estimation into shape and energy components. The shape is estimated through codebook mapping while the energy is estimated separately from the transition-band signal. This segmentation allows independent optimization of both parameters, preventing errors in shape estimation from propagating to energy estimation.
Solution Approach 2:
The patent uses the transition-band signal as an intermediary to estimate high-band energy independently of the spectral envelope shape estimation. This intermediary approach decouples the energy estimation from the shape estimation process, allowing each to be optimized separately and reducing the impact of shape estimation errors on energy accuracy.
Data Source
Figure 1~3
Figure 4
Figure 5~6
AI summary
A method (100) includes receiving (101) an input digital audio signal comprising a narrow-band signal. The input digital audio signal is processed (102) to generate a processed digital audio signal. A high-band energy level corresponding to the input digital audio signal is estimated (103) based on an estimated enery of a transition- band of the processed digital audio signal within a predetermined upper frequency range of a narrow-band bandwidth. A high-band digital audio signal is generated (104) based on the high-band energy level and an estimated high-band spectrum corresponding to the high-band energy level.