Audio Bandwidth Extension Using Spectral Envelope and Flatness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In UWB audio encoding, the absence of effective methods for efficiently encoding and decoding high-frequency signals at very low bitrates leads to inaccuracies and noise in the reconstructed high-frequency part, affecting the subjective sense of hearing.
Innovation Solution
An audio processing method that filters an audio signal into low-frequency and high-frequency components, performs spectral envelope and flatness extraction, and quantizes this information to create a bandwidth-extended bitstream, which is then decoded to reconstruct the high-frequency part accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a higher bitrate is allocated to the low-frequency part of the signal, then the encoding efficiency of the low-frequency part is improved, but the bitrate available for high-frequency encoding is reduced
Solution Approach 1:
The audio signal is segmented into low-frequency and high-frequency parts, with different encoding strategies applied to each. The low-frequency part receives higher bitrate for accurate encoding, while the high-frequency part uses bandwidth extension techniques to reconstruct from spectral envelope and flatness information, resolving the bitrate allocation contradiction.
Solution Approach 2:
The patent changes the encoding parameters for different frequency bands. Instead of using uniform bitrate allocation, the system uses spectral envelope extraction and spectral flatness extraction specifically for high-frequency bands, allowing efficient encoding with lower bitrate by leveraging the mathematical properties of spectral characteristics.
2Quantity of substance
If the high-frequency part is encoded at very low bitrate, then the bitrate consumption is reduced, but the accuracy and quality of the reconstructed high-frequency signal deteriorates
Solution Approach 1:
The patent introduces spectral envelope information and spectral flatness information as intermediary parameters to bridge the low-frequency encoded signal and the high-frequency reconstruction. These intermediary parameters capture the essential characteristics of the high-frequency signal, enabling accurate reconstruction from minimal bitrate data through mathematical modeling rather than direct high-bitrate encoding.
Solution Approach 2:
The system creates a simplified representation (copy) of the high-frequency signal's spectral characteristics through envelope and flatness extraction. This simplified spectral model is then used to reconstruct the high-frequency part, achieving accurate reconstruction without requiring the full original high-frequency data at high bitrate.
3Loss of information
If spectral envelope information is extracted from both low-frequency and high-frequency spectra, then the completeness of spectral information is improved, but the processing complexity increases
Solution Approach 1:
The patent applies different processing qualities to different frequency bands. Spectral envelope extraction is applied to both low-frequency and high-frequency spectra, but spectral flatness extraction is applied only to the high-frequency spectrum where it is most critical for maintaining perceptual quality. This localized application of processing techniques maintains information completeness while managing computational complexity.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
An audio processing method and apparatus, and an electronic device, a computer-readable storage medium and a computer program product. The method comprises: performing filtering processing on an audio signal, so as to obtain a low-frequency signal and a high-frequency signal (101); coding the low-frequency signal to obtain a code stream of the low-frequency signal (102); performing frequency-domain transformation processing on the low-frequency signal, so as to obtain a low-frequency spectrum, and performing frequency-domain transformation processing on the high-frequency signal, so as to obtain a high-frequency spectrum (103); performing spectral-envelope extraction processing on the low-frequency spectrum and the high-frequency spectrum, so as to obtain spectral-envelope information, and performing spectral-flatness extraction processing on the high-frequency spectrum, so as to obtain spectral-flatness information (104); and performing quantization coding processing on the spectral-flatness information and the spectral-envelope information, so as to obtain a frequency-band expanded code stream, and forming a coded code stream by combining the frequency-band expanded code stream and the code stream of the low-frequency signal (105)..