Adaptive Noise-Floor Control in High-Frequency Audio Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional audio signal encoding techniques, such as Spectral Band Replication (SBR), often result in a significant difference between original and restored high-frequency signals, leading to diminished sound quality due to excessive noise addition, especially in voiced sound sections.
Innovation Solution
A method and apparatus that calculate and update the noise-floor level of high-frequency signals based on the presence of voiced or unvoiced sounds in low-frequency signals, adjusting the noise level accordingly to improve the accuracy of high-frequency signal encoding and decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If noise is added to the high frequency signal to compensate for tonal differences, then the difference between original and restored high frequency signals is reduced, but excessive noise is added especially in voiced sound sections
Solution Approach 1:
The patent applies different noise-floor levels to different sections of the signal based on their characteristics. Specifically, it identifies voiced and unvoiced sections in the low frequency signal and applies corresponding noise-floor levels to the replicated high frequency sections. This local differentiation allows noise compensation in unvoiced sections while avoiding excessive noise in voiced sections, thereby resolving the contradiction between restoration accuracy and noise control.
Solution Approach 2:
The patent dynamically adjusts the noise-floor level parameter based on the signal characteristics. By calculating voicing levels from the low frequency signal and using these to modulate the noise-floor level applied to the high frequency signal, the system changes the noise parameter adaptively. This allows the noise level to be increased when needed (unvoiced sections) and decreased when harmful (voiced sections), resolving the contradiction between needing noise for restoration accuracy and avoiding excessive noise.
2Ease of manufacture
If a fixed noise-floor level is used for high frequency signal replication, then the encoding process is simple, but the tonal characteristics of original high frequency signals are not accurately preserved
Solution Approach 1:
The patent performs preliminary analysis of the low frequency signal to calculate voicing levels before replicating the high frequency signal. By pre-calculating the noise-floor level based on the low frequency signal characteristics, the system prepares the appropriate noise parameters in advance. This preliminary action allows the encoding process to remain relatively simple while ensuring that the correct noise level is applied to preserve tonal characteristics, thus resolving the contradiction between simplicity and accuracy.
Solution Approach 2:
The patent uses the low frequency signal as feedback to determine the appropriate noise-floor level for high frequency replication. By analyzing the voicing characteristics of the low frequency signal and using this information to control the noise level in the high frequency signal, the system creates a feedback loop that ensures tonal accuracy. This feedback mechanism allows the system to maintain relatively simple encoding while achieving accurate tonal characteristics through adaptive noise control.
Data Source
AI summary
A method and apparatus to encoding or decoding an audio signal is provided. In the method and apparatus, a noise-floor level to use in encoding or decoding a high frequency signal is updated according to the degree of a voiced or unvoiced sound included in the signal.