Harmonic Bandwidth Extension Using Adaptive Non-Linear Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current harmonic bandwidth extension techniques for audio signals face challenges in generating high-quality synthesized high-band signals, particularly in distinguishing between speech and music, leading to weak correlation between the synthesized and actual high-band signals.
Innovation Solution
The method involves an encoder that separates an audio signal into low-band and high-band components, selects a non-linear processing function based on the characteristics of the low-band signal, and generates adjustment parameters to improve the correlation between the synthesized high-band signal and the actual high-band signal, using a combination of analysis filter banks, non-linear processing functions, and parameter estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single non-linear function is used to generate the synthesized high-band signal for all low-band signals, then the device complexity is reduced, but the quality of the synthesized high-band signal deteriorates for certain signal types (speech vs. music)
Solution Approach 1:
The system dynamically selects from multiple non-linear processing functions based on the characteristics of the low-band signal. The encoder determines whether the signal is speech or music and selects the appropriate function (e.g., cubic function for speech, quadratic function for music), allowing the processing to adapt to different signal types rather than using a fixed single function.
Solution Approach 2:
The system changes the parameters of the non-linear processing by selecting different functions from a set of available functions. Each function has different mathematical characteristics (cubic, quadratic, etc.) that are suited for different signal types. The selection process adjusts which function is applied based on signal analysis, optimizing the synthesis quality for each case.
2Productivity
If the high-band portion is not fully encoded and transmitted to improve coding efficiency, then the loss of information increases, but the productivity is improved
Solution Approach 1:
Instead of transmitting the full high-band signal, the system creates a synthesized copy of the high-band portion by applying non-linear processing functions to the low-band signal. This synthesized high-band signal is then combined with the transmitted high-band data to reconstruct the full bandwidth signal at the decoder, reducing transmission requirements while maintaining quality.
Solution Approach 2:
The encoder performs preliminary non-linear processing on the low-band signal to generate an extended low-band signal that contains information useful for reconstructing the high-band portion. This preliminary extension allows the decoder to synthesize the high-band signal more accurately with less transmitted data, improving coding efficiency.
3Reliability
If the bandwidth is extended from narrowband (3.4 kHz) to super wideband (16 kHz) to improve signal quality, then the use of energy increases, but the quality of signal reconstruction is improved
Solution Approach 1:
The encoder performs preliminary non-linear processing on the low-band signal to generate an extended low-band signal that contains information useful for reconstructing the high-band portion. This preliminary extension allows the decoder to synthesize the high-band signal more accurately with less transmitted data, improving coding efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method includes separating, at a device, an input audio signal into at least a low-band signal and a high-band signal. The low-band signal corresponds to a low-band frequency range and the high-band signal corresponds to a high-band frequency range. The method also includes selecting a non-linear processing function of a plurality of non-linear processing functions. The method further includes generating a first extended signal based on the low-band signal and the non-linear processing function. The method also includes generating at least one adjustment parameter based on the first extended signal, the high-band signal, or both.