Noise Parameter Reconstruction for Speech Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech coding standards, such as DTX/CNG in AMR and ITU's silence compression scheme, face issues with discontinuous noise generation due to fixed SID frame intervals, leading to discontinuous noise spectra and poor listening quality when SID frames are sent at adaptive intervals or when there are significant changes in spectral parameters.
Innovation Solution
A method and apparatus for noise generation that allows SID frames to be sent at either fixed or adaptive intervals, using previously received noise parameters to determine a random value range for noise reconstruction, ensuring a smooth transition and natural noise generation by updating noise parameters incrementally based on differences and intervals between SID frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If SID frames are sent at fixed intervals using linear interpolation, then the decoding process is simple, but the noise spectrum becomes discontinuous and listening quality deteriorates when spectral parameters change significantly
Solution Approach 1:
The patent applies dynamics by making the noise generation process adaptive rather than static. The decoder dynamically adjusts noise parameters based on received SID frames, using exponential weighting factors that adapt to changes in spectral parameters. This allows the system to maintain smooth noise transitions while handling variable transmission intervals and significant spectral changes, resolving the contradiction between simple decoding and continuous noise spectrum.
Solution Approach 2:
The patent changes parameters by introducing exponential weighting factors and using non-linear interpolation methods instead of simple linear interpolation. The spectral parameters are transformed using exponential weighting, and the noise generation uses these transformed parameters to maintain continuity. This parameter transformation approach enables the system to handle significant spectral parameter changes while maintaining smooth noise transitions.
2Productivity
If SID frames are sent at adaptive intervals to improve communication efficiency, then system capacity increases, but noise generation becomes discontinuous and listening quality suffers
Solution Approach 1:
The patent makes the noise generation system dynamic by adapting to variable SID frame intervals. The exponential weighting factors are adjusted based on the timing and content of received SID frames, allowing the system to maintain continuous and natural-sounding noise generation even when SID frames arrive at irregular intervals. This dynamic adaptation resolves the contradiction between improved communication efficiency and maintained noise continuity.
Solution Approach 2:
The patent implements feedback mechanisms where the decoder continuously monitors received SID frames and adjusts noise generation parameters accordingly. The exponential weighting factors are updated based on feedback from the timing and spectral content of received frames, enabling the system to compensate for adaptive interval variations and maintain smooth noise transitions throughout the communication process.
3Ease of manufacture
If linear interpolation is used to estimate noise parameters between SID frames, then computation is simple, but the noise transition appears unnatural when spectral parameters change significantly
Solution Approach 1:
The patent transforms the interpolation approach by applying exponential weighting to spectral parameters and using non-linear transformation methods. Instead of simple linear interpolation, the system uses exponential weighting factors that emphasize recent SID frame data while still incorporating historical information. This parameter transformation maintains computational feasibility while producing much more natural-looking noise transitions that adapt to significant spectral changes.
Solution Approach 2:
The patent introduces dynamic weighting factors that adjust the interpolation behavior based on the magnitude of spectral parameter changes. When spectral parameters change significantly, the exponential weighting automatically adapts to provide smoother transitions. This dynamic approach maintains relative computational simplicity while dramatically improving the naturalness of noise transitions compared to fixed linear interpolation.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The disclosure provides a method for noise generation, including: determining an initial value of a reconstructed parameter; determining a random value range based on the initial value of the reconstructed parameter; taking a value in the random value range randomly as a reconstructed noise parameter; and generating noise by using the reconstructed noise parameter. The disclosure also provides an apparatus for noise generation.