Adaptive Spectrum Parameter Decoding for Speech Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech decoding technologies fail to accurately determine spectrum parameters for error frames, leading to deteriorated speech quality when bad frames occur continuously, as they either ignore temporal variations or weaken the relevance between previous and current frame parameters.
Innovation Solution
A decoding method and device that adaptively shift the spectrum parameter of a good frame towards a constant mean value based on the number of continuous bad frames, using a spectrum parameter calculation unit to determine the spectrum parameter of the current bad frame, thereby maintaining relevance and improving speech quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the LSF of the previous frame is directly used as the LSF of the error frame, then the calculation is simple, but the temporal variation of spectrum parameter is ignored resulting in poor speech quality when bad frames occur continuously
Solution Approach 1:
The patent applies dynamics by making the spectrum parameter replacement adaptive rather than static. The replacement LSF is dynamically adjusted based on the number of continuous bad frames through a weighting factor that varies with time, allowing the system to adapt to continuous error conditions while maintaining reasonable computational complexity.
Solution Approach 2:
The patent changes the parameter of spectrum parameter replacement by introducing a time-varying weighting factor that depends on the number of continuous bad frames. This transforms the fixed replacement approach into a parameter-adjusted approach, improving speech quality under continuous error conditions without significantly increasing complexity.
2Reliability
If the LSF is shifted towards a constant mean value, then the relevance between previous and current frame is maintained, but when bad frames occur continuously the relevance is weakened resulting in suboptimal spectrum parameter
Solution Approach 1:
The patent makes the balancing between previous frame LSF and mean LSF dynamic by adjusting the weighting factor based on the number of continuous bad frames. When errors are continuous, the weight of previous frame LSF is reduced; when errors are isolated, the weight is increased, optimally preserving relevance while handling continuous errors.
Solution Approach 2:
The patent introduces feedback by using the detection result of continuous bad frames to adjust the replacement strategy. The system monitors the error condition and feeds this information back into the LSF replacement calculation, allowing adaptive adjustment of the weighting factor to maintain optimal speech quality.
3Measurement precision
If multiple historical frames are used for spectrum parameter replacement, then the accuracy may be improved, but the memory usage and calculation complexity increase
Solution Approach 1:
The patent extracts only the essential information needed for accurate replacement by focusing on the nearest good frame's LSF and the count of continuous bad frames. This selective extraction avoids the need to store and process multiple historical frames, reducing memory usage and computational complexity while maintaining accuracy.
Data Source
AI summary
A decoding method and device are provided. The spectrum parameter of a current bad data frame is determined. Specifically, a number of continuous bad frames that occur currently is determined. A spectrum parameter of a good data frame before the current bad data frame is determined. And a constant mean value of a spectrum parameter is determined. Then, the spectrum parameter of the good data frame is adaptively shifted towards the constant mean value of the spectrum parameter according to the number of the continuous bad data frames to calculate and obtain spectrum parameter information of the current bad frame. When the continuous bad data frames occur, the relevance between the spectrum parameter of the nearest good frame and the spectrum parameter of the current bad frame is gradually reduced, so that more accurate spectrum parameter of the current bad data frame can be obtained, thereby obtaining a better speech quality under a same code rate and a same frame error rate.