PVQ Shape Search With Adaptive Bit Length Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Structured Pyramid Vector Quantization (PVQ) for speech and audio coding faces challenges in achieving computationally efficient shape search while maintaining high Signal to Noise Ratio (SNR), especially in implementations with fixed precision Digital Signal Processors (DSPs), particularly for high-rate coding scenarios where the number of unit pulses and dimensions are high.
Innovation Solution
The method involves iteratively adding unit pulses in an inner dimension search loop, determining possible upshifts in correlation values based on maximum absolute values, and dynamically adjusting bit word lengths to optimize calculations, ensuring efficient PVQ shape search by employing longer bit word lengths when necessary to maintain precision and avoid overflow.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If unconstrained vector quantization is used to achieve optimal quantization performance, then Signal to Noise Ratio is improved, but complexity and memory capacity requirements increase significantly
Solution Approach 1:
The patent segments the quantization process into structured components (pyramid structure with hierarchical levels) rather than treating it as an unconstrained optimization problem. This segmentation allows the system to achieve good SNR performance while maintaining manageable complexity through organized search procedures at different hierarchical levels.
Solution Approach 2:
The patent implements dynamic search strategies that adapt to the specific characteristics of the input signal. The search complexity and precision are dynamically adjusted based on signal properties, allowing the system to maintain high SNR when needed while reducing complexity for less demanding cases.
2Measurement precision
If higher rate speech and audio coding is used to improve quality, then Signal to Noise Ratio is improved, but the number of unit pulses and dimensions increase leading to higher search complexity
Solution Approach 1:
The patent divides the high-dimensional search space into hierarchical pyramid levels, where coarser levels provide initial guidance and finer levels refine the solution. This segmentation reduces the effective search complexity even when the total number of dimensions and unit pulses is high, enabling high-rate coding with acceptable computational load.
Solution Approach 2:
The patent performs preliminary coarse quantization at higher hierarchical levels before conducting fine search at lower levels. This preliminary action provides a good initial estimate that guides the subsequent fine search, reducing the search space and computational complexity required to achieve high SNR in high-rate coding scenarios.
3Device complexity
If fixed precision DSP is used to reduce device complexity, then implementation cost is reduced, but precision is lost leading to lower Signal to Noise Ratio
Solution Approach 1:
The patent employs dynamic precision management where the computational precision is adaptively adjusted based on the current search state and signal characteristics. This allows fixed-precision DSPs to achieve high SNR by using higher precision only when and where needed, rather than maintaining high precision throughout the entire computation process.
Solution Approach 2:
The patent changes numerical parameters (precision requirements, scaling factors) dynamically during the quantization process based on the current state. This allows the system to maintain high SNR performance on fixed-precision DSPs by adjusting parameter precision to match the actual computational needs at each stage.
4Measurement precision
If longer bit word lengths are used to maintain precision in limited precision DSP environments, then Signal to Noise Ratio is improved, but computational complexity and memory requirements increase
Solution Approach 1:
The patent dynamically adjusts bit word lengths based on the current computational stage and signal characteristics. Longer bit word lengths are used only when and where precision is critical for maintaining SNR, while shorter lengths are used in stages where precision requirements are lower, thereby reducing overall computational complexity and memory requirements.
Solution Approach 2:
The patent applies different precision levels (bit word lengths) to different parts of the computation based on local requirements. Critical computations that directly impact SNR use longer bit word lengths, while less critical operations use shorter lengths, optimizing the trade-off between precision and complexity.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
An encoder and a method therein for Pyramid Vector Quantizer, PVQ, shape search, the PVQ taking a target vector x as input and deriving a vector y by iteratively adding unit pulses in an inner dimension search loop. The method comprises, before entering a next inner dimension search loop for unit pulse addition, determining, based on the maximum pulse amplitude, maxampy, of a current vector y, whether more than a current bit word length is needed to represent enloopy, in a lossless manner in the upcoming inner dimension loop. The variable enloopy is related to an accumulated energy of the vector y. The performing of this method enables the encoder to keep the complexity of the search at a reasonable level. For example, it enables the encoder to apply an increased precision loop only when it may be needed, by analyzing whether the "worst case scenario" in the upcoming inner loop would require an inner loop with a higher precision than the one currently used.