LPC Filter Quantization Using Multi-Reference Super-Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech and audio coding techniques face challenges in efficiently quantizing Linear Prediction Coefficients (LPC) filters, particularly in reducing the bit rate while maintaining subjective quality, especially in applications like multimedia and wireless communication, where variable bit rate quantization and super-frame structures are not fully utilized.
Innovation Solution
The method involves differential quantization of LPC filters relative to multiple references, including past or future quantized filters, and using a selector to choose the reference that provides the lowest distortion at a given bit rate or the lowest bit rate for a given target distortion level, along with switched absolute or differential quantization to adapt to changing signal conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If fixed number of bits per frame is used for LPC filter quantization in CELP frames, then device complexity is reduced, but bit rate efficiency deteriorates
Solution Approach 1:
The patent applies dynamics by transitioning from fixed-bit quantization to variable-bit quantization. The quantization device dynamically adjusts the number of bits allocated to LPC filter quantization based on the selected reference filter (past or future) and signal conditions. This allows the system to optimize bit rate efficiency by using more bits when necessary (e.g., when future reference is selected) and fewer bits when sufficient (e.g., when past reference suffices), thereby resolving the contradiction between device complexity and bit rate efficiency.
Solution Approach 2:
The patent changes the parameter of bit allocation from fixed to variable. By introducing a mechanism that selects between past and future reference filters and adjusts bit allocation accordingly, the system modifies the quantization parameter dynamically. This parameter change enables better bit rate efficiency without significantly increasing device complexity, as the adjustment is based on predetermined reference selection rather than complex real-time analysis.
2Loss of information
If variable bit rate quantization is implemented, then bit rate efficiency is improved, but device complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-selecting reference filters (past or future) based on predetermined criteria before actual quantization occurs. The system prepares reference filter selections in advance, which simplifies the variable bit rate quantization process during execution. This preliminary preparation reduces the computational burden during real-time operation, thereby improving bit rate efficiency without proportionally increasing device complexity.
3Loss of information
If super-frame structure is utilized for LPC filter quantization, then bit rate is reduced, but adaptability to changing signal conditions deteriorates
Solution Approach 1:
The patent applies dynamics by enabling the quantization device to adapt to changing signal conditions within the super-frame structure. By selecting between past and future reference filters dynamically based on signal characteristics, the system maintains adaptability while operating within the super-frame framework. This allows the system to reduce bit rate through super-frame utilization while simultaneously adapting to varying signal conditions through reference filter selection.
Solution Approach 2:
The patent changes parameters by adjusting reference filter selection and bit allocation based on signal conditions within the super-frame structure. This parameter adjustment enables the system to maintain adaptability to changing signals while benefiting from the bit rate reduction achieved through super-frame organization. The system modifies quantization parameters dynamically to balance bit rate efficiency and signal adaptability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A device and a method for quantizing, in a super-frame including a sequence of frames, LPC filters calculated during the frames of the sequence. The LPC filter quantizing device and method comprises: an absolute quantizer for first quantizing one of the LPC filters using absolute quantization; and at least one quantizer of the other LPC filters using a quantization mode selected from the group consisting of absolute quantization and differential quantization relative to at least one previously quantized filter amongst the LPC filters. For inverse quantizing, at least the first quantized LPC filter is received and an inverse quantizer inverse quantizes the first quantized LPC filter using absolute inverse quantization. If any quantized LPC filter other than the first quantized LPC filter is received, an inverse quantizer inverse quantizes this quantized LPC filter using one of absolute inverse quantization and differential inverse quantization relative to at least one previously received quantized LPC filter.