Multi-Channel Signal Quantization via Frequency Band Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity and inefficiency of quantizing channel level differences (CLDs) in multi-channel signal processing, particularly for stereo speech signals, hinder the effectiveness of current audio processing technologies.
Innovation Solution
A method and apparatus that categorize frequency bands and quantify only those with matching categories, reducing the data to be processed and transmitted, and utilizing differences between adjacent bands to minimize bit redundancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all CLDs of stereo speech signals are quantized using a scalar quantizer, then the quantization accuracy is improved, but the computational complexity increases and quantization efficiency decreases
Solution Approach 1:
The patent divides frequency bands into different categories (e.g., low, mid, high frequency bands) and applies different quantization strategies to each category. Instead of uniformly quantizing all 31 frequency bands, the system segments them into groups where only certain categories are quantized in each frame, reducing the total number of quantization operations while maintaining accuracy where needed.
Solution Approach 2:
The patent applies different quantization precision levels to different frequency band categories based on their importance. Critical frequency bands receive full quantization precision, while less critical bands use coarser quantization or are skipped entirely in certain frames, optimizing the balance between accuracy and complexity for each local region of the frequency spectrum.
2Measurement precision
If all CLDs of stereo speech signals are quantized, then the sound field representation accuracy is improved, but the processing efficiency decreases
Solution Approach 1:
The patent implements periodic quantization where different categories of frequency bands are quantized in alternating frames or according to a periodic pattern. For example, odd-frequency bands are quantized in even frames while even-frequency bands are quantized in odd frames, ensuring all bands are periodically updated without requiring full quantization in every frame, thus improving efficiency while maintaining representation accuracy.
Solution Approach 2:
The patent performs preliminary classification of frequency bands into categories before quantization, and uses temporal correlation to predict which bands need quantization based on previous frames. This preliminary action identifies redundant quantization operations that can be skipped, improving processing efficiency while maintaining sound field representation accuracy through intelligent selection of quantization targets.
3Measurement precision
If the scalar quantizer processes all frequency bands, then the quantization precision is maintained, but the data transmission volume increases
Solution Approach 1:
The patent extracts and transmits only the essential quantization data by selectively quantizing only certain frequency band categories in each frame rather than all bands. This extraction approach removes redundant data from the transmission stream while preserving the most important sound field information, reducing data volume while maintaining quantization precision for critical parameters.
Solution Approach 2:
The patent applies partial quantization by processing only a subset of frequency bands in each frame rather than all bands uniformly. This partial action approach transmits sufficient information for acceptable sound field representation without the excessive data volume that would result from complete quantization of all 31 frequency bands in every frame, optimizing the trade-off between precision and data efficiency.
Data Source
AI summary
A method, an apparatus, and a system for encoding and decoding multi-channel signals are disclosed. The method for encoding multi-channel signals includes: determining the category of an index corresponding to a channel level difference (CLD) which needs to be quantized in a current frame; quantizing the CLD of at least one frequency band whose index category is the same as the determined category of the index in the current frame, and obtaining quantized data.