Transient Audio Encoding with Multi-Resolution Filter Banks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing audio encoding techniques fail to optimize performance, quality, and compression, particularly in transient portions of audio signals, as they do not effectively adapt resolution and bit allocation based on signal characteristics.

Innovation Solution

The proposed solution involves a variable-resolution audio encoding technique that identifies transients within audio frames, applies multi-resolution filter bank analysis, and generates variable-length code book segments based on quantization data magnitudes, using different window functions within frames containing transients to achieve more accurate encoding with fewer bits.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If uniform resolution filtering is applied to all portions of the frame, then device complexity is reduced and processing is simplified, but encoding accuracy deteriorates in transient portions

Engineering Contradiction:
Improveencoding accuracyVSAvoidfiltering complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The frame is divided into multiple portions with different filtering resolutions. Transient portions are identified and processed with higher resolution filtering, while non-transient portions use lower resolution filtering. This segmentation allows the system to apply appropriate processing complexity to different regions, improving overall encoding accuracy without uniformly increasing device complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different filtering resolutions are applied to different portions of the frame based on local signal characteristics. Transient portions receive higher resolution filtering to capture rapid changes, while steady-state portions use lower resolution filtering. This local adaptation of quality ensures optimal encoding accuracy where needed without wasting computational resources elsewhere.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If fixed bit allocation is used for all transform data samples, then device complexity is reduced and processing is simplified, but encoding quality deteriorates for transient signals

Engineering Contradiction:
Improveencoding qualityVSAvoidbit allocation complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The bit allocation scheme is made dynamic and adaptive rather than fixed. The system adjusts the number of bits allocated to each transform data sample based on local signal characteristics, particularly identifying transient portions and allocating more bits to them. This dynamic adaptation improves encoding quality for transient signals without requiring complex manual configuration.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The bit allocation parameter is changed from a fixed value to a variable that adapts to signal characteristics. By monitoring transient detection results and signal energy distribution, the system modifies bit allocation parameters in real-time to optimize encoding quality. This parameter adaptation allows higher quality encoding for transient portions while maintaining efficiency in steady-state portions.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If variable-resolution filtering is applied to transient portions, then encoding accuracy is improved, but the number of bits required for transmission increases

Engineering Contradiction:
Improveencoding accuracyVSAvoidnumber of bits
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

Variable-resolution filtering is applied selectively only to transient portions of the signal rather than uniformly across the entire frame. This localized application improves encoding accuracy where it is most needed (in transient portions) while avoiding unnecessary bit consumption in steady-state portions, thus optimizing the balance between accuracy and bit rate.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system applies higher resolution filtering partially, only to the extent necessary for transient portions, rather than applying it excessively to the entire signal. This partial action approach achieves the minimum necessary improvement in encoding accuracy for transient regions without incurring the full cost of applying variable-resolution filtering to all portions, thereby controlling bit consumption.

Inventive Principle:
Principle #16Partial or excessive action

4Measurement precision

If different window functions are applied within a frame, then encoding accuracy is improved, but device complexity increases

Engineering Contradiction:
Improveencoding accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The frame is segmented into multiple portions, each processed with an appropriate window function based on local signal characteristics. Transient portions use window functions optimized for capturing rapid changes, while steady-state portions use window functions optimized for frequency resolution. This segmentation approach improves encoding accuracy without requiring the system to manage multiple window functions simultaneously across the entire frame.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The window function parameter is changed from a single fixed function to a variable that can be selected based on signal characteristics. The system dynamically selects appropriate window functions for different portions of the frame, adapting to transient or steady-state conditions. This parameter adaptation improves encoding accuracy while using control mechanisms to manage processing complexity.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2054883B1Audio encoding system
Publication Date: 2010.10.27 DIGITAL RISE TECH CO LTD
  • EP2054883B1 patent drawingFigure 1
  • EP2054883B1 patent drawingFigure 2~4
  • EP2054883B1 patent drawingFigure 5~6

AI summary

Provided are, among other things, systems, methods and techniques for encoding an audio signal, in which is obtained a sampled audio signal which has been divided into frames. The location of a transient within one of the frames is identified, and transform data samples are generated by performing multi-resolution filter bank analysis on the frame data, including filtering at different resolutions for different portions of the frame that includes the transient. Quantization data are generated by quantizing the transform data samples using variable numbers of bits based on a psychoacoustical model, and the quantization data are grouped into variable-length segments based on magnitudes of the quantization data. A code book is assigned to each of the variable-length segments, and the quantization data in each of the variable-length segments are encoded using the code book assigned to such variable-length segment.