Audio Encoder Switching via Baseband Signal Buffer Population
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech coding technologies face challenges in maintaining high voice quality at low bit rates, leading to perceptually significant distortion and artifacts when switching between different encoding technologies, particularly due to frame boundary issues and energy mismatches.
Innovation Solution
The implementation of a system that uses a modified discrete cosine transform (MDCT) encoder for frames with high-frequency components and an algebraic code-excited linear prediction (ACELP) encoder for frames without substantial high-frequency components, where the MDCT encoder generates a baseband signal used to populate the ACELP encoder's target signal buffer, reducing the need for buffer resets and filter state reinitialization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If speech coding switches between different encoding technologies (e.g., MDCT and ACELP) to adapt to varying signal characteristics, then coding quality and adaptability are improved, but frame boundary artifacts and energy mismatches occur due to buffer resets and filter state reinitialization
Solution Approach 1:
The patent applies preliminary action by pre-populating the target signal buffer with extrapolated values before the encoder switch occurs. This advance preparation ensures that when the switch happens, the buffer already contains appropriate data, preventing the need for reset operations that would cause frame boundary artifacts and energy mismatches.
Solution Approach 2:
The patent introduces an intermediary mechanism (the target signal buffer with extrapolated values) that mediates between different encoder technologies. This intermediary allows smooth transitions by providing continuous data during switch points, acting as a bridge that prevents discontinuities and artifacts during technology switching.
2Ease of operation
If buffer resets and filter state reinitialization are performed when switching encoders, then encoder switching is simplified, but signal continuity and perceived quality deteriorate due to frame boundary artifacts
Solution Approach 1:
By performing preliminary population of the target signal buffer with extrapolated values before the switch, the system maintains signal continuity without requiring complex reset operations. This preliminary action ensures that the transition is seamless, preserving both ease of operation and signal stability.
Solution Approach 2:
The patent applies beforehand cushioning by preparing the target signal buffer in advance with appropriate extrapolated values. This cushioning effect absorbs the potential discontinuities that would otherwise occur during encoder switching, protecting signal continuity while maintaining operational simplicity.
3Device complexity
If a single encoder is used for all frames, then device complexity is reduced, but coding quality deteriorates due to inability to handle different signal characteristics (speech vs. music)
Solution Approach 1:
The patent applies segmentation by dividing the encoding process into different paths based on signal characteristics. The system segments frames into speech-like (handled by ACELP) and music-like (handled by MDCT) categories, allowing each encoder to specialize in its optimal domain while maintaining overall system simplicity through a unified switching framework.
Solution Approach 2:
The patent implements dynamics by making the encoder selection adaptive and dynamic based on the characteristics of each frame. Rather than a fixed single-encoder architecture, the system dynamically switches between encoders and uses a dynamic buffer population strategy that adapts to the current signal type, enhancing versatility without significantly increasing complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A particular method includes encoding a first frame of an audio signal using a first encoder. The method also includes generating, during encoding of the first frame, a baseband signal that includes content corresponding to a high band portion of the audio signal. The method further includes encoding a second frame of the audio signal using a second encoder, where encoding the second frame includes processing the baseband signal to generate high band parameters associated with the second frame.