Sinusoidal Trajectory Segmentation for High Frequency Audio Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The MPEG-H 3D Audio Codec faces challenges in accurately representing and reconstructing high frequency tonal components, particularly for sounds with prominent high frequency partials, leading to distortion and artifacts due to the limitations of the eSBR tool, which struggles with highly pitched sounds and frequency variations.
Innovation Solution
The proposed High Frequency Sinusoidal Coding (HFSC) tool uses sinusoidal modeling to encode prominent tonal components as sinusoidal trajectories, partitioning them into segments for efficient coding and decoding, and integrating this process within the MPEG-H 3D Audio Core Encoder to enhance the representation of high frequency components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If the eSBR tool is used for bandwidth extension in MPEG-H 3D Audio Codec, then the audio signal can be compressed, but distortion and artifacts occur in high frequency tonal components
Solution Approach 1:
The audio signal is segmented into tonal components and non-tonal components. The HFSC tool specifically processes tonal components using sinusoidal modeling, while other components are handled by the eSBR tool. This segmentation allows targeted processing that preserves high frequency tonal accuracy while maintaining overall compression efficiency.
Solution Approach 2:
The HFSC tool acts as an intermediary between the eSBR tool and the final audio output. It receives the compressed audio signal, identifies and processes tonal components using sinusoidal modeling to correct distortions, then passes the corrected signal to the output. This intermediary processing layer resolves the contradiction by fixing high frequency artifacts without compromising the overall compression achieved by eSBR.
2Productivity
If Huffman codes are used for encoding DCT coefficients, then coding efficiency improves by about 10%, but memory requirements increase for Huffman tables storage
Solution Approach 1:
The invention extracts only the essential Huffman tables needed for encoding the specific DCT coefficient patterns that occur in sinusoidal coding. By analyzing the actual data characteristics and removing unnecessary table entries, the solution achieves the 10% coding efficiency improvement while minimizing memory requirements to only what is strictly necessary.
3Device complexity
If sinusoidal trajectories are partitioned into segments for encoding, then coding complexity is reduced, but additional processing steps are required
Solution Approach 1:
Sinusoidal trajectories are partitioned into segments that correspond to groups of segments (GOS) in the bitstream structure. This segmentation allows the decoder to process trajectories in manageable chunks, reducing memory requirements and computational complexity while maintaining accuracy. The segment boundaries are carefully chosen to align with natural discontinuities in the sinusoidal data.
Solution Approach 2:
The encoder performs preliminary segmentation of sinusoidal trajectories into segments before encoding. This preliminary organization of data into GOS-aligned segments simplifies the subsequent encoding and decoding processes, as the structure is already optimized for the bitstream format. The preprocessing work reduces the complexity of the main encoding algorithm.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments provide an audio signal encoding method comprising the steps of: collecting audio signal samples (114), determining sinusoidal components (312) in subsequent frames, estimation of amplitudes (314) and frequencies (313) of the components for each frame, merging thus obtained pairs into sinusoidal trajectories, splitting particular trajectories into segments, transforming (318, 319) particular trajectories to the frequency domain by means of a digital transform performed on segments longer than the frame duration, quantization (320, 321) and selection (322, 323) of transform coefficients in the segments, entropy encoding (328), outputting the quantized coefficients as output data (115), wherein segments of different trajectories starting within a particular time are grouped into Groups of Segments (GOS), and the partitioning of trajectories into segments is synchronized with the endpoints of a Group of Segments (GOS).