Subband Linear Prediction in Critically Sampled Filterbanks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio source coding systems face challenges in efficiently encoding subband predictors at low bit rates while minimizing noise shaping artifacts and aliasing effects, particularly in critically sampled filterbanks and linear prediction applications.
Innovation Solution
Implementing a compact description of subband predictors using signal models and direct subband domain predictors, along with cross-subband predictor terms, to reduce computational complexity and aliasing artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If a subband predictor is used in a critically sampled filterbank, then audio signal compression is achieved, but aliasing artifacts appear in the output signal
Solution Approach 1:
The audio signal is divided into multiple subbands using a filterbank, allowing independent prediction and processing in each subband. This segmentation enables targeted noise shaping in each subband while managing aliasing artifacts through cross-subband prediction terms that compensate for aliasing in critically sampled filterbanks.
Solution Approach 2:
Cross-subband prediction terms are introduced as intermediary elements that compensate for aliasing artifacts. These terms use prediction coefficients from adjacent subbands to cancel out aliasing effects, acting as a mediator between the subband prediction process and the final reconstructed signal.
2Loss of substance
If backwards adaptive prediction is used to avoid transmitting predictor description, then bit rate is reduced, but prediction quality degrades at low bit rates
Solution Approach 1:
The predictor description is determined in advance at the encoder and transmitted to the decoder before the actual prediction process. This preliminary action ensures both encoder and decoder use the same predictor parameters, maintaining prediction quality while still achieving bit rate reduction through efficient coding of the predictor description.
Solution Approach 2:
The patent uses model parameters (such as frequency parameters for sinusoidal models) to describe the predictor instead of transmitting full predictor coefficient sets. This parameter change enables compact representation of the predictor at low bit rates while maintaining prediction accuracy through efficient parameter-to-coefficient conversion.
3Productivity
If long term prediction with large delay is implemented, then periodic signal redundancy is exploited, but computational complexity increases due to time domain to subband conversion
Solution Approach 1:
The patent replaces the mechanical conversion process from time domain to subband domain with a direct subband domain prediction approach. By formulating long term prediction directly in the subband domain using appropriate prediction masks and coefficients, the need for additional filterbank operations is eliminated, reducing computational complexity while maintaining prediction effectiveness.
4Ease of manufacture
If in-band prediction is used, then implementation is simple, but noise shaping is limited and alias artifacts persist
Solution Approach 1:
The patent extends prediction from the traditional single-subband (one-dimensional) approach to multi-subband (multi-dimensional) prediction. By incorporating cross-subband prediction terms that utilize information from adjacent subbands, the system achieves improved noise shaping and aliasing compensation while maintaining reasonable implementation complexity through structured prediction masks.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6a
AI summary
The present document relates to audio source coding systems. In particular, the present document relates to audio source coding systems which make use of linear prediction in combination with a filterbank. A method for estimating a first sample (615) of a first subband signal in a first subband of an audio signal is described. The first subband signal of the audio signal is determined using an analysis filterbank (612) comprising a plurality of analysis filters which provide a plurality of subband signals in a plurality of subbands from the audio signal, respectively. The method comprises determining a model parameter (613) of a signal model; determining a prediction coefficient to be applied to a previous sample (614) of a first decoded subband signals derived from the first subband signal, based on the signal model, based on the model parameter (613) and based on the analysis filterbank (612); wherein a time slot of the previous sample (614) is prior to a time slot of the first sample (615); and determining an estimate of the first sample (615) by applying the prediction coefficient to the previous sample (614).