Subband Filter Compression for Lower-Complexity QMF Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The computational complexity of filtering in the Quadrature Mirror Filter (QMF) domain, particularly for long impulse response filters like Head Related Transfer Functions (HRTF) filters, is substantial, limiting efficient multi-channel audio processing and rendering over headphones.
Innovation Solution
A filter compressor that examines input subband filter impulse responses, selects higher-value coefficients, and constructs compressed subband filter impulse responses by omitting or setting to zero lower-value coefficients, reducing the number of active filter taps and thereby decreasing computational complexity while maintaining audio quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If long impulse response filters (e.g., HRTF filters) are used in the QMF domain for multi-channel audio processing, then audio quality and spatial rendering accuracy are improved, but computational complexity increases substantially
Solution Approach 1:
The patent extracts and removes insignificant filter coefficients from the QMF domain filter impulse responses. By examining the magnitude of coefficients and selectively removing those below a threshold, the filter length is reduced while maintaining the essential filtering characteristics. This extraction of unnecessary elements directly reduces computational complexity while preserving audio quality.
Solution Approach 2:
The patent applies different treatment to different parts of the filter impulse response based on their significance. Coefficients are examined individually and those with magnitudes below a threshold are set to zero, while significant coefficients are retained. This local differentiation allows the filter to maintain high quality where needed while reducing complexity in less critical regions.
2Productivity
If the number of filter taps is reduced to decrease computational complexity, then processing speed and efficiency are improved, but filtering accuracy and audio quality may deteriorate
Solution Approach 1:
The patent changes the parameters of the filter by selectively modifying coefficient values based on their magnitude. A threshold parameter is introduced, and coefficients below this threshold are set to zero while others are preserved. This parameter-based selection allows the filter to maintain accuracy for significant frequency components while reducing the number of operations for less significant ones.
Solution Approach 2:
The patent applies partial action by not removing all coefficients but only those that are insignificant. This selective removal approach ensures that the essential filtering functionality is preserved while achieving computational reduction. The filter maintains sufficient accuracy by retaining enough significant coefficients to capture the important spectral characteristics.
3Measurement precision
If all filter coefficients are retained to maintain filtering accuracy, then audio quality is preserved, but the data rate and memory requirements increase
Solution Approach 1:
The patent extracts and removes redundant coefficient data from the filter representation. By identifying and eliminating coefficients with magnitudes below a threshold, the data rate required to store and transmit filter information is reduced. This extraction maintains filtering accuracy by preserving only the significant coefficients that contribute meaningfully to the audio output.
Data Source
AI summary
A filter compressor for generating compressed subband filter impulse responses from input subband filter impulse responses corresponding to subbands, which include filter impulse response values at filter taps, includes a processor for examining the filter impulse response values from at least two input subband filter input responses to find filter impulse response values having higher values and at least one filter impulse response value having a value being lower than the higher values, and a filter impulse response constructor for constructing the compressed subband filter impulse responses using the filter impulse response values having the higher values, wherein the compressed subband filter impulse responses do not include filter impulse response values corresponding to filter taps of the at least one filter impulse response value having the lower value or include zero-valued values corresponding to filter taps of the at least one filter impulse response value having the lower value.