FIR Filter Coefficient Update for Fast Band Gain Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal processing circuits face challenges in quickly and accurately generating FIR filters with adjusted frequency characteristics, leading to increased processing load and reduced processing speed, as they often require converting between IIR and FIR filters with equivalent transfer functions.
Innovation Solution
A filter generator that performs fast-Fourier-transforming, filtering, and inverse-fast-Fourier-transforming of audio signals, utilizing a weighting factor storage unit, gain coefficient storage, and calculation units to generate new filter coefficients by multiplying the coefficients of the preceding filter by a correction gain, allowing only the coefficients of the adjusted band to be updated.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an IIR filter is used to adjust the gain of a desired frequency band, then the adjustability is improved, but arithmetic errors accumulate and arithmetic accuracy deteriorates
Solution Approach 1:
The frequency band is divided into multiple bands, and weighting factor information is stored for each band. This segmentation allows selective adjustment of specific frequency bands using FIR filters without affecting other bands, achieving both adjustability and arithmetic accuracy.
Solution Approach 2:
The patent changes the parameter representation from direct gain values to weighting factor information that can be multiplied with FFT-transformed audio signals. This parameter transformation enables FIR filters to achieve IIR-like adjustability while maintaining arithmetic accuracy through straightforward multiplication operations.
2Reliability
If an FIR filter is used to maintain high arithmetic accuracy, then arithmetic errors do not accumulate, but it becomes difficult to generate a filter by partially changing the frequency characteristics
Solution Approach 1:
The frequency spectrum is segmented into multiple bands through FFT transformation, and weighting factors are independently adjustable for each band. This segmentation enables partial changes to frequency characteristics while maintaining the arithmetic accuracy of FIR filters.
Solution Approach 2:
The patent transitions from time-domain filter coefficients to frequency-domain weighting factors through FFT transformation. This dimensional change from time domain to frequency domain enables independent adjustment of specific frequency bands while maintaining FIR filter arithmetic accuracy.
3Adaptability or versatility
If a signal processing circuit converts between IIR and FIR filters with equivalent transfer functions to achieve both adjustability and accuracy, then both adjustability and arithmetic accuracy are improved, but the processing load increases and processing speed decreases
Solution Approach 1:
The patent extracts only the necessary weighting factor information for specific frequency bands from the complete filter transfer function. By working directly with FFT-transformed signals and applying weighting factors in the frequency domain, it eliminates the need for complex IIR-FIR conversion while maintaining both adjustability and accuracy.
Solution Approach 2:
The patent introduces FFT transformation as an intermediary step between the audio signal and the FIR filter. This intermediary transformation to the frequency domain enables efficient band-specific adjustment through simple multiplication, avoiding the computational burden of IIR-FIR conversion while achieving both adjustability and arithmetic accuracy.
Data Source
Figure 1
Figure 2
Figure 3(a)~3(b)
AI summary
A filter generator (100) generates a filter on the basis of band information (frequency) and gain characteristics (gain value) set by a user. The filter generator (100) obtains weighting factor information on the basis of the band information selected by the user and calculates a gain difference between a gain value used in a preceding filtering process and the new gain value selected by the user. The filter generator (100) then obtains a correction gain by multiplying the weighting factor information by the gain difference and generates a filter by multiplying a coefficient of the filter used in the preceding filtering process by the correction gain.