Embedded Audio Equalizer Using MDCT, KBD Window, and OLA

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Audio signal equalizers in embedded systems face high computational demands due to the use of cascade infinite impulse response (IIR) filters, making them inefficient for processing audio signals without excessive computation.

Innovation Solution

The implementation of a Kaiser-Bessel-derived (KBD) window and Overlap-and-Add (OLA) processing method, which window and convert time domain audio signals into frequency bands, equalize them, and then convert back to time domain using Modified Discrete Cosine Transform (MDCT) and Inverse MDCT, while applying KBD windows and OLA processing to minimize signal distortion and allow user-defined filter settings.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a cascade IIR filter is used for audio equalization, then the equalization function is achieved, but the computational load becomes excessively high for embedded systems

Engineering Contradiction:

Improveequalization functionVSAvoidcomputational load

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The audio signal is divided into multiple overlapping sampling blocks that are processed independently through the MDCT-transform domain equalization pipeline. This segmentation allows parallel processing of different signal segments, reducing the overall computational burden on embedded systems while maintaining continuous audio coverage through the overlap-add method.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces the traditional time-domain IIR filter mechanism with a transform-domain approach using MDCT (Modified Discrete Cosine Transform). This substitution moves the equalization operation from direct time-domain filtering to a more efficient frequency-domain processing paradigm, significantly reducing computational complexity for embedded implementations.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Use of energy by moving object

If transform domain processing is used for equalization, then computational efficiency is improved, but signal distortion may occur during time domain to frequency domain conversion

Engineering Contradiction:

Improvecomputational efficiencyVSAvoidsignal distortion

Core Design Contradiction:

Use of energy by moving objectVSReliability

Solution Approach 1:

A KBD (Kaiser-Bessel Derived) window function is applied to each sampling block before the MDCT transformation. This preliminary windowing action tapers the signal at block boundaries, preventing spectral leakage and artifacts that would otherwise cause distortion during the transform. The same KBD window is applied in reverse during the inverse transform to restore the signal properly.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces an overlap-add mechanism as an intermediary process between transformed blocks. By overlapping adjacent sampling blocks and adding them together, the system ensures continuous signal reconstruction without gaps or artifacts, effectively mediating between the segmented transform processing and the continuous time-domain output signal.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS10693430B2Audio signal processing method and audio equalizer

Publication Date: 2020.06.23 REALTEK SEMICON CORP

AI summary

An audio signal processing method and an audio equalizer, both implemented in an embedded system without incurring excessive computation, use a Kaiser-Bessel-derived (KBD) window and an Overlap-and-Add (OLA) processing to eliminate signal distortion of a time domain audio signal during signal conversion and to generate filters according to audio effects desired by a user.