Psychoacoustic optimization of a robust feedback active noise controller for headphones

Area of Science:

• Acoustics and Audio Engineering
• Control Systems Engineering
• Signal Processing

Background:

• Active noise cancellation in headphones often uses feedback (FB) controllers.
• Designing broadband FB controllers requires careful tuning to address noise across various frequencies.
• Mixed-sensitivity H-infinity control offers a robust framework for controller design.

Purpose of the Study:

• To develop an optimization routine using a genetic algorithm (GA) for mixed-sensitivity H-infinity feedback controllers.
• To evaluate the effectiveness of GA-optimized controllers targeting different acoustic metrics: sound pressure level (SPL), perceived loudness, and perceived sharpness.
• To assess the performance of controllers designed for headphones with varying spectral characteristics.

Main Methods:

• Implementation of a genetic algorithm (GA) for optimizing the boundaries of a mixed-sensitivity H-infinity controller.
• Design of feedback (FB) controllers with distinct target functions: SPL, perceived loudness, and perceived sharpness.
• Testing with audio signals exhibiting diverse spectral properties to evaluate controller performance.

Main Results:

• The genetic algorithm (GA) approach successfully optimized controllers for minimizing sound pressure level (SPL) and perceived loudness.
• Controllers optimized for perceived loudness and SPL demonstrated effective noise cancellation.
• Optimization targeting perceived sharpness yielded results that are only recommended to a limited extent, indicating challenges in precisely controlling this metric.

Conclusions:

• Genetic algorithm (GA) optimization is a viable method for designing feedback (FB) controllers for active noise cancellation in headphones, particularly for SPL and loudness targets.
• Controlling perceived sharpness using this optimization approach presents limitations and requires further investigation.
• The study highlights the trade-offs and challenges in optimizing controllers for multiple, potentially conflicting, acoustic perception goals.