Perceptual optimization for headphone listening
Loading...
Date
2025-02-04
Authors
Ζάχος, Παναγιώτης
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This dissertation explores advanced techniques in audio signal processing to enhance the listening experience of headphone users via the perceptual optimization of algorithms applied in personal audio devices. It draws inspiration from established market trends coupled with common issues encountered during modern audio reproduction scenarios, as described in the Motivation section of Chapter 1. In such scenarios, the user may be exposed to different types of disturbing noise, which reduce the overall audio quality and lead to a less enjoyable listening experience.
To address these challenges, this dissertation focuses on adaptive equalization and system identification, novel time-domain psychoacoustic as well as spatially aware active noise control systems, and binaural sound source localization.
The theoretical foundations for the techniques developed in this dissertation are established in Chapter 2, covering audio equalization, adaptive filtering, active noise control, phased array processing, and Artificial Intelligence, with a focus on Neural Networks. The dissertation concludes with a summary of contributions in the field of audio engineering and potential future research directions in Chapter 8.
Chapter 3 introduces an adaptive equalization method using a parallel biquad filter structure, capable of compensating for headphone frequency responses with low filter order, making it suitable for real-time applications. This method requires minimal user interaction and effectively models arbitrary system responses. It is evaluated using both objective and subjective criteria, demonstrating its efficacy.
Chapter 4 presents a psychoacoustic ANC system that computes a masking threshold based on the streamed audio signal and models a residual noise filter to cancel unmasked noise. This approach reduces noise in the most audible frequency regions while disregarding inaudible noise components, achieving a more perceptually compliant result. It leverages psychoacoustic principles to improve noise cancellation without compromising audio quality.
Chapters 5 and 6 discuss a spatially aware, targeted beamforming ANC technique that utilizes a novel, purpose-designed disturbance metric to enhance noise suppression in specific directions. This system dynamically steers the beamformer toward the most disturbing source, improving the overall listening experience.
Chapter 7 proposes a binaural parameter compression technique applied to a sound source localization model. The method significantly reduces the size of input features—Interaural Level Difference (ILD) and Interaural Time Difference (ITD)—while maintaining high localization performance across a range of reverberant acoustic scenes and source types, including speech and music.
Description
Keywords
Headphones, Equalization, Active noise control, Psychoacoustic models, Auditory perception, Beamforming, Direction of arrival estimation