This is published under the terms of CC BY-NC.

Downloaded: 781 times

Abstract

The cocktail party effect refers to a challenging problem in speech perception where one is able to selectively attend to one sound source in a noisy and multi-talk environment. The recent studies in neuroscience and psychoacoustics shed light on how the human brain solves the cocktail party problem, that inspires many computational solutions. With the advent of novel physiological techniques and deep learning algorithms, it is now possible to effectively detect auditory attention based on brain signals. In this paper, we provide a comprehensive overview of the most recent EEG-based auditory attention detection techniques and the methods to evaluate their performance. We examine both statistical and deep learning approaches, exploring their strengths and limitations. Furthermore, we also point out the gaps between the state-of-the-art and the practical needs in real-world applications. We also offer an overview of the available resources for EEG-based auditory attention detection research.