Auditory Attention Detection in Real-Life Scenarios Using Common Spatial Patterns from EEG

Auditory attention detection (AAD) methods based on electroencephalography (EEG) could be used in neuro-steered hearing devices to help hearing-loss people improve their hearing ability. However, previous studies have mostly obtained EEG data in laboratory settings which limits the practical application of neuro-steered hearing devices. In this study, we employ a common spatial pattern (CSP) algorithm to perform AAD using EEG signals collected by a wireless mobile EEG system, from real-life scenarios when people are walking and sitting. The results show that the CSP method can achieve AAD accuracy between 81.3% and 87.5% when using different decision windows (1 s- 30 s), which is better than previous methods based on linear mapping methods and convolutional neural networks (CNN). This proves that the CSP algorithm can decode people's attention efficiently even outside the laboratory. Analysis of EEG frequency bands shows that the delta and beta bands have high activity in attention tasks.