Design of customisable audible brainstem response data for training and testing of audiological AI-software

This document comprises a compendium of the design of electroencephalography data for an application to train and test deep-learning networks to be applied on auditory brainstem responses. Despite their essentially artificial nature, the data are derived from experimental sources and are representative of the characteristic behaviour of audiological practice. It is possible to customise them to meet specific requirements, such as different levels of loudness, noise levels, and the amplitude and latency relationship between the six waves, which is characteristic of the responses. The data set under consideration comprises a total of 105 time series, which are divided into three distinct loudness classes: loud (70 dB), medium loud (50 dB), and silent (30 dB).