Automated prediction of loudness growth curve using EEG signals

This paper introduces an innovative and automated approach for the prediction of loudness growth curves based on auditory brainstem responses (ABRs), harnessing the power of deep learning and signal processing techniques. Hearing loss, affecting a significant portion of the global population, calls for accurate and efficient assessment methods to improve the quality of life for affected individuals. Our method entails preprocessing ABR signals, extracting informative features via empirical wavelet transform with Fourier Bessel series expansion, and subsequently mapping these features to loudness growth estimates using multi-target regression. Through evaluation employing mean squared error and Frechet distance, our approach demonstrates acceptable performance and consistency across subjects and stimulus levels. Importantly, it overcomes limitations inherent in existing methods that primarily rely on click ABRs and psychoacoustic measures.