An audio system for a headrest having a set of R right speaker(s), a set of L left speaker(s), and an electronic chain configured to emit a sound signal from at least a right input signal and a left input signal. The electronic chain includes a separating module configured to obtain, from the right input signal and the left input signal (LIN), at least four mono audio components, and a sound spatialization module configured to apply, to each mono audio component delivered by the separating module, a binaural filter previously selected from the plurality of binaural filters, and dedicated to the audio component, producing a sound whose source position is associated with the previously selected binaural filter.