Eliciting adaptation to non-individual HRTF spectral cues with multi-modal training presence

Spatial hearing relies on the exploitation of acoustic cues, the well known Head Related Transfer Function (HRTF), by the auditory system. Binaural synthesis is the creation of 3D virtual sound sources based on the reproduction of these acoustic cues to the listener’s ears over headphones.[1,2] From these HRTFs, one can extract inter-aural time differences (ITD), principally linked with the lateralization of sound sources, as well as spectral cues, which are more concerned with elevation. A common problem is that HRTFs are extremely morphologically dependent (shape of the outer ears, head dimensions, torso) thus providing idiosyncratic cues. Therefore, methods for generating individualized HRTFs is the topic of numerous studies [3,4] in order to avoid typical binaural synthesis artifacts such as front/back confusions or angular distortions primarily in the vertical plane.[5] In fact, although we can claim high authenticity with binaural rendering, it is not always a natural situation (for example the lack of dynamic cues due to head movements in a static rendering [6]) and, even in the best conditions, some artifacts are still present. It is often found that a period of adaptation is needed to adjust to rendering anomalies, and studies using binaural spatialization mention trial sessions for the subjects to become acquainted with the technique. This is often term ed the “learning effect”.[5,6,7] An interesting extension of this phenomenon is to consider the individualization problem in a reverse sense, that is, instead of adapting HRTFs to the individual, one forces the auditory system to adapt to the non-individual HRTF. There is evidence of ongoing spatial calibration in the adult human auditory system. Work by Hofman et al [8] showed that subjects with modified pinnae (using inserts), following a dramatical degradation of sound elevation perception, steadily reacquired localization performance over time. The point of the present study is to make further investigations on this auditory localization adaptation in the context of a virtual sound environment. Many questions can be raised, such as how to elicit adaptation, which acoustic cues are involved, what is the duration of adaptation, how long does it last, or where in the auditory system does it occur. This study focuses on adaptation to the spectral component of the HRTF, as was done by Hofman, but directed towards virtual rendering with the condition that adaptation should be rapid. Rapid auditory training for lateralization with paired audio/visual stimuli or induced plasticity with ventriloquism effect has been previously demonstrated.[9,10,11] Here we induce adaptation to non-individual HRTFs spectral cues with a quick exploration of the spatial map by an auditory-kinesthetic process. This idea is in contrast to recent studies involving spatial hearing in blind subjects which have reported that early-blind subjects, commonly believed to have better localization performance through hyper-compensation of their visual loss, exhibit less accuracy in elevation estimation, suggesting that the auditory system may require visual feedback for calibration.[12,13] Some exceptions were noted which lead to the hypothesis that auditory calibration may be achieved via other multimodal interactions. A proprioceptive feedback of auditory spatial information has been used by giving blindfolded subjects control of a virtual sound source spatialized at the hand position. The principle is that the listener associates the source position with the acoustic cues used for the binaural rendering through the constant and innate awareness of one’s own hand position. This modality has the advantage of offering natural interactivity with perception/action coupling and is not limited to the visual field, thus allowing the user to explore the entire auditory sphere.