Personal devices (smartphones, tablets, etc.) have multiple functions and are the main vehicle for distributing musical contents today. As the audience drifts towards these new devices, binaural listening over headphones is gaining ground and symbolizes the very notion of personal experience by theoretically giving access to the reproduction of three-dimensional sound scenes. This binaural reproduction is based on the dynamic filtering of the sound sources by Head-Related Transfer Functions (HRTF) previously measured on the head of a listener or of a dummy head. However, the individual dependency of these HRTFs has limited their general public use to date.
As part of the BiLi project (FUI funding), the HRTF measurement system in IRCAM’s anechoic chamber received a major hardware and software upgrade that significantly increases the spatial resolution compared to the available databases. The new spatial sampling (1680 directions) allows a decomposition into high order spherical harmonics, useful for various postprocessing steps (spatial interpolation, HOA / binaural transcoding).
The sharing of HRTF databases within the international community has motivated the standardization of an exchange format. Resulting from an international collaboration, the AES69 SOFA (Spatially Oriented Format for Acoustics) format, approved by the Audio Engineering Society’s standardization body, allows the storage of spatial acoustic data such as HRTFs or Spatial Room Impulse Responses (SRIRs). IRCAM has set up an OPenDAP (Open-source Project for Network Data Access Protocol) server that hosts various HRTFs databases in SOFA format and to which client applications (web applications, Matlab, etc.) can send download requests (a specific HRTF, a complete head, etc.).
Work on the individualization of HRTFs is continuing, notably in the field of HRTF estimations that do not require either acoustic or morphological measurements. Some methods exploit available databases to guide the user in the selection of the most appropriate set of HRTFs (RASPUTIN project, p. 36). A proposed new approach is based on the use of deep learning methods (HAIKUS project, p. 37) and blind analysis on binaural signals recorded with in-ear microphones under unsupervised conditions (reverberant environment, any signals, sources and moving listener).
IRCAM Team : Acoustic and Cognitive Spaces
