Changing the HRTF profile.
Unique to each individual, HRTFs are not universal. The profiles offered by default in HOLOPHONIX for binaural spatialization suit a majority of people. It is possible to import other HRTF files in SOFA format (“Spatially Oriented Format for Acoustics”)[1] in order to select a filtering profile better suited to one’s own hearing.
Adjusting an HRTF profile to one’s morphology.
The measurements taken refer to a subject, to the head of a listener and therefore to a certain interaural distance (the distance between the two ears of the listener who participated in the measurements). This reference distance can be modified by the user, directly impacting the Interaural Time Difference (ITD). The “Interaural Time Delay” parameter gives the user the ability to adjust the model being used.
Compensation filtering of the listening system: the headphones.
It is also possible to introduce compensation filtering for the headphones being used. Headphones introduce filtering related to their own frequency response. The principle is similar to loudspeaker equalization that would be applied for multi-channel spatialization in a room.
The SOS data format rather than HRIR.
Impulse responses can be approximated by a network of cascaded filters (passing the signal from one filter to another) called biquadratic filters[2] of second order (SOS, Second Order Section). The advantage is a reduced processing cost compared to a convolution product.
Compensation of the electroacoustic measurement chain.
It is more advantageous to use measured models whose measurement chain has been compensated. Each SOFA file should ideally specify the following criteria:
- Whether the measurement chain has been compensated or not.
- Sampling frequency of the measurements.
- Measurement format: HRIR or SOS.
- SOS order (number of cascaded filters) if applicable.
[1] “SOFA convention,” 2013.: https://www.sofaconventions.org/mediawiki/index.php/SOFA_(Spatially_Oriented_Format_for_Acoustics).
[2] A biquadratic filter, in digital audio, is a combination of an incoming audio sample (x(n)) with the two previously input samples (x(n-1)) and (x(n-2)) and the two previously output samples (y(n-1)) and (y(n-2)). This combination yields the filtered sample (y(n)) according to the relation: (y(n)=b_{0}x(n)+b_{1}x(n-1)+b_{2}x(n-2)-a_{1}y(n-1)-a_{2}y(n-2)). The coefficients (a_{i}) and (b_{i}) are constants calculated according to the filter type, its gain, frequency, and quality factor. Simple to use and implement, it allows the main filters used in audio to be modelled at very low computational cost. Cascaded (SOS), it enables the design of complex filters at high filter orders.