GB2620138A - Method for generating a head-related transfer function - Google Patents

Abstract

A method for generating a Head-Related Transfer Function, HRTF, for an individual user, the method comprising: obtaining a predetermined default HRTF model; obtaining a hearing factor for a user; and generating a personalised HRTF model for the user by modifying the default HRTF model based on the obtained hearing factor. A separate invention discloses the creation of a default HRTF database by the averaging of individual hearing factors. Preferably the hearing factors comprises at least one of an interaural time delay (ITD), and interaural level difference (ILD), a spectral peak or notch corresponding to a physical feature of the user, and a height compensation factor. Preferably a head measurement of the user is obtained and the ITD is calculated based on said measurement.

Description

METHOD FOR GENERATING A HEAD-RELATED TRANSFER FUNCTION TECHNICAL FIELD

The following disclosure relates to methods for generating head-related transfer functions (HRTFs). HRTFs are used for simulating, or compensating for, how sound is received by a listener in a 3D space. For example, HRTFs are used in 3D audio rendering, such as in virtual surround sound for headphones.

BACKGROUND

HRTFs (Head Related Transfer Functions) describe the way in which a person hears sound in 3D, and can change depending on the position of the sound source. Typically, in order to calculate a received sound Kt t), a signal x(f; 0 transmitted by the sound source is combined with (e.g. multiplied by, or convolved with) the transfer function 11(0.

HRTFs are individual to each person and depend on things like the size of the head and shape of the ear. In 3D audio rendering, it is beneficial to try and personalise the HRTF filters used to best match the person listening to the audio.

For example, this can mean that the person will hear audio rendered through headphones in a similar way to how they hear 3D audio in real life.

However, generating an HRTF for a specific user conventionally requires collecting a large amount of data about how the user perceives sound from audio sources at different 3D positions, and it is difficult to expect the user to go through a lengthy configuration process to obtain this data. As a result, rather than providing a personalised HRTF, sound systems typically provide a limited number of pre-set HRTFs, and the user can select a most suitable one of the pre-set HRTFs.

Accordingly, it is desirable to provide a way to generate an HRTF that is personalised for an individual user, without requiring a lot of data about the user or requiring a lengthy configuration process.

SUMMARY

According to a first aspect, the present disclosure provides a method for generating a Head-Related Transfer Function, HRTF, for an individual user, the method comprising: obtaining a predetermined default HRTF model; obtaining a hearing factor for a user; and generating a personalised HRTF model for the user by modifying the default HRTF model based on the obtained hearing factor Herein, an HRTF model can be an HRTF (i.e. a transfer function indicating how an audio signal is changed between being output by an audio source and being heard by the user). Alternatively, the HRTF model can be a parametric description of an HRTF (i.e. a set of amplitude parameters each of which indicates an amplitude of a component audio filter in the HRTF, such that the when the personalised HRTF model is combined with the set of component audio filters, the HRTF of the user is constructed).

By using a template, the method can start with a HRTF model that is roughly suitable for all users, and the process for customizing the template can be as elaborate or simple as the individual user is willing to tolerate, while improving the accuracy with which the HRTF model represents the user's natural experience of listening to audio in a 3D environment.

In a first embodiment of the first aspect, the default HRTF model comprises a first set of component audio filters and the personalised HRTF model comprises a second set of component audio filters. By storing the HRTF model as a set of component audio filters, the HRTF model can be immediately used for filtering an audio signal.

According to the first embodiment, modifying the default HRTF model may comprise at least one of: adding a component audio filter to the first set of component audio filters to generate the second set of component audio filters; removing a component audio filter from the first set of component audio filters to generate the second set of component audio filters; and performing signal processing on a component audio filter of the first set of component audio filters to generate the second set of component audio filters.

In a second embodiment of the first aspect, the default HRTF model comprises a first set of parameters and the personalised HRTF model comprises a second set of parameters. By storing the HRTF models as sets of parameters, it becomes unnecessary to include frequency spectra in the models, and the data size and computational complexity of the models can be reduced.

According to the second embodiment, modifying the default HRTF model may comprise at least one of: adding a parameter to the first set of parameters to generate the second set of parameters; removing a parameter from the first set of parameters to generate the second set of parameters; and modifying a parameter of the first set of parameters to generate the second set of parameters.

According to the second embodiment, the method may further comprise synthesizing a personalised HRTF based on the second set of parameters and a basis set of component audio filters.

According to the first aspect, the hearing factor may comprise at least one of: an interaural time delay; an interaural level difference; a spectral peak or notch corresponding to a physical feature or set of features of the user; and a height compensation factor.

Furthermore, the method may comprise: obtaining a head measurement of the user; calculating the interaural time delay based on the head measurement.

Alternatively, the method may comprise: obtaining an image of an ear, head and/or torso of the user; performing image processing on the image to identify the physical feature; calculating the spectral peak or notch based on the physical 25 feature.

According to a second aspect, the present disclosure provides a method of generating a default Head-Related Transfer Function, HRTF, model, the method comprising: for each of a plurality of users and each of a plurality of averageable hearing factors, obtaining the averageable hearing factor for the user; calculating an average of each of the averageable hearing factors for the plurality of users; and generating the default HRTF model based on the averageable hearing factors.

In a first embodiment of the second aspect, the default HRTF model comprises a first set of component audio filters.

In a second embodiment of the second aspect, the default HRTF model comprises a first set of parameters.

According to the second embodiment of the second aspect, the method may further comprise synthesizing a default HRTF based on the first set of parameters and a basis set of component audio filters.

In embodiments of the second aspect, the averageable hearing factors may include at least one of: an interaural time delay; an interaural level difference and a low-order spectral response.

In embodiments of the second aspect, the averageable hearing factors may exclude at least one of: spectral peaks and notches corresponding to physical features; and a height compensation factor The first aspect and the second aspect may be combined into a single method comprising generating a default HRTF model followed by generating an HRTF for an individual user According to a third aspect, there is provided an HRTF generator configured to perform a method according to the first aspect and/or the second aspect. The HRTF generator may be implemented in headphones, in a base unit configured to communicate with headphones, or may be independent from headphones. For example, the HRTF generator may be implemented in a server or cloud service.

The HRTF generator may be implemented using a memory and processor, wherein the memory stores computer-readable instructions which, when executed by the processor, cause the processor to perform a method according to the first aspect or according to the second aspect. Alternatively, the HRTF generator may comprise hardware, such as an ASIC, which is specifically adapted to perform a method according to the first aspect or according to the second aspect.

According to a fourth aspect, there is provided a computer program comprising computer-readable instructions which, when executed by one or more processors, cause the one or more processors to perform a method according to the first aspect or according to the second aspect.

According to a fifth aspect, there is provided a non-transitory storage medium storing computer-readable instructions which, when executed by one or more 10 processors, cause the one or more processors to perform a method according to the first aspect or according to the second aspect.

According to a sixth aspect, there is provided a signal comprising computer-readable instructions which, when executed by one or more processors, cause the one or more processors to perform a method according to the first aspect or according to the second aspect.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1A schematically illustrates HRTFs in the context of a real sound source offset from a user; Fig. 1B schematically illustrates an equivalent virtual sound source offset from a user in audio provided by headphones; Fig. 2 illustrates head width as a hearing factor for generating an HRTF; Fig. 3 illustrates obtaining pinna features as hearing factors for generating an HRTF; Figs. 4A and 4B schematically illustrate examples of a personalised HRTF model according to the invention; Fig. 5 schematically illustrates a method for generating a HRTF for an individual user; Figs. 6A and 6B schematically illustrate methods for generating an HRTF for an individual user; Fig. 7 schematically illustrates a method of generating a default HRTF model.

DETAILED DESCRIPTION

Fig. 1A schematically illustrates HRTFs in the context of a real sound source offset from a user.

As shown in Fig. 1A, the real sound source 10 is in front of and to the left of the user 20, at an azimuth angle 0 in a horizontal plane relative to the user 20. The effect of positioning the sound source 10 at the angle 0 can be modelled as a frequency-dependent filter hL(61) affecting the sound received by the user's left ear 21 and a frequency-dependent filter hR(0) affecting the sound received by the user's right ear 22. The combination of hL(0) and hR(0) is a head-related transfer function (HRTF) for azimuth angle O. More generally, the position of the sound source 10 can be defined in three dimensions (e.g. range r, azimuth angle 0 and elevation angle cp), and the HRTF can be modelled as a function of three-dimensional position of the sound source relative to the user.

The sound received by the each of the user's ears is affected by numerous hearing factors, including the following examples.

* The distance wH between the user's ears 21, 22 (which is also called the "head width" herein) causes a delay between sound arriving at one ear and the same sound arriving at the other ear (an interaural time delay). This distance wH is illustrated in Fig. 2. Other head measurements can also be relevant to hearing and specifically relevant to interaural time delay, including head circumference, head depth and/or head height.

* Each of the user's ears has a different frequency-dependent sound sensitivity (i.e. the user's ears have an interaural level difference).

* The shape of the user's outer ear (pinna) creates one or more resonances or antiresonances, which appear in the HRTF as spectral peaks or notches.

Fig. 3 illustrates pinna features 320, 330. In this example the pinna features are contours of the ear shape which affect how sound waves are directed to the auditory canal 310. The length and shape of the pinna feature affects which sound wavelengths are resonant or anfiresonant with the pinna feature, and this response also typically depends on the position and direction of the sound source. Fig. 3 is referenced again later in the specification when describing how to obtain pinna features in a method of generating an HRTF for an individual user. Further spectral peaks or notches may be associated with other physical features of the user. For example, the user's shoulders and neck may affect how sound is reflected towards their ears. For at least some frequencies, more remote physical features of the user such as torso shape or leg shape may also be relevant.

Each of these factors may be dependent upon the position of the sound source. As a result, these factors are used in human perception of the position of a sound source.

Wien the sound source is distant from the user, the HRTF is generally only dependent on the direction of the sound source from the user On the other hand, when the sound source is close to the user the HRTF may be dependent upon both the direction of the sound source and the distance between the sound source and the user Fig. 1B schematically illustrates an equivalent virtual sound source offset from a user in audio provided by headphones 30. Herein "headphones" generally includes any device with an on-ear or in-ear sound source for at least one ear, including VR headsets and ear buds.

As shown in Fig. 1B, the virtual sound source 10 is simulated to be at the azimuth angle 0 in a horizontal plane relative to the user 20. This is achieved by incorporating the HRTF for a sound source at azimuth angle 0 as part of the sound signal emitted from the headphones. More specifically, the sound signal from left speaker 31 of the headphones 30 incorporates hi_(0) and the sound signal from right speaker 32 of the headphones 30 incorporates hR(0). Additionally, inverse filters hie and h-lRo may be applied to the emitted signals to avoid perception of the "real" HRTF of the left and right speakers 31, 32 at their positions LO and RO close to the ears.

In general, HRTFs are complex and cannot be straightforwardly modelled as continuous function of frequency and sound source position. Instead, HRTFs are commonly stored as tables of HRTFs for a finite set of sound source positions, and interpolation may be used for source sources at other positions. An HRTF for a given sound source position may be stored as a Finite Impulse Response (FIR) filter, for example. In one case, the set of sound source positions may simply include positions spaced across a range of azimuth angles 0 (without addressing effects of range or elevation). In some cases, elevation may be modelled, for example by using a correcting factor that affects left and right ears symmetrically.

Even with such finite sets, a significant amount of data must be obtained to model the frequency-dependent and position-dependent HRTF for an individual user.

Accordingly, the present invention seeks to provide a more convenient way of generating a personalised HRTF model suitable for simulating 3D sound sources in headphones.

Figs. 4A and 4B schematically illustrate the information content of example personalised HRTF models according to the invention.

The personalised HRTF model 400 includes a default HRTF model 410 and one or more modifying elements.

In the example of Fig. 4A, the default HRTF model 410 comprises a first set of component audio filters 411, 412, and the personalised HRTF model comprises one or more personalised component audio filters 421, 423.

Each component audio filter may correspond to a different sound source position and a respective ear, as discussed above.

Furthermore, multiple component audio filters may be combined to give the HRTF for a single sound source position and single ear For example, one component audio filter may correspond to a response characteristic of the inner ear (which may be independent of sound source position). Another component may correspond to a low-order spectral response of the outer ear. Yet another component may correspond to a spectral peak or notch associated with a shape feature of the outer ear (pinna) or associated with other physical features of the user such as the shape of the head, neck or shoulders.

Further component audio filters may be based on other known properties of hearing in 3D space. For example, the first set of component audio filters may comprise a low-pass filter associated with sound source positions behind the user The combination of the first set of component audio filters 411, 412 and the personalised component audio filters 421, 423 gives a second set of component audio filters which form a personalised HRTF.

In some cases, a personalised component audio filter 421 replaces a corresponding element 411 of the first set to generate the second set. For example, a component audio filter corresponding to a low-order spectral response of an average outer ear (e.g. calculated from measured responses by a set of test users) is replaced with a low-order spectral response calculated from measured responses of the specific user In other cases, a personalised component audio filter 423 may be added to the first set to generate the second set. For example, in a case where the default HRTF model 410 is designed to be independent of individual ear shape, the default HRTF model 410 may have no component audio filters corresponding to spectral peaks or notches associated with the shape of the outer ear (pinna). The personalised HRTF model may become more accurate than the default HRTF model for a given user by adding a component audio filter 423 which is based on ear shape.

In yet further cases, generating the second set of component filters may comprise performing signal processing on a component audio filter of the first set to generate the second set. For example, if the head width or other head measurement of a specific user is different from an average head measurement on which the default HRTF model was based, then the interaural time delay (ITD) between the left and right ears 21,22 may need to be adjusted. In one example, this is achieved for a user ITD A and an average ITD A0 by multiplying the frequency-domain function d(f) = ei2Thf(°0-°) by a corresponding component audio filter of the first set.

On the other hand, Fig. 2B illustrates a personalised HRTF model 450 comprising a set of parameters. The personalised HRTF model can be used to synthesize a personalised HRTF based on the set of parameters, together with a basis set of component audio filters. By storing the personalised HRTF model as a set of parameters, it becomes unnecessary to include frequency spectra in the model, and the data size of the model can be reduced. In one implementation, the basis set comprises a set of Gaussian spherical functions as spatial distributions, where each Gaussian spherical function is used to weight a respective frequency domain filter. Parametric modelling of the Gaussian curves was used to provide a basis set which can be tweaked parametrically as necessary. This also has the advantage of providing an efficient way to store the HRTF model in memory.

The default HRTF model 460 comprises a first set of parameters. These may correspond to one or more hearing factors averaged for a sample set of users. The first set of parameters may for example include a default head measurement 461 and a default interaural level difference 462.

The personalised HRTF model 450 further comprises one or more personalised parameters 471, 473, 474.

The combination of the first set of parameters 461, 462 and the personalised parameters 471, 473, 474 gives a second set of parameters which form a personalised HRTF model In some cases, a personalised parameter 471 replaces a corresponding element 461 of the first set to generate the second set. For example, a user head measurement parameter 471 may replace a default head measurement parameter 461 In other cases, a personalised parameter 473, 474 may be added to the first set to generate the second set. For example, the default HRTF model 460 may not include any features which depend upon the physical shape of a user, e.g. ear shape (pinna), because it is not meaningful to average these features across multiple users. The personalised HRTF model may become more accurate than the default HRTF model for a given user by adding an indication of one or more physical features such as pinna features 473, 474 for a given user. For example, personalised parameters 473, 474 may correspond to the pinna features 320 and 330 shown in Fig. 3.

Once the personalised HRTF model 450 is generated in parametric form, these parameters can be used to synthesize a personalised HRTF. For example, a basis set of component audio filters may be defined, and the personalised HRTF may be constructed from the set of component audio filters, by selecting and weighting component audio filters according to the second set of parameters.

Additionally or alternatively, the second set of parameters may be used as inputs to a process for generating customised audio filters. For example, a pinna notch may have a generally defined spectral shape corresponding to a general shape of the human ear, but the frequency position may depend on the user. In that case, a user-specific pinna notch may be generated based on each of the pinna feature parameters 473, 474, and included as part of the personalised HRTF.

Fig. 5 schematically illustrates a method for generating a HRTF for an individual user. This method may be performed by an HRTF generator. The HRTF generator may be implemented in a set of headphones 30, in a base unit configured to communicate with the headphones, or may be independent from the headphones. In one example, the HRTF generator could be implemented in an interactive audio-visual system such as a game console which is associated with the headphones 30. In another example, the HRTF generator may be implemented in a server or cloud service. The HRTF generator may be implemented using a general-purpose memory and processor together with appropriate software. Alternatively, the HRTF generator may comprise hardware, such as an ASIC, which is specifically adapted to perform the method.

Referring to Fig. 5, at step S510, the HRTF generator obtains a default HRTF model 410, 460. The default HRTF model have been previously generated by the HRTF generator or may have been generated by a separate device. Details of the process of generating a default HRTF model are discussed below with reference to Fig. 7.

At step S520, the HRTF generator obtains a hearing factor for a user.

In one embodiment the HRTF generator is configured to obtain an interaural time delay as a hearing factor. Referring back to Figs. 1A and 2, it can be readily understood that the interaural time delay for a given sound source position depends upon the width wH of the user's head and upon the angle of the sound source relative to the user. In one example, the width of the users head is input by the user via a user interface such as a game controller, keyboard, touchscreen or voice command. Alternatively, the width of the user's head may be obtained from a photo. For example, the HRTF generator may be connected to a camera, or the user may electronically provide a photo to the HRTF generator. As a further alternative, the width of a users head may be sensed by the headphones 30 based on a distance between the left and right speakers 31, 32 when worn by the user.

Once the width wH of the user's head (and/or another head measurement for the user) has been obtained by the HRTF generator, the HRTF generator can calculate the interaural time delay for one or more virtualised sound source positions based on the head measurement.

Referring back to Fig. 3, in another embodiment, the HRTF generator is configured to calculate a spectral peak or notch corresponding to a pinna shape feature 320, 330. For example, the HRTF generator may obtain an image 300 of an ear 21 of the user The HRTF generator may then use known image processing techniques such as edge detection to identify a feature of the pinna, and then calculate a spectral peak or notch based on the feature. The feature may for example be a ridge contour having a measurable length, and resonances or antiresonances may be predicted based on the length of the ridge contour.

In another embodiment, the HRTF generator is configured to obtain a height compensation factor as the hearing factor for the user. Auditory perception of elevation angle is different from perception of sound source azimuth angle in a horizontal plane, because it is not always possible to use differences between the two ears (e.g. interaural time delay) to identify the elevation angle. Nevertheless, the shape of the outer ear (pinna) can have some effect on sound source height perception as with horizontal perception. Additionally or alternatively, the height compensation factor may include the user's height or dynamic elevation. This may be used to simulate sound sources above or below the user's head in a virtualised space, or may be taken into account when simulating sound reflections off the ground as part of the HRTF. For example, vertical motion of the headphones 30 may be tracked and provided to the HRTF generator Further hearing factors may be measured with in-ear microphones, or may be measured based on physiological feedback or conscious feedback from the user For example, the user may be invited to listen to sound received from a virtual source according to multiple HRTFs and indicate which is a better simulation of the equivalent real sound source.

At step S530, the HRTF generator generates a personalised HRTF model for the user by modifying the default HRTF model based on the obtained hearing factor Referring to Fig. 4A, when the default HRTF model comprises a first set of component audio filters, step S530 may comprise adding a new personalised component audio filter 421, 423 and/or modifying or removing a default component audio filter 411, in order to generate a second set of component audio filters for the personalised HRTF model.

An example of this process is shown in the schematic flowchart of Fig. 6A, wherein the default HRTF model comprises a default set of component audio filters 610, and a hearing factor 620 is obtained for the user The hearing factor 620 is used to generate a personalised component audio filter 630. The personalised component audio filter 630 may for example be selected from a table of predetermined component audio filters based on the hearing factor 620. Alternatively, the personalised component audio filter 630 may be generated as a function of a predetermined component audio filter based on the hearing factor 620. Furthermore, the personalised component audio filter 630 may be generated by processing a filter of the default set of component audio filters 610.

As another implementation, the "personalised component audio filter 630" may comprise a decision to omit one of the default set 610 from the personalised HRTF.

The personalised component audio filter 630 is then combined with the default set of audio filters 610 at step 640, to produce the personalised HRTF 660 comprising a personalised set of component audio filters Referring to Fig. 4B, when the default HRTF model comprises a first set of parameters, step S530 may comprise adding a new personalised HRTF parameter 471, 473, 474 and/or modifying or removing a default parameter 461, in order to generate a second set of parameters for the personalised HRTF model.

An example of this process is shown in the schematic flowchart of Fig. 6B, wherein the default HRTF model comprises a default set of parameters 615, and a hearing factor 620 is obtained for the user.

The hearing factor 620 is used to generate a personalised HRTF parameter 635. The personalised HRTF parameter 635 may for example be selected from a table of predetermined parameters based on the hearing factor 620. Alternatively, the personalised HRTF parameter 635 may be generated as a function of a predetermined parameter based on the hearing factor 620. Furthermore, the personalised HRTF parameter 635 may be generated by modifying a parameter of the default set of parameters 615. As another implementation, the "personalised HRTF parameter 635" may comprise a decision to omit one of the default set of parameters 615 from the personalised HRTF.

The personalised HRTF parameter 635 is then combined with the default set of HRTF parameters 615 at step 645, to produce the personalised set of HRTF parameters 650.

The personalised set of HRTF parameters 650 is then used as the input to a model for generating a personalised HRTF 660. For example, the personalised HRTF 660 may be synthesized based on the personalised parameters 650 and a basis set of component audio filters. This step may be performed separately from generating the personalised HRTF model, meaning that the personalised HRTF model can be stored efficiently, and any updates to the basis set of component audio filters can be easily applied to generate a new personalised HRTF for each user.

Steps S520 and S530 may be repeated for multiple hearing factors in order to further personalise the HRTF. This may be a step by step process, and has the advantage that the default HRTF, and each personalised HRTF generated in step S530, can be used by the user. As a result, the personalisation process can be spread out over time, and can be stopped whenever the user considers the personalisation to be adequate.

The above description assumes that a default HRTF model is already available in order to perform personalisation. There are known techniques for generating an HRTF model and such existing models may be used as the default HRTF model.

Fig. 7 schematically illustrates an example method of generating a default HRTF model, based on averaging data for a plurality of users (which may include or exclude the user for whom the personalised HRTF will be generated). This method may be performed by the same computer device which generates the personalised HRTF, or may be performed by a separate computer device For example, while the personalised HRTF may be generated by a personal user device, the default HRTF may be generated once by an organisation providing a personalised HRTF service to multiple users.

At step S710, for each of the plurality of users and each of a plurality of hearing factors, the hearing factor is obtained for the user. The hearing factors may include, for example, head measurement, interaural time delay, interaural level difference, and low-order spectral response properties. Levels and response properties of the users' ears may be measured with in-ear microphones. Additionally, levels and response properties of the users' ears may be measured based on conscious and/or physiological feedback from the users. However, some hearing factors are not suitable for inclusion in the default HRTF. For example, it is not meaningful to obtain an average shape of users' ears or other physical features and include corresponding average spectral peaks or notches.

At step S720, an average is calculated for each hearing factor, to generate an average hearing factor 411, 412, 461, 462.

At step S730, the default HRTF model is generated based on the hearing factors.

This may be implemented for each averageable hearing factor similarly to stages 620, 630 and 640 of Fig. 6A or stages 620, 635 and 645 of Fig. 6B, in order to generate a default HRTF model 410, 460 as previously described.

The default HRTF model may itself be specialised in various ways. For example, the default HRTF model may be calculated for a specific use case such as a specific type of room or a specific simulated environment (such as the environment in a specific video game). With the above-described techniques, different default HRTF models can be straightforwardly combined with different user-specific personalisations as required.

Claims (1)

1. CLAIMS1. A method for generating a Head-Related Transfer Function, HRTF, for an individual user, the method comprising: obtaining a predetermined default HRTF model; obtaining a hearing factor for a user; and generating a personalised HRTF model for the user by modifying the default HRTF model based on the obtained hearing factor 2. A method according to claim 1, wherein the default HRTF model comprises a first set of component audio filters and the personalised HRTF model comprises a second set of component audio filters.3. A method according to claim 2, wherein modifying the default HRTF model comprises at least one of: adding a component audio filter to the first set of component audio filters to generate the second set of component audio filters; removing a component audio filter from the first set of component audio filters to generate the second set of component audio filters; and performing signal processing on a component audio filter of the first set of component audio filters to generate the second set of component audio filters.4. A method according to claim 1, wherein the default HRTF model comprises a first set of parameters and the personalised HRTF model comprises a second set of parameters.5. A method according to claim 4, wherein modifying the default HRTF model comprises at least one of: adding a parameter to the first set of parameters to generate the second set of parameters; removing a parameter from the first set of parameters to generate the second set of parameters; and modifying a parameter of the first set of parameters to generate the second set of parameters.6. A method according to claim 4 or claim 5, further comprising synthesizing a personalised HRTF based on the second set of parameters and a basis set of component audio filters.7. A method according to any preceding claim, wherein the hearing factor comprises at least one of: an interaural time delay; an interaural level difference; a spectral peak or notch corresponding toa physical feature of the user; and a height compensation factor A method according to claim 7, comprising: obtaining a head measurement of the user; calculating the interaural time delay based on the head measurement.9. A method according to claim 7, comprising: obtaining an image of an ear, head and/or torso of the user; performing image processing on the image to identify the physical feature; calculating the spectral peak or notch based on the physical feature.10. A method of generating a default Head-Related Transfer Function, HRTF, model, the method comprising: for each of a plurality of users and each of a plurality of averageable hearing factors, obtaining the averageable hearing factor for the user; calculating an average of each of the averageable hearing factors for the plurality of users; and generating the default HRTF model based on the averageable hearing factors.11. A method according to claim 10, wherein the default HRTF model comprises a first set of component audio filters.12. A method according to claim 10, wherein the default HRTF model comprises a first set of parameters 13. A method according to claim 12, further comprising synthesizing a default HRTF based on the first set of parameters and a basis set of component audio filters 14. A method according to any of claims 10 to 12, wherein the averageable hearing factors include at least one of: an interaural time delay; an interaural level difference; and a low-order spectral response.15. A method according to any of claims 10 to 13, wherein the averageable hearing factors exclude at least one of: spectral peaks and notches corresponding to physical features; and a height compensation factor