Method for generating auricle-containing acoustical artificial head model on basis of statistical shape model

Abstract

Disclosed in the present invention is a method for generating an auricle-containing acoustical artificial head model on the basis of a statistical shape model. The method comprises the following steps: preprocessing a 3D graphics database of complete head surfaces of subjects, performing splitting to obtain an auricle-free head model dataset and an extended auricle model dataset, each of which has a checkable region; in a rigidly aligned coordinate system, using a statistical shape model to separately generate an auricle-free head average model and an extended auricle average model, which match each other, wherein a significant feature of the two models is having index information for the checkable regions; and assembling the auricle-free head average model and the extended auricle average model on the basis of the checkable regions to generate a complete external contour model of an auricle-containing acoustical artificial head model. The present invention uses the alignment of local physiological features of auricles, making the structure of the acquired extended auricle average model more accurate and representative; and high-order geometric features of the auricles and checkable regions nearby are retained, making an acoustical artificial head of the present invention more representative.

Description

A method for generating an acoustic artificial head model containing an auricle based on a statistical shape model Technical Field

[0001] The present application relates to the field of acoustic technology, and in particular to a method for generating an acoustic artificial head model containing an auricle based on a statistical shape model. Background Art

[0002] Humans perceive sound using both ears, and binaural sound pressure signals contain the primary information about the sound. From a physical perspective, sound waves emitted by a sound source are reflected and scattered by the human body before ultimately reaching the ears (Xie Bosun, Head-Related Transfer Function and Virtual Hearing [M]. National Defense Industry Press, 2008). Therefore, the shape of the human body has a significant impact on the sound information received.

[0003] In the field of acoustic technology, the sound pressure signals picked up by the two ears can be used to directly study binaural sound reproduction. The head-related transfer function (HRTF) can also be obtained through the binaural signals in the anechoic chamber and its binaural localization factors can be extracted for the study of virtual auditory reproduction. It has important application prospects in fields such as virtual reality (Xie B. Head-related transfer function and virtual auditory display [M]. J. Ross Publishing, 2013.). However, the experimental measurement process is cumbersome, and the influence of experimental conditions on data accuracy cannot be ignored (Li S, Peissig J. Measurement of head-related transfer functions: A review [J]. Applied Sciences, 2020, 10 (14): 5014.), so obtaining binaural sound signals through numerical simulation is also a successful application. Based on the 3D model, binaural sound signals can be easily obtained under different acoustic environments and different sound source positions (Katz BF G. Boundary element method calculation of individual head-related transfer function. I. Rigid model calculation [J]. The Journal of the Acoustical Society of America, 2001, 110 (5): 2440-2448.), which has broad application prospects. With the development of 3D imaging and 3D printing technology, it has become a research hotspot in the interdisciplinary field of acoustics.

[0004] Furthermore, based on a large number of 3D models of human subjects' heads, various methods have been used to generate universal artificial heads, greatly facilitating research in virtual auditory playback and binaural hearing. Numerous artificial heads exist, each with its own distinct uses and design methods. Table 1 lists some representative artificial heads.

[0005] Table 1 Existing artificial heads and their design methods

[0006] A literature review revealed that existing artificial heads have not published their design concepts and technical details. In particular, none of these artificial heads consider the coupling relationship between the auricle and the head surface (coupling state parameters, such as the auricle angle θ). Among existing artificial heads, the KEMAR artificial head has the auricle and head physiological structural characteristics closest to those of the patented solution. However, its official documentation only provides the dimensional parameters of the European physiological structure used as a reference, and its specific design and technical details are not disclosed. However, the head-related transfer function (HRTF), which is crucial for binaural hearing research (defined as the acoustic transmission process from a free-field point source to the two ears, which contains most of the sound source localization information and is a crucial data foundation for virtual auditory playback), particularly the auricle structural characteristics, dimensional parameters, auricle angle, and rotation angle, all exhibit significant individual differences, significantly impacting the HRTF. Therefore, in the application of artificial heads related to acoustics and hearing, preserving more representative auricle structural characteristics and morphological parameters of the coupling between the auricle and the head surface is of great scientific significance and application value.

[0007] Due to the complexity of the auricle's morphology, existing auricle models for artificial heads are generated by only designing a localized portion of the auricle using limited-dimensional data. However, a significant advantage of the present invention is that the method employed incorporates high-order geometric features of the auricle and the nearby head surface, allowing complete morphological parameters such as the auricle's structural characteristics and the auricle's angle of opening to be reflected in the model. In practical applications involving the use of acoustic artificial heads, retaining morphological parameters such as the auricle's angle of opening, which characterizes the coupling of the auricle to the head, is of great significance for the analysis of binaural sound signal characteristics and the study of binaural hearing localization factors. Therefore, the present invention can generate an artificial head model that is more suitable for acoustic research, and has both scientific significance and application prospects.

[0008] Summary of the Invention

[0009] To effectively preserve the more personalized auricle structure and the morphological parameters of the auricle coupled to the head surface, the present invention provides a method for generating an acoustic artificial head model using a statistical shape model. This method is suitable for designing acoustic artificial head models suitable for binaural hearing research. This method aims to preserve the complex local physiological structural characteristics of the auricle, as well as the auricle-head coupling relationship and ear morphological parameters such as the auricle angle in practical applications of acoustic artificial heads. This method is of great significance for the analysis of binaural sound signal characteristics and the study of binaural hearing localization factors. Given the difficulties of current artificial head models in quickly and accurately capturing the characteristics of the human population, a method for generating an acoustic artificial head model containing the auricle based on a statistical shape model is proposed.

[0010] The present invention is achieved through at least one of the following technical solutions.

[0011] A method for generating an acoustic artificial head model containing an auricle based on a statistical shape model comprises the following steps:

[0012] (1) Preprocess the complete 3D graphics database of the subject's head surface and then split it into a head model dataset without auricle and an extended auricle model dataset with a check area;

[0013] (2) In the rigidly aligned coordinate system, the matching average head model M without auricle is generated by the statistical shape model method. H and the extended auricle average model M P ,The salient features of the two groups of models are the index information of ,checked areas;

[0014] (3) Average model M of the head without auricle based on the selected area H and the extended auricle average model M P Splicing is performed to generate the external contour model M of the complete acoustic artificial head model including the auricle HP , used for the design and production of acoustic artificial head model entities.

[0015] Furthermore, in step (1), the number of subjects must meet the modeling requirements, and all complete head models must be marked and rigidly aligned so that all models are located at the same position in space with feature points defined by physiological structures, i.e., feature point markings, and non-rigid registration is performed to obtain a complete head dataset, thereby eliminating data errors caused by translation or rotation.

[0016] Furthermore, the preprocessing includes removing noise, filling missing data points, and performing plane section closing processing on the neck.

[0017] Furthermore, the auricle notch of the head model without auricle is the root of the auricle; the outer diameter of the selected area is the angled cone surface set in the direction of the two ear polar axes with the center of the head as the starting point, which produces a closed loop notch with the head surface near the auricle, and the closed loop notch is set as the radius R', which is larger than the maximum radius R of the auricle projected on the head surface along the two ear polar axes. max , that is, R′>R max to ensure that the intercepted extended auricle model has complete appearance features; the checked area is the area between the auricle root notch and the closed loop notch.

[0018] Furthermore, in step (2), the auricle-free head dataset and the extended auricle dataset that have been rigidly aligned and non-rigidly registered in step (1) are respectively used to establish a statistical shape model of the auricle-free head / extended auricle for the auricle-free head data and the extended auricle data through Gaussian process, which is used to describe and limit the reasonable probability distribution range of the complete contour of the model, and their mean function is statistically calculated to obtain the average model M of the auricle-free head. H and the extended auricle average model M P .

[0019] Furthermore, the splicing is based on the auricle-free head model dataset and the extended auricle model dataset obtained in step (1), and the index information between the two datasets is obtained by the check area, and the auricle-free head average model M is obtained. H and the extended auricle average model M P Perform vector superposition coupling to complete the splicing; in the splicing process, filter and smooth the edge of the model to generate the external contour model M of the complete acoustic artificial head model including the auricle HP .

[0020] Furthermore, in the splicing process, the average model M of the head without auricle is used. H The selected area is used as the reference, and the average model M of the head without pinna H The mesh density in the selected area should be equal to the expanded auricle sub-model M P The mesh density in the selected area.

[0021] Furthermore, the auricle average model M is extended P The grid size should not be larger than 1 / 6 of the wavelength of the audible upper frequency of 20kHz.

[0022] Furthermore, the outer contour model M of the acoustic artificial head model containing the auricle is produced by shell extraction and offset design. HP entity.

[0023] Furthermore, the outer contour model M of the acoustic artificial head model containing the auricle is made by injection molding, reverse molding, or 3D printing. HPentity.

[0024] Compared with the existing technology, the beneficial effects of the present invention are:

[0025] (1) The local physiological characteristics of the auricle are aligned so that the obtained extended auricle average model M P The structure is more accurate and representative;

[0026] (2) The high-order geometric features of the auricle and the surrounding selected areas are retained, so that the complete morphological parameters such as the auricle angle can be reflected in the model. In the analysis of binaural sound signal characteristics and the study of binaural hearing localization factors, the auricle structural features and auricle morphological parameters are of great significance. Since the present invention better retains the auricle structural features and auricle morphological parameters, the acoustic artificial head designed by the present invention is more representative and more practical than the traditional artificial head. The acoustic artificial head designed by the present invention has significant advantages over the traditional artificial head in terms of accuracy and representativeness. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] FIG1 is a flow chart of a method for generating an acoustic artificial head model including an auricle based on a statistical shape model according to an embodiment;

[0028] FIG2 is a partial cross-sectional schematic diagram of a complete head model according to an embodiment;

[0029] FIG3 is a partial cross-sectional diagram of a head model without auricles after only the auricles are removed according to an embodiment;

[0030] FIG4 is a cross-sectional schematic diagram of an expanded auricle model according to an embodiment;

[0031] FIG5 is a diagram showing the effect of separating the auricle using a conical closed-loop notch on the head contour according to an embodiment;

[0032] Figure 6 is a diagram showing the splicing effect of the embodiment;

[0033] FIG7 is a comparison diagram of the artificial head model obtained by the present invention and the artificial head obtained by the conventional method. DETAILED DESCRIPTION

[0034] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0035] The present invention is based on a pre-processed human head model (partial cross-sectional outline reference Figure 2) basic database, which is split into a head model without auricles (partial cross-sectional outline reference Figure 3) data set with a check area and an extended auricle model data set (reference Figure 4), and constructs an average model of the head without auricles (model M) by using a statistical shape model method. H ) and the extended auricle average model (model MP ), and finally, based on the selected area, the two average models are spliced ​​together to generate a complete artificial head model (model M HP ), which can be used to simulate the acoustic (sound wave) signals received by both ears (i.e., the sound wave signals propagating from a specific sound source to both ears), and can also obtain a complete 3D model of an acoustic artificial head after a design such as shell extraction, and can produce its solid model by various methods, including but not limited to injection molding, 3D printing and other technologies. The implementation process is shown in Figure 1, which mainly includes: step (1), pre-processing the complete head surface 3D graphic database of a certain number of subjects, and then splitting it into a dataset of head models without auricles and a dataset of extended auricle models with check areas; step (2), in a rigidly aligned coordinate system, generating the matching average models M of the head without auricles by statistical shape modeling. H and the extended auricle average model M P The significant feature of the two groups of models is the index information of the selected area; Step (3) is to select the model M based on the selected area. H and Model M P Splice and generate a complete acoustic artificial head model with auricle outer contour model M HP , which can be used to design and produce an acoustic artificial head model entity. The present invention aims to design an acoustic artificial head model suitable for binaural hearing research. In particular, the method of the present invention can effectively preserve more representative auricle contour features and the relative positional relationship between the auricle and the head surface.

[0036] As shown in FIG1 , this embodiment is the complete process flow for generating an acoustic artificial head model including an auricle, and is also applicable to the method for generating an acoustic artificial head model including an auricle based on a statistical shape model described in this embodiment. The specific steps include the following:

[0037] Step 1: Establish a basic database, perform preprocessing, and generate a head model dataset without auricle and an extended auricle model dataset respectively.

[0038] Step 1.1: Use a 3D scanner or other imaging device to collect 3D head scan data from multiple individuals. Preprocess the data, including removing noise, filling missing data points, and performing planar cross-section closure of the neck, to establish a basic database. All data collection requires prior consent from the subjects and compliance with ethical review requirements.

[0039] In particular, the size of the database should generally meet the following statistical requirements:

[0040] Assuming that a basic database has been obtained and the database sample capacity meets the ISO 15535:2007 standard to ensure the reliability of the results, the limiting formula is as follows:

[0041] Among them, 1.96 is the critical value (Z value) of the 95% confidence interval of a standard normal distribution; CV is the coefficient of variation of the mean; SD is the standard deviation of the human body size under investigation; a is the relative accuracy that the investigation aims to obtain. The data with the largest head length is used here to obtain the maximum sample minimum capacity N. It should be noted that this preprocessing needs to be performed on the model that has been marked, aligned, and registered.

[0042] Step 1.2: Create registration template and split template

[0043] Step 1.2.1. Create a registration template. Sample a representative human head model as the initial template shape and fully mark its physiological features. The complete physiological feature marking points must include all detailed features such as the face, neck, head outline, and auricle, generating a registration template that corresponds to the physiological features.

[0044] Step 1.2.2, split the template. Divide the registration template obtained in step 1.2.1 into three areas, namely, the auricle area, the check area and the rest of the head area. The auricle notch of the head model without auricle is the root of the auricle (see Figure 3); the outer diameter of the check area is a cone surface with a specific angle set in the direction of the two ear polar axes with the center of the head as the starting point, and a closed loop notch (radius R') is generated with the head surface near the auricle. The radius is larger than the maximum radius R of the auricle projected on the head surface along the two ear polar axes. max , that is, R′>R max To ensure that the intercepted extended auricle model has complete shape features; the selected area is the area between the auricle root notch and the closed loop notch (refer to Figures 3 and 4);

[0045] As a specific embodiment, a human Frankfurt surface is established based on the left and right tragus points and the right orbital inferior margin point, and a coordinate system is established by the symmetric surface of the human head and the line connecting the ear canal opening; then, with the line connecting the ear canal opening as the central axis and the origin of the coordinate axis as the vertex, a bilaterally symmetrical conical surface is created with a 30° inclination angle (the auricle is basically within a 20° inclination range of the polar axis of both ears starting from the center of the head), thereby dividing the outer edge of the selected area (with the area away from the central axis as the outer edge), and taking the auricle root notch as the inner edge of the selected area, thereby completing the definition of the selected area, that is, the area between the auricle root notch and the closed loop notch. The auricle area is defined as the auricle model within the auricle root notch, and the rest of the head area is the area of ​​the complete head model excluding the auricle area, that is, the closed loop notch. The union of the auricle area and the checked area is the extended auricle area, and the union of the remaining head areas and the checked area is the head area without auricle; the corresponding template parts are the extended auricle model and the head model without auricle, and the corresponding point cloud index is the data form of the point cloud area division.

[0046] Step 1.3: Label, align, and register all models in the database

[0047] Step 1.3.1. Marking. Combine anatomical knowledge to mark some key physiological features of all human head models. Compared to the complete marking in step 1.2.1, only some key feature points need to be marked here. Key features include the feature point of the ear canal opening, the left and right tragus points, and the right orbital margin point.

[0048] Step 1.3.2, Alignment. Establish the Frankfurt surface of the human body based on the left and right tragus points and the right orbital margin point, and establish a coordinate system based on the symmetry plane of the human head and the line connecting the ear canal opening. Perform rigid alignment on the models so that all head models are located in the same position in space based on the physiological structure definition to eliminate data errors caused by translation or rotation.

[0049] Step 1.3.3: Registration. The obtained physiological structure feature points and the non-rigid iterative closest point (NICP) registration algorithm are used to register the head model surface data based on the template to obtain a point distribution model with corresponding relationships, thereby generating a complete model dataset.

[0050] Step 1.4: Split the dataset based on the template, create a head dataset without auricle and a dataset with extended auricle, and define the check area.

[0051] Step 1.4.1: Establish a dataset of heads without pinnae. Based on the index of the head model without pinnae obtained in step 1.2.2, extract the dataset of head models without pinnae from the complete head model set that has been aligned and registered in step 1.3.

[0052] Step 1.4.2: Based on the extended auricle model index obtained in step 1.2.2, the extended auricle model dataset is extracted from the complete head model that has been aligned and registered in step 1.3.

[0053] As an example, the mesh size of the extended auricle model should be smaller than 2.8mm, one-sixth the minimum audible sound wavelength, to fully describe the extended auricle structure. The non-extended auricle model (the portion outside the extended auricle model) should include essential features of the human head while omitting non-critical structures. See Figure 3 for a schematic diagram of the relevant processing results.

[0054] Step 2: Build a statistical shape model and obtain the average model

[0055] Step 2.1, using the aligned and registered auricle-free head dataset and the extended auricle dataset obtained in step 1, calculate the distribution and reference shape of each point. For the dataset, the spatial three-dimensional coordinates of the human body surface point can be represented by the vector x i,j To express, for the i-th subject, the body surface shape Γ i The human body surface shape vector S i (A set of vectors x on physiological structure i,j ∈R 3 ) to discretely represent:

[0056] In the formula The superscripts x, y, and z represent the corresponding spatial coordinates, respectively. The subscript i represents the subject number, j represents the point number, and p represents the total number of points in the template. 3 The dimension of the representative vector is 3, Represents the x, y, z coordinates of the last point of the model.

[0057] Based on the human body surface shape vector S of each subject sample shown in formula (4) i , the data sets can be integrated.

[0058] In a statistical sense, the shape category is usually modeled by the multivariate normal distribution N(μ,Σ), which is determined by the assumption that when there are enough samples, any variable of the shape vector S obeys the normal distribution. The average human surface shape vector S of all N subjects in the auricle-free head model dataset or the extended auricle model dataset can be averaged to calculate the average human surface shape vector μ and its covariance matrix Σ, and S~N(μ,Σ) is the point distribution model, which describes the distribution state of the model points. Based on this, it can be considered that the average human surface shape vector μ represents the reference shape Γ R :

[0059] Where N represents the total number of subjects.

[0060] Step 2.2: Establish a Gaussian process, using the vector S in step 2.1 i , we can define Γ R A set of discrete functions s on i (x k ), function s i (x k ) represents the S of the i-th subject i The kth value of the vector, s(x), represents the distance from the reference shape Γ RThe mapping relationship between any variable x on the shape Γ and the corresponding variable x on another shape Γ. The Gaussian process shown in Equation (7) is uniquely defined by the mean function μ(x) described in Equation (8) and the covariance function Σ(x,x′) described in Equation (9):

[0061] Where x k represents Γ on the reference shape R The kth variable on the template is any spatial coordinate of any point on the template. m represents the total number of variables in the template, and its value is equal to the total number of points in the template multiplied by 3, that is, m = 3p. GP is the abbreviation of Gaussian process, which can also be modified to N to represent normal distribution. The meaning of x is x k The continuous form of Γ R Any variable at any point on the , x ′ represents a point different from x, s i (x) represents the x from the reference shape Γ R The shape mapped to the i-th subject, s i (x′) represents the distance from the reference shape Γ to the R The shape mapped to the i-th subject.

[0062] Performing Karhunen-Loève expansion on the covariance function Σ(x,x′) yields a set of orthogonal basis functions φ i (x) and the corresponding eigenvalue λ i , by introducing the coefficient α i , s(x) can be described as follows:

[0063] Similar to principal component analysis, retain its low-order approximation as follows:

[0064] Where r represents the order required for the low-order approximation of s(x).

[0065] pass Method, the above basis function φ can be calculated i Approximate value of (x) Finally, the shape is represented as:

[0066] Where p represents the total number of points in the template, Represents the spatial coordinate values ​​of x, y, and z after the p-th point on the template is mapped to the mean function μ(x). Represents the approximate value of the mapping of the p-th point on the template to the above basis function The spatial coordinate values ​​of x, y, and z are represented above, S represents the shape vector of any human body surface, Q is the feature matrix, and α represents the coefficient vector corresponding to the feature matrix.

[0067] Formula (15) means decomposing any shape into average shapes and a set of characteristic shape variation modes u i The linear combination of , any shape can be represented in the form of a coefficient vector. This method is applied to the complete head / extended auricle model dataset to obtain the average model M of the head without auricle. H and the extended auricle average model M P .

[0068] Generate the matching average model M of the head without pinna H and the extended auricle average model M P , including but not limited to using any subset of the no-pinna head model dataset and the extended pinna model dataset to obtain the no-pinna head average model M in an equivalent manner H and the extended auricle average model M P Used for the design and production of acoustic artificial head model entities.

[0069] Step 3: Average model M of the head without auricle based on the selected area H and the extended auricle average model M P Splice and generate a complete acoustic artificial head model with auricle outer contour model M HP .

[0070] Step 3.1: Average model M of the head without pinna H and the extended auricle average model M P The point cloud data of the selected area are recorded as P H and P P , taking the average model M of the head without auricle H As a benchmark, the transformation matrix T is calculated based on the minimum mean square error, and the transformation matrix is ​​used to implement the model M H and Model M P overlap.

[0071] where M′ P represents the transformed extended auricle average model M P ;

[0072] Step 3.2: Design the filter matrix W to select the area point cloud data P H Linear filtering is performed on the edge of the template to smooth the template.

[0073] Where ω i Represents the weight coefficient, which is used to perform weighted superposition on the matrix.

[0074] Step 3.3: Perform vector superposition operation on the model, based on the transformed extended auricle average model M′ P Superimpose the average model of the head without pinna M H , then add the check area WP for filtering H , realize the average model M of the head without pinna H With average extended pinna model M P The target shape M is obtained by coupling HP Expressed as: M HP =WM H +M′ P +M H (18)

[0075] Target shape M HP That is, the outer contour of the acoustic artificial head model including the pinna, the filtering and smoothing effects are shown in Figure 6.

[0076] Generate the external contour model M of the acoustic artificial head model including the pinna based on the statistical shape model HP Used for the design and production of acoustic artificial head model entities, including but not limited to designing hollow or dense models by methods such as shell extraction, offset, injection molding, mold casting, 3D printing, etc. for entity production.

[0077] Figures c and d in Figure 7 are KEMAR artificial heads produced abroad, which are widely used in fields such as acoustic measurement. The model is designed based on the average physiological parameters of the Western human body. The difference between the present invention and the KEMAR artificial head (Figures a and b) is that the method for obtaining the appearance of the artificial head and the population for obtaining the appearance are fundamentally different. The comparison in Figure 7 is only in appearance, using the isometric view (Figures a and c) and the rear view (Figures b and d) as examples.

[0078] The above is a specific implementation of the present invention. By establishing a template with a corresponding relationship with physiological characteristics, constructing a statistical shape model, and generating a complete outer contour of an acoustic artificial head model containing the auricle, etc., an acoustic artificial head model containing the auricle is generated. While retaining key physiological parameters such as the auricle angle, it solves the difficulties in formulating the auricle model of the traditional artificial head model, bringing substantial progress and application value to acoustic measurement and related research fields.

Claims

1. A method for generating an acoustic artificial head model containing an auricle based on a statistical shape model, characterized in that: The following steps are involved: (1) Preprocessing the complete head surface 3D graphic database of the subject, and then splitting it into a head model dataset without auricle and an extended auricle model dataset with a selected area; (2) In a rigidly aligned coordinate system, the matching average head model M without auricle is generated by statistical shape modeling. H and the extended auricle average model M P ,The salient feature of the two groups of models is the index information of the ,checked regions; (3) Average model M of the head without auricle based on the selected area H and the extended auricle average model M P Splicing is performed to generate the external contour model M of the complete acoustic artificial head model including the auricle HP , used for the design and production of acoustic artificial head model entity.

2. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: In step (1), the number of subjects must meet the modeling requirements, and all complete head models must be marked and rigidly aligned so that all models are located at the same position in space with feature points defined by physiological structure, i.e., feature point markings, and non-rigid registration is performed to obtain a complete head dataset, thereby eliminating data errors caused by translation or rotation.

3. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: The preprocessing includes removing noise, filling missing data points, and performing plane section closing processing on the neck.

4. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: The auricular notch of the head model without auricle is the root of the auricle; the outer diameter of the selected area is the angled cone surface set in the direction of the two ear polar axes with the center of the head as the starting point, and a closed loop notch is generated with the head surface near the auricle. The closed loop notch is set as the radius R', and the radius R' is greater than the maximum radius R of the auricle projected on the head surface along the two ear polar axes. max , that is, R′>R max to ensure that the intercepted extended auricle model has complete appearance features; the checked area is the area between the auricle root notch and the closed loop notch.

5. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: In step (2), the auricle-free head dataset and the extended auricle dataset that have been rigidly aligned and non-rigidly registered in step (1) are respectively used to establish a statistical shape model of the auricle-free head / extended auricle through a Gaussian process for the auricle-free head data and the extended auricle data, which is used to describe and limit the reasonable probability distribution range of the complete contour of the model, and their mean functions are statistically calculated to obtain the average auricle-free head model M H and the extended auricle average model M P .

6. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: The splicing is based on the auricle-free head model dataset and the extended auricle model dataset obtained by preprocessing in step (1), using the index information between the two datasets provided by the check area to generate the auricle-free head average model M. H and the extended auricle average model M P Perform vector superposition coupling to complete the splicing; during the splicing process, filter and smooth the edge of the model to generate an external contour model M of the complete acoustic artificial head model including the auricle HP .

7. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: During the splicing process, the average model M of the head without auricle is used. H The selected area is used as the reference, and the average model M of the head without auricle H The mesh density in the selected area should be equal to the expanded auricle sub-model M P The mesh density in the selected area.

8. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: Extended auricle average model M P The grid size should not be larger than 1 / 6 of the wavelength of the audible upper frequency limit 20 kHz.

9. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: The outer contour model M of the acoustic artificial head model with a hollow or dense auricle is produced by shelling and offset design. HP entity.

10. The method for generating an acoustic artificial head model containing an auricle based on a statistical shape model according to claim 1, characterized in that: The outer contour model M of the hollow or dense acoustic artificial head model containing the auricle is made by injection molding, reverse molding, and 3D printing methods HP entity.

Images