Vehicle sound area calibration method, vehicle, storage medium and program product

Abstract

The invention provides a vehicle sound area calibration method, a vehicle, a storage medium and a program product. The interior of a cabin of the vehicle is divided into a plurality of sound areas to be calibrated; the method comprises the following steps: acquiring a sound source scanning grid and a calibration audio sample set for the vehicle; traversing a plurality of different scanning point combinations, and executing the following operations on each traversed scanning point combination: according to the positioning spectrum of each calibration audio sample, determining the maximum value of the matching degree of each calibration audio sample on each scanning point in the scanning point combination, the sound area corresponding to the maximum value of the matching degree is used as a predicted sound area of the corresponding calibration audio sample; determining the sound area positioning accuracy of the scanning point combination according to the consistency of the predicted sound area of each calibration audio sample and the corresponding real sound area; and selecting the scanning point combination with the highest sound area positioning accuracy from the traversed scanning point combinations as a calibration result.

Description

Technical Field

[0001] This specification relates to the field of vehicle technology, and in particular to a vehicle audio zone calibration method, a vehicle, a storage medium, and a program product. Background Technology

[0002] With the development of automotive intelligence, the vehicle cabin has become a "third living space" integrating work and entertainment. To ensure the independence and privacy of human-machine interaction for different passengers, the industry has used sound zone localization and separation technology to divide the physical space of the cabin into multiple independent sound zones, enabling the system to accurately identify voice commands in each sound zone and provide personalized audio services. Therefore, how to efficiently and accurately complete sound zone calibration to meet the rapid adaptation needs of different vehicle models has become a key challenge for the industry.

[0003] In related technologies, sound zone separation typically relies on deep neural networks. However, due to significant differences in acoustic parameters among different vehicle models in terms of size, microphone layout, and interior materials, each model requires the separate collection of a large amount of data and retraining of the model. This calibration process is time-consuming, usually taking about two months, which seriously affects the efficiency and cost of feature deployment. Summary of the Invention

[0004] In view of this, this specification provides a vehicle audio zone calibration system, method, vehicle, storage medium, and program product to address the deficiencies in the related art.

[0005] Specifically, this specification is implemented through the following technical solution: According to a first aspect of this specification, a method for vehicle acoustic zone calibration is provided, wherein the interior of the vehicle is divided into multiple acoustic zones to be calibrated; the method includes: Acquire a sound source scanning grid and a calibration audio sample set for the vehicle. The scanning points of the sound source scanning grid cover the spatial orientation inside the cabin. Each calibration audio sample in the calibration audio sample set is associated with the real sound region where it was collected. For each calibrated audio sample, the localization spectrum of the audio signal it contains is calculated. The localization spectrum is used to characterize the degree of matching between the audio signal and the spatial orientation corresponding to each scanning point. The algorithm iterates through multiple different combinations of scan points, where each combination consists of scan points selected from the sound source scanning grid for each calibrated audio region. For each traversed combination of scan points, the following operations are performed: based on the localization spectrum of each calibrated audio sample, the algorithm determines the maximum matching degree of each calibrated audio sample at each scan point in the combination, and uses the audio region corresponding to the maximum matching degree as the predicted audio region of the corresponding calibrated audio sample; based on the consistency between the predicted audio region of each calibrated audio sample and the corresponding real audio region, the algorithm determines the localization accuracy of the scan point combination. The combination of scanning points with the highest accuracy in locating the vocal range is selected from the traversed combinations of scanning points as the calibration result, and each scanning point in the calibration result is used as the target scanning point for locating the corresponding vocal range to be calibrated.

[0006] According to a second aspect of this specification, a vehicle acoustic zone localization method is provided, wherein the interior of the vehicle is divided into multiple acoustic zones, and each acoustic zone is configured with at least one target scanning point obtained by the acoustic zone calibration method as described in any one of the first aspects; the method includes: Acquire a multi-channel audio signal composed of multiple microphones of the vehicle, wherein the multi-channel audio signal contains the target voice signal to be located; The corresponding localization spectrum is calculated based on the multi-channel audio signal. The localization spectrum is used to characterize the degree of matching between the target speech signal and the spatial orientation corresponding to each target scanning point. The target voice region to which the target speech signal belongs is determined based on the target scan point corresponding to the highest matching degree. According to a third aspect of this specification, a vehicle is provided, comprising: a processor and a memory for storing processor-executable instructions; wherein the processor executes the executable instructions to implement the steps of the method as described in either the first or second aspect.

[0007] According to a fourth aspect of this specification, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method as described in either the first or second aspect.

[0008] According to a fifth aspect of this specification, a computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of the methods described in either the first or second aspect.

[0009] This specification describes a calibration method that constructs a sound source scanning grid covering the cabin and collects calibration samples from known sound zones. Based on a physical acoustic model, it calculates the matching spectrum between the samples and each grid point, and then searches for the optimal combination of scanning points that maximizes positioning accuracy. This method eliminates the reliance on deep neural network models and massive amounts of training data, eliminating the need for repeated model training for different vehicle models. Calibration can be completed with only a small number of samples and a limited number of optimization searches. Compared to traditional model training and calibration methods, this method significantly shortens the calibration time and allows for rapid adaptation and deployment by configuring corresponding configuration files for different vehicle models. It also improves the efficiency of subsequent environment configuration and reduces development and maintenance costs. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in this specification, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of this specification, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0011] Figure 1 This is a schematic flowchart illustrating an exemplary embodiment of a vehicle sound zone calibration method in this specification; Figure 2 This is a schematic diagram of the coordinates of a vehicle-mounted microphone shown in an exemplary embodiment of this specification; Figure 3 This is a schematic diagram illustrating a dense scanning grid as shown in an exemplary embodiment of this specification; Figures 4a-4f This is a schematic diagram of a target scan point in a dense scan grid, as illustrated in an exemplary embodiment of this specification; Figure 5 This is a flowchart illustrating an exemplary embodiment of a vehicle sound zone localization method as shown in this specification; Figure 6 This is a schematic diagram of the structure of a device shown in an exemplary embodiment of this specification; Figure 7 This is a schematic diagram of the structure of a vehicle sound zone calibration device shown in an exemplary embodiment of this specification; Figure 8 This is a schematic diagram of the structure of a vehicle sound zone positioning device shown in an exemplary embodiment of this specification. Detailed Implementation

[0012] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this specification.

[0013] The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of this specification. The singular forms “a,” “the,” and “the” as used in this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0014] It should be understood that although the terms first, second, third, etc., may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this specification, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0015] The embodiments of the vehicle sound zone calibration method in this specification will be described in detail below with reference to the accompanying drawings.

[0016] Figure 1 This is a schematic flowchart illustrating an exemplary embodiment of a vehicle sound zone calibration method described in this specification. Figure 1 As shown, the interior of the vehicle's cabin is divided into multiple pitch-fixing zones; the method includes: Step S102: Obtain the sound source scanning grid and calibration audio sample set for the vehicle. The scanning points of the sound source scanning grid cover the spatial orientation inside the cabin. Each calibration audio sample in the calibration audio sample set is associated with the real sound region where it was collected.

[0017] The solution described in this specification first obtains two types of basic data for calibration: a sound source scanning grid and a calibration audio sample set. The sound source scanning grid covers the spatial orientation of the vehicle cabin through a series of scanning points, with each scanning point corresponding to a preset spatial orientation. The calibration audio sample set contains audio samples collected in actual sound zones, and each sample is labeled with the actual sound zone information at the time of collection, providing a reference label for subsequent sound zone matching. Those skilled in the art will understand that a sound zone is usually associated with a vehicle seat, such as the driver's seat, front passenger seat, rear left seat, and rear right seat. Each sound zone represents an independent voice interaction space, and the aforementioned calibration audio samples are recorded within each such sound zone, with their source sound zone clearly labeled.

[0018] The construction process for obtaining the aforementioned sound source scanning grid can be based on the specific acoustic geometry of the vehicle cabin. Specifically, a geometric model of the cabin's interior can first be obtained, reflecting the cabin's basic spatial structure and dimensions. Based on this geometric model, dense scanning points are generated within the target area at preset spatial intervals, thereby forming a sound source scanning grid covering the cabin's interior space.

[0019] Each scanning point in the sound source scanning grid is not simply a geometric point, but is associated with a key acoustic parameter: the steering vector. Assuming a sound source is located in the spatial orientation corresponding to that scanning point, the steering vector of that point can be understood as the theoretical phase difference relationship between the sound waves emitted by that source and the multiple microphones configured in the vehicle. This relationship forms the basis for subsequent sound source orientation matching calculations. It should be noted that these multiple microphones can typically be positioned in fixed locations within the cabin, such as on the roof, to collect mixed speech signals from all audio ranges, without requiring each audio range to have its own dedicated microphone.

[0020] Specifically, the aforementioned steering vector can be mathematically represented as an M-dimensional complex vector, where M can be the number of microphones in the vehicle. The m-th element of this vector... This can be considered as the calculation result for the k-th frequency band, and its expression is shown in Formula 1: in, The imaginary unit, The signal sampling rate, This represents the total number of frequency bands used in the Fast Fourier Transform. The core parameter in the formula... This is the theoretical time difference between the arrival of a sound wave at the m-th microphone and its arrival at a preset phase center. The aforementioned phase center is a reference point used for calculation, and its location can be set according to convenience, such as the geometric center of a vehicle cabin, the geometric center of the microphone array, or the origin of the sound source scanning grid. Furthermore, this phase center can typically be set as the origin of the sound source scanning grid. Therefore, as mentioned earlier, the complex exponential function essentially encodes the theoretical phase difference generated when a sound wave arrives at different microphones from a specific direction.

[0021] In addition, the aforementioned theoretical time difference It can be determined by the physical path of sound wave propagation, and the calculation formula is as shown in Formula 2 above: in, For the speed of sound, Let be the straight-line distance from the m-th microphone to the phase center. Where is the distance... Based on the microphone's three-dimensional coordinates in the cockpit coordinate system The following formula 3 is used to calculate: It's worth noting that the microphone's coordinates can be obtained through simple physical measurements of the actual vehicle, such as by the vehicle manufacturer's sound zone calibration personnel. Figure 2The lengths of the line segments formed by the different microphones A, B, C, and D (AB, AC, CD, etc.) and scanning points p1, p2, p3, and p4 shown in the diagram can be obtained, or they can be directly obtained based on an existing 3D digital model of the cockpit. This specification does not impose any restrictions on the comparison. Thus, by combining formulas 1-3, the corresponding steering vector for each scanning point in the sound source scanning grid can be pre-calculated, providing a complete theoretical basis for subsequent sound source orientation matching.

[0022] In summary, based on the aforementioned geometric relationship between the microphone and the cockpit, and the formula for calculating the steering vector, a sound source scanning grid for zone calibration can be constructed. As mentioned earlier, this grid can consist of a large number of scanning points evenly arranged at preset spatial intervals of, for example, 5 centimeters, thereby densely covering the possible speech emission areas of each zone within the cockpit.

[0023] In one embodiment, each scan point can correspond to a unique scan direction index. It also possesses a pre-calculated steering vector based on Formula 1. .like Figure 3 As shown, points can be placed at fixed intervals along the front-back and left-right directions of the vehicle to form a two-dimensional planar grid. If 30 points are placed in the front-back direction and 42 points are placed in the left-right direction, then the total number of scanning points is... This means that the system has prepared 1260 potential spatial orientation templates for calibration, among which, Figure 3 In the diagram, x represents the phase center, ○ represents the microphone, and □ represents the scanning orientation.

[0024] The establishment of the above dense grid can discretize the continuous physical space into a series of candidate points, so that the subsequent pitch range calibration process is transformed into a data-driven optimization problem of searching for and selecting the optimal representative point for each pitch range among these candidate points.

[0025] Furthermore, the execution entity for each step of the vehicle audio zone calibration method described in this specification can be flexibly configured according to the actual product form and implementation scenario. The entire process of the above method can be completed independently by a single entity, such as being executed entirely locally in the vehicle by the onboard computing unit, or entirely outside the vehicle by an independent cloud server or dedicated calibration equipment. The above method can also be jointly and collaboratively executed by the vehicle and the cloud, or other external computing units. For example, the vehicle can be responsible for data collection and preprocessing, while the cloud can be responsible for time-consuming dense grid calculations and optimization searches. Finally, the calibration results are sent to the vehicle in the form of, for example, a target scan point configuration file, thus facilitating adaptation to different system architectures and resource constraints.

[0026] Step S104: For each calibrated audio sample, calculate the localization spectrum of the audio signal it contains. The localization spectrum is used to characterize the degree of matching between the audio signal and the spatial orientation corresponding to each scanning point.

[0027] For each calibrated audio sample acquired, its audio signal localization spectrum can be calculated. This localization spectrum characterizes the degree of matching between the audio signal and the spatial orientation corresponding to each scanning point in the sound source scanning grid. Essentially, it extracts the feature distribution of the above samples in the spatial dimension through signal processing methods, providing data basis for subsequent sound region determination.

[0028] For each calibrated audio sample, the above-mentioned localization spectrum calculation process can be implemented based on the previously obtained steering vector.

[0029] In one embodiment, the sound zone calibration system in this specification can calculate the covariance matrix between signals collected by multiple microphones based on the audio signal of the calibration audio sample, and calculate the matching degree between the audio signal and each scanning point based on the covariance matrix and the steering vector corresponding to each scanning point in the sound source scanning grid. The matching degrees corresponding to each scanning point then constitute the localization spectrum. Specifically, the covariance matrix statistically characterizes the spatial phase relationship generated when the audio signal of the calibration audio sample arrives at different microphones, serving as the mathematical basis for subsequent azimuth matching. Furthermore, in the process of calculating the matching degree value between the audio signal and each scanning point in the grid, it essentially compares the spatial characteristics of the actual collected signal, i.e., the covariance matrix, with the theoretical spatial template corresponding to each scanning point, i.e., the steering vector, thereby quantifying the probability that the audio signal originates from the azimuth of each scanning point. Finally, by arranging the matching values ​​corresponding to all scan points in order, a complete localization spectrum is formed. This localization spectrum can be used as a vector or graph, and its dimension is equal to the total number of scan points. The value of each dimension intuitively reflects the degree of matching between the audio signal of the current sample and the corresponding spatial orientation, providing direct data basis for subsequent scan point combination search and optimization.

[0030] Based on the above steps, the aforementioned localization spectrum can be achieved through specific mathematical operations. First, assume that a total of [data / samples] have been collected in each register. A set of calibration samples, sample index used This indicates, and the samples collected at that time The true vocal range index is So for the first A calibrated audio sample, in the frequency band The covariance matrix on is denoted as Combining the sound source scanning grid with the first The guiding vector corresponding to each scan point The matching degree between the sample and the scan point can be calculated using the following formula 4. : in, to For the frequency band range used for the samples, This represents the conjugate transpose of the steering vector. The essence of this operation is to represent the actual spatial features of the samples across multiple frequency bands. Projected onto scan point Theoretical orientation template The sum of these sums represents the sum of the values ​​of the sample and the scan point. The matching degree reflects the likelihood that the sample signal originates from that location.

[0031] Step S106: Traverse multiple different combinations of scanning points, wherein each combination of scanning points consists of scanning points selected from the sound source scanning grid for each calibrated audio region; and perform the following operations on each traversed combination of scanning points: determine the maximum matching degree of each calibrated audio sample on each scanning point in the combination of scanning points according to the localization spectrum of each calibrated audio sample, and take the audio region corresponding to the maximum matching degree as the predicted audio region of the corresponding calibrated audio sample; determine the localization accuracy of the combination of scanning points according to the consistency between the predicted audio region of each calibrated audio sample and the corresponding real audio region.

[0032] By iterating through all By calculating each scan point according to Formula 4 above, the corresponding value for the sample can be obtained. 3D localization spectral vector This vector comprehensively represents the matching degree distribution between the audio signal of the currently calibrated audio sample and each spatial orientation in the sound source scanning grid, providing core data input for the statistically optimal scan point combination search in this step. Specifically, based on the localization spectrum of each sample, the maximum matching degree of the sample at each scan point in the combination can be determined, and the pitch region corresponding to the maximum value is taken as the predicted pitch region of the sample. Subsequently, the consistency between the predicted pitch regions and the actual pitch regions of all samples is statistically analyzed, thereby obtaining the pitch region localization accuracy corresponding to the scan point combination. This process is essentially an optimization process for searching the optimal scan point combination.

[0033] In other words, after obtaining the localization spectral vectors of all calibrated samples, the process can proceed to the traversal search of scan point combinations and the evaluation stage of each subsequent step. This allows for the selection of the most representative scan point for each calibrated pitch region from a vast pool of potential spatial orientations, resulting in the highest overall accuracy for pitch region determination based on this set of points.

[0034] Regarding the definition and traversal process of scan point combinations, a scan point combination refers to a set of points formed by selecting one scan point for each tone region to be calibrated from the sound source scan grid. Assume the vehicle has... A region of pitch to be calibrated, for example Corresponding to the driver's seat, front passenger seat, rear left seat, and rear right seat, and the total number of points in the scanned grid is [number missing]. Then a combination can be represented as Each of them Representative assigned to the first For each vocal range's scan point index, the system needs to traverse a large number of such different combinations for evaluation.

[0035] When constructing the above combination of scan points, prior knowledge based on the physical layout of the vehicle compartment can be introduced as a constraint to further improve search efficiency, narrow the search space, and accelerate the optimization process. Specifically, the construction process of the above combination of scan points can be subject to at least one of the following constraints: 1. Relevance Constraint: Based on the symmetrical layout of the multiple to-be-calibrated sound zones in the vehicle cabin, the process of selecting scanning points for symmetrically arranged sound zones is correlated. For example, in a common left-right symmetrical layout, this constraint can be specifically implemented as follows: the scanning points selected for left-right symmetrically arranged sound zones, such as the driver's seat and passenger seat, are symmetrical in the horizontal direction with respect to the longitudinal center plane of the vehicle.

[0036] 2. Sequence Constraint: Based on the positional order of the multiple to-be-calibrated sound zones in the front-rear direction of the vehicle, the coordinate relationship of the scan points selected for different sound zones in the front-rear direction of the vehicle is restricted. For example, this constraint can be specifically implemented as follows: the coordinates of the scan points selected for the rear sound zone in the front-rear direction of the vehicle do not exceed, that is, the corresponding coordinate values ​​are not greater than the corresponding coordinates of the scan points selected for the front sound zone on the same side.

[0037] By introducing such reasonable physical constraints, the number of invalid or physically unreasonable combinations that need to be traversed can be effectively reduced without sacrificing calibration accuracy, thereby significantly improving the computational efficiency of the calibration process.

[0038] Step S108: Select the combination of scanning points with the highest accuracy in locating the sound region from the traversed combinations of scanning points as the calibration result. Each scanning point in the calibration result is used as the target scanning point for locating the corresponding sound region to be calibrated.

[0039] From all the combinations of scan points traversed, the combination with the highest accuracy in register localization can be selected as the final calibration result, which can be called a combination evaluation process. Each scan point in this result is determined as the target scan point for the corresponding register to be calibrated, and is used as the direction matching reference in the subsequent actual register localization process in this specification, thereby achieving efficient and accurate register localization.

[0040] Specifically, for each combination of scan points traversed, the system can perform the following operations to evaluate its performance: 1. Sample prediction: For the first sample in the calibration sample set... Each sample, based on its localization spectral vector Considering only the four scan points in this combination Corresponding matching degree value Find the maximum value among them; the pitch range number corresponding to this maximum value is determined as the predicted pitch range for that sample. .

[0041] 2. Correctness judgment: The predicted tone range is... Compared with the sample's previously known true vocal range Compare. If If the sound register is correctly located, the sample is considered to have been correctly located; otherwise, it is considered incorrect.

[0042] 3. Accuracy calculation: For all samples in the sample set... Repeat steps 1 and 2 for each sample, count the number of samples that are correctly located, and then calculate the accuracy of the sound region localization of the scan point combination on the entire sample set, that is, the proportion of the number of correct samples to the total number of samples.

[0043] In summary, by traversing a massive number of candidate combinations and calculating the accuracy of each region's localization, the system can ultimately select the combination of scanning points with the highest accuracy, denoted as . This combination represents the optimal calibration result obtained in the first round of optimization search, with each scan point in the combination being determined as the target scan point for its corresponding vocal range. For example... Figure 4a As shown, the solid black markers represent the optimal scanning positions selected for each vocal register. This process transforms the vocal register calibration problem into a data-driven, quantifiable combinatorial optimization problem, eliminating the need for complex model training.

[0044] Furthermore, after determining the first target scan point for each tone range through the aforementioned steps, i.e., the initial calibration result, iterative optimization can be performed to improve the robustness and accuracy of the calibration results. More specifically, this iterative process aims to gradually expand more scan points for each tone range based on the existing target scan point, forming a "scan point cluster" that can more comprehensively characterize the acoustic properties of that tone range.

[0045] In one embodiment, each round of iterative optimization includes the following steps: based on the currently determined target scan point, traverse multiple different combinations of new scan points, wherein the new scan point combinations are composed of other scan points selected from the sound source scan grid for each tone region to be calibrated, excluding the target scan point; simultaneously, for each new scan point combination traversed, merge it with the target scan point of the corresponding tone region to form an extended scan point combination, and calculate the tone region localization accuracy of the extended scan point combination; finally, select the new scan point combination that makes the tone region localization accuracy the highest, and add the corresponding scan point to the calibration result.

[0046] Specifically, the above steps can also include a preliminary step of fixing historical results. First, the target scan points already determined in the current calibration results are fixed. Then, all scan points already selected as target scan points are removed from the entire sound source scan grid, and the remaining scan points form the candidate scan point set for the current iteration. The system then iterates through the candidate point set, selecting one scan point for each calibrated pitch region to form all possible new scan point combinations. This step is similar to the initial search, but the search scope is limited to unoccupied scan points. For each new scan point combination encountered, it is merged with the existing target scan points for the corresponding pitch region to form a new extended scan point combination containing more scan points. Then, based on the localization spectra of all calibration samples, the overall pitch region localization accuracy of this extended scan point combination is evaluated. The evaluation method is as follows: for each sample, in its extended scan point combination, for all scan points corresponding to each pitch region, including original points and new points, the maximum matching degree is found, and the pitch region corresponding to the maximum value is taken as the predicted pitch region. This is then compared with the true pitch region to calculate the accuracy. Meanwhile, among all the new scan point combinations traversed in this round of iteration, the new scan point combination that maximizes the accuracy of the expanded scan point combination is selected, and the scan points in this combination are added to the target scan point set of the corresponding tone range, thereby updating the calibration results and completing one round of iteration.

[0047] By repeating the above iterative process, multiple target scanning points can be gradually added to each vocal range. For example... Figure 4a , 4b Figures 4c, 4d, 4e, and 4f sequentially illustrate examples of optimized results where each register is expanded from one scan point to six scan points. This iterative mechanism enables the calibration system to adaptively capture richer spatial acoustic features of each register, thereby achieving more stable and accurate register localization in practical applications.

[0048] It should be noted that, with Figure 4f For example, the maximum of 6 scanning directions per pitch region shown is only a specific embodiment. In practical applications, the final number of target scanning points for each pitch region is not fixed, but can be flexibly adjusted according to the actual pitch region localization accuracy requirements. For example, when the accuracy has reached a preset threshold, the iteration can be stopped early; if higher localization robustness is required, the number of iterations can be increased to expand the number of scanning points for each pitch region. This flexibility allows this method to be precisely configured according to different performance requirements and computing resources.

[0049] Furthermore, the method described in this specification may also include a closed-loop correction mechanism to ensure the long-term adaptability and robustness of the calibration results in a real-world vehicle environment. Specifically, when the system performs actual sound zone localization based on the current calibration results, i.e., the target scanning points corresponding to each sound zone, if localization errors or abnormal results occur in certain use cases, these erroneous use cases can be used to iteratively optimize and correct the calibration results.

[0050] In one embodiment, if the sound region localization result based on the above calibration result is abnormal, the corresponding erroneous audio samples can be collected and the corresponding actual sound region can be recorded to form an erroneous sample set. At the same time, the above erroneous sample set is merged into the above calibration audio sample set as an extended calibration sample set. Finally, based on the above extended calibration sample set, multiple different combinations of scanning points are re-traversed to update the above calibration result.

[0051] Specifically, when a sound zone localization error is detected, the system or calibration personnel can collect the corresponding erroneous audio samples and accurately record the actual sound zone of the passenger emitting the sample, thus forming a targeted error sample set. These samples reflect the shortcomings of the current calibration results in specific scenarios or acoustic conditions. Next, the aforementioned error sample set is merged with the initial calibration audio sample set to form a broader and more challenging extended calibration sample set. Furthermore, the system can re-search for the combination of scan points that achieves the highest overall localization accuracy based on the extended samples containing erroneous cases, especially in cases of previous errors, thereby updating and outputting a new generation of calibration results. It is evident that this corrective step upgrades the sound zone calibration method from a static "offline calibration" scheme to a resilient system with continuous self-optimization capabilities. This allows the calibration results to better adapt to complex real-world acoustic environments without restarting large-scale data acquisition and calibration processes. Maintenance and performance improvement of the calibration model can be completed solely through incremental error sample collection and rapid re-optimization, significantly reducing the algorithm maintenance costs after mass production of the vehicle model.

[0052] As mentioned earlier, after obtaining the optimal sound zone scanning orientation, i.e., the set of target scanning points, it can be directly saved as a lightweight configuration file. In actual deployment, the same core sound zone localization algorithm library can be adapted to different vehicle models and cabin layouts by loading configuration files corresponding to different vehicle models. This realizes a maintenance mode where one algorithm can adapt to multiple vehicle models, completely solving the problems of algorithm fragmentation and high maintenance costs caused by the "one vehicle, one model" approach in traditional solutions. At the same time, since it completely avoids the training process relying on deep neural network models, as well as the massive data collection and complex acoustic simulation work required for training, this method can shorten the traditional calibration cycle of several months to a very short time.

[0053] Figure 5This is a flowchart illustrating an exemplary embodiment of a vehicle sound zone localization method, as shown in this specification. Figure 5 As shown, the interior of the vehicle's cabin is divided into multiple sound zones, each sound zone being equipped with at least one target scanning point obtained through the aforementioned sound zone calibration method; the method includes: Step S502: Obtain a multi-channel audio signal composed of multiple microphones of the vehicle, wherein the multi-channel audio signal includes the target voice signal to be located.

[0054] During actual vehicle operation, when sound zone localization is required, the vehicle-associated sound zone localization system first collects audio signals in real time using multiple microphones located within the cabin. These microphones work synchronously, combining the collected signals to form a multi-channel audio signal. This signal contains the target voice signal to be located, such as the user's voice command, and may also be mixed with ambient noise and other interference sounds.

[0055] Considering the unique characteristics of the in-vehicle acoustic environment, sounds generated by the vehicle's own audio system, such as stereos and navigation announcements, can be picked up by microphones, creating strong echo interference that severely impacts the accuracy of subsequent sound zone localization. Therefore, after obtaining the original multi-channel audio signal, this manual recommends performing echo cancellation preprocessing on it.

[0056] Specifically, the vehicle system simultaneously acquires M-channel microphone signals and R-channel reference signals. The M-channel microphone signals are directly picked up by the in-cabin microphone array and include target speech, ambient noise, and content played by the vehicle's audio system. The R-channel reference signals are directly taken from the output of the vehicle's audio system, such as the raw audio signal of currently playing music or navigation announcements. Both types of signals are input together to the vehicle's echo cancellation module.

[0057] This echo cancellation module, based on algorithms such as adaptive filtering, uses a reference signal to estimate and cancel the echo components associated with the microphone signal. After this step, the system output retains the M-channel audio signal, which has significantly reduced or even eliminated the echo interference from the vehicle's audio system, thus obtaining a multi-channel audio signal mainly composed of target speech and ambient sound. This lays the foundation for subsequent high-precision spatial feature extraction and sound zone localization.

[0058] Step S504: Calculate the corresponding positioning spectrum based on the multi-channel audio signal. The positioning spectrum is used to characterize the degree of matching between the target speech signal and the spatial orientation corresponding to each target scanning point.

[0059] Based on the acquired multi-channel audio signals, the system can calculate a localization spectrum that characterizes the degree of matching between the target speech signal and each preset target scanning point. The target scanning point is the optimal spatial location determined in advance for each vocal region using the aforementioned vocal region calibration method. This localization spectrum is generated through a signal processing algorithm, and the value in each dimension directly reflects the acoustic feature similarity between the current target speech signal and the spatial location represented by the corresponding target scanning point, i.e., the degree of matching.

[0060] The calculation process of the localization spectrum in the register localization process can be divided into three steps: 1. Calculate the covariance matrix of the target speech signal. Based on the preprocessed multi-channel audio signal, the system first calculates the covariance matrix of the target speech signal acquired from multiple microphones. This matrix statistically represents the spatial phase relationship of the target speech signal as it arrives at different microphones, and is the core data carrier reflecting the location characteristics of the sound source. This step is consistent with the principle of calculating the sample covariance matrix in the localization and calibration stage, the only difference being that the object processed here is the real-time acquired speech signal to be located.

[0061] 2. Calculating the matching degree based on the covariance matrix and steering vector. The system then uses the aforementioned covariance matrix, combined with the pre-stored steering vector corresponding to each target scan point, to calculate the matching degree between the target speech signal and each target scan point. The steering vector is a theoretical acoustic template pre-calculated for each target scan point during the calibration phase, representing the ideal phase difference of the sound wave arriving at each microphone from the direction of that scan point. The matching degree is typically calculated using algorithms such as beamforming, that is, by comparing the covariance matrix of the actual signal with the theoretical steering template of each scan point, the probability that the target speech signal originates from the direction of each target scan point is quantified.

[0062] 3. Constructing a localization spectrum from matching scores. This involves arranging the matching scores of the target speech signal calculated above with all target scan points in order of scan point location, thus constructing a localization spectrum for localization. This localization spectrum is a vector whose dimension equals the total number of target scan points, and the value of each element represents the degree of matching between the target speech signal and the corresponding target scan point. At this point, the system has completed the conversion from the original multi-channel signal to spatial orientation matching features, providing a direct basis for the final vocal region determination.

[0063] Furthermore, in the actual process of voice region localization, the calculation of the above covariance matrix can be optimized to accurately extract the spatial features of the target speech signal from the mixed signal that may contain background noise, reverberation and other interference. This optimization can be achieved by introducing a speech masking code.

[0064] In one embodiment, the system first obtains the speech masking code of the target speech signal. This masking code can be generated by a pre-installed speech wake-up module, a speech activity detection module, or a deep learning speech separation model. It is a two-dimensional time-frequency matrix used to characterize the weight or probability of each time-frequency unit in the original multi-channel audio signal belonging to the target speech signal. Specifically, the system can convert the echo-cancelled multi-channel audio signal... As input, where Frequency band number, This is the frame number. Simultaneously, the system acquires the masking code of the target speech signal. Used to characterize each time-frequency point Probability of belonging to the target speech signal: when the target speech signal dominates at this time-frequency point. ,otherwise .

[0065] Based on this, the weighted covariance matrix of the target speech can be calculated using Formula 5. : in, For frequency bands used for sound range localization, The effective time period of the target speech, such as the duration of the wake word. This represents the conjugate transpose of a multi-channel audio signal vector. The core of this calculation lies in assigning higher weights to the time-frequency points of the target speech indicated by the masking code, and then accumulating the correlation between the microphone signals. Mathematically, this process can be viewed as filtering the covariance estimate of the original signal using the masking code, thereby significantly suppressing the contribution of noise and interference, resulting in a final weighted covariance matrix. It can more purely and accurately represent the spatial phase relationship of the target speech signal itself among multiple microphones.

[0066] Of course, the above embodiments are only discussed under the ideal situation where only one type of speech signal exists in the multi-channel audio signal. The system can use blind source separation technology to generate the speech masking code of the target speech signal to deal with complex scenarios such as multiple speakers mixing in the vehicle. Specifically, the system can first input the multi-channel audio signal obtained after echo cancellation into the blind source separation algorithm module. This algorithm decomposes the mixed signal into multiple separated audio signals by utilizing the statistical independence between signals without relying on prior information about the sound source. Ideally, each output signal corresponds to an independent sound source, such as the voices of users in different voice ranges, such as the driver and passenger. After obtaining the multiple separated audio signals, the system can identify and determine which one corresponds to the target speech signal to be located, i.e., the target separated audio signal. The determination criteria can be preset audio conditions, such as: selecting the one with the strongest signal energy within a specific time period, such as the wake-up word period; or identifying the signal containing the valid wake-up word through the voice wake-up detection of the voice wake-up module; or combining other sensor information for comprehensive judgment. Finally, after determining the target audio signal, the system can generate the corresponding voice masking code by analyzing the time-frequency energy distribution of the signal. Specifically, the signal at each time and frequency point can be calculated. The energy or power on the surface is processed through normalization, threshold judgment, etc., and transformed into a probability value between 0 and 1.

[0067] Furthermore, the system can also analyze the calculated weighted covariance matrix. Phase Transform (PHAT) processing is performed to improve the robustness of the pitch localization algorithm to changes in signal amplitude and to make purer use of the phase information that is crucial for localization.

[0068] Specifically, the PHAT transformation affects the covariance matrix. Each complex element in (indicating the first) The and the first The microphone signal is in the frequency band The cross-correlation values ​​are normalized as shown in Formula 6: in, This represents the modulus of a complex number, i.e., the amplitude. This is a very small positive number used to avoid the denominator being zero, ensuring the stability of numerical calculations. The physical meaning of this transformation is that it converts each cross-correlation value... Divided by its own amplitude This yields a complex number with a modulus approximately equal to 1. ,in This is the original phase angle. This operation completely eliminates the influence of signal amplitude fluctuations, such as changes in speaker volume and differences in energy at different frequencies, on the correlation value. This ensures that the processed matrix elements retain and enhance only the core phase difference information between the sound reaching different microphones. In summary, the covariance matrix after PHAT transformation contains spatial phase relationships that are more robust to background noise and room reverberation.

[0069] After obtaining the covariance matrix after PHAT transformation Then, the system generates the final sound source localization spectrum through beamforming calculations. The calculation formula is as follows: in, This number represents the direction of sound source scanning. During the register localization stage, this number directly corresponds to the target scanning point pre-determined and stored for each register using the aforementioned calibration method. Each register may be configured with multiple target scanning points, which together constitute the directional template for that register. To be consistent with the scanning direction That is, the first Frequency band corresponding to each target scanning point The steering vector. This vector is pre-calculated and stored during the calibration phase, and is the theoretical phase difference template of the sound wave arriving at each microphone from the azimuth of the target scanning point. Represents the steering vector The conjugate transpose of . It is obtained after weighted and PHAT transformation, frequency band The covariance matrix on the array represents the spatial phase relationship of the current target speech signal generated between the actual microphone arrays. : This refers to the frequency band range used for sound region localization, and it is usually selected as the frequency band that is most effective for human voice localization.

[0070] In the above formula 7 This can be viewed as beamforming power calculation. Essentially, it involves calculating the spatial features of the actually observed target speech signal (…). ), to the Theoretical orientation template of each target scanning point ( Projection and matching are performed on the frequency band. This operation yields a scalar value representing the degree of matching between the actual signal and the theoretical template in this frequency band. Subsequently, in the effective frequency band... The matching results of all frequency bands are summed up to obtain the target speech signal and the first frequency band. Overall matching degree of the orientation of each target scanning point .

[0071] In summary, the matching degree can be calculated using the formulas described above. All Arranged according to the scan point sequence, this forms the localization spectrum used for sound region determination. This localization spectrum visually displays the distribution of the matching degree between the target speech signal and each pre-stored target location, and the direction of the scan point corresponding to its peak is the most likely source location of the sound determined by the algorithm.

[0072] Step S506: Determine the target audio region to which the target speech signal belongs based on the target scanning point corresponding to the highest matching degree.

[0073] Based on the calculated localization spectrum, the system identifies the target scanning point that best matches the target speech signal. Since each target scanning point is pre-assigned to a specific vocal region, the vocal region to which the highest matching point belongs is determined as the source location of the target speech signal, i.e., the target vocal region ultimately identified by the system. This step completes the mapping from signal spatial features to specific physical vocal regions, achieving precise vocal region localization.

[0074] Figure 6 This is a schematic structural diagram of an electronic device according to an exemplary embodiment. Please refer to... Figure 5 At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other necessary hardware. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it, forming a vehicle audio zone calibration device and a vehicle audio zone positioning device at the logical level. Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0075] Corresponding to the embodiments of the aforementioned vehicle audio zone calibration method, this specification also provides embodiments of a vehicle audio zone calibration device.

[0076] Please refer to Figure 7 , Figure 7 This is a schematic diagram illustrating the structure of a vehicle audio zone calibration device as shown in an exemplary embodiment. Figure 7 As shown, the interior of the vehicle's cabin is divided into multiple pitch-fixing zones; the device includes: The sample acquisition unit 702 is used to acquire a sound source scanning grid and a calibration audio sample set for the vehicle. The scanning points of the sound source scanning grid cover the spatial orientation inside the cabin, and each calibration audio sample in the calibration audio sample set is associated with the real sound region where it was collected. The localization spectrum calculation unit 704 is used to calculate the localization spectrum of the audio signal contained in each calibrated audio sample, wherein the localization spectrum is used to characterize the degree of matching between the audio signal and the spatial orientation corresponding to each scanning point. The scan point combination traversal unit 706 is used to traverse multiple different scan point combinations, wherein each scan point combination consists of scan points selected from the sound source scan grid for each calibrated sound region; and performs the following operations on each traversed scan point combination: based on the localization spectrum of each calibrated audio sample, determines the maximum matching degree of each calibrated audio sample on each scan point in the scan point combination, and takes the sound region corresponding to the maximum matching degree as the predicted sound region of the corresponding calibrated audio sample; and determines the sound region localization accuracy of the scan point combination based on the consistency between the predicted sound region of each calibrated audio sample and the corresponding real sound region. The calibration result determination unit 708 is used to select the combination of scanning points with the highest accuracy in locating the sound region from the traversed combination of scanning points as the calibration result, and each scanning point in the calibration result is used as the target scanning point for locating the corresponding sound region to be calibrated.

[0077] Optionally, the sample acquisition unit 702 is specifically used for: Obtain the geometric model of the interior of the cabin; Based on the geometric model, the sound source scanning grid is constructed according to a preset spatial interval. Each scanning point in the sound source scanning grid corresponds to a steering vector. The steering vector is used to characterize the phase difference relationship between the sound wave and the multiple microphones of the vehicle from the spatial orientation of the corresponding scanning point.

[0078] Optionally, the positioning spectrum calculation unit 704 is specifically used for: Based on the audio signal of the calibrated audio sample, calculate the covariance matrix among the signals collected by the multiple microphones; Based on the covariance matrix and the steering vector corresponding to each scan point in the sound source scanning grid, the matching degree between the audio signal and each scan point is calculated. The localization spectrum is composed of the matching degree corresponding to each scanning point.

[0079] Optionally, the device further includes: The scanning point combination extension unit is used to traverse multiple different new scanning point combinations based on the currently determined target scanning point. The new scanning point combination consists of other scanning points selected from the sound source scanning grid for each pitch region to be calibrated, excluding the target scanning point. For each newly added scan point combination encountered, it is merged with the target scan point of the corresponding vocal range to form an extended scan point combination, and the vocal range localization accuracy of the extended scan point combination is calculated. Select the new combination of scanning points that achieves the highest accuracy in locating the sound region, and add the corresponding scanning points to the calibration results.

[0080] Optionally, the process of constructing the scan point combination is subject to at least one of the following constraints: Based on the symmetrical layout of the multiple to-be-calibrated pitch zones in the vehicle cabin, the process of selecting scanning points for symmetrically arranged pitch zones is subject to correlation constraints. Based on the positional order of the multiple to be calibrated pitch zones in the front-rear direction of the vehicle, the coordinate relationship of the scanning points selected for different pitch zones in the front-rear direction of the vehicle is constrained sequentially.

[0081] Optionally, if the constraint includes an associative constraint, the associative constraint is such that the scan points selected for the left-right symmetrically arranged sound zones are symmetrical in the horizontal direction with respect to the longitudinal center plane of the vehicle. When the constraint includes a sequence constraint, the sequence constraint is such that the coordinates of the scan point selected for the rear audio zone in the vehicle's longitudinal direction do not exceed the corresponding scan point selected for the front audio zone.

[0082] Optionally, the device further includes: An extended calibration sample set determination unit is used to collect corresponding erroneous audio samples and record the corresponding actual audio regions when the sound region localization result based on the calibration result is abnormal, so as to form an erroneous sample set. The erroneous sample set is incorporated into the calibration audio sample set to serve as an extended calibration sample set; Based on the extended calibration sample set, multiple different combinations of scan points are re-traversed to update the calibration results.

[0083] Please refer to Figure 8 , Figure 8 This is a schematic diagram illustrating the structure of a vehicle sound zone positioning device as an exemplary embodiment. Figure 8 As shown, the interior of the vehicle's cabin is divided into multiple sound zones, each sound zone being equipped with at least one target scanning point obtained through the aforementioned sound zone calibration method; the device includes: The multi-channel audio signal acquisition unit 802 is used to acquire multi-channel audio signals composed of multiple microphones of the vehicle, wherein the multi-channel audio signals include the target voice signal to be located. The positioning spectrum calculation unit 804 is used to calculate the corresponding positioning spectrum based on the multi-channel audio signal. The positioning spectrum is used to characterize the degree of matching between the target speech signal and the spatial orientation corresponding to each target scanning point. The target voice region determination unit 806 is used to determine the target voice region to which the target speech signal belongs based on the target scanning point corresponding to the highest matching degree.

[0084] Optionally, the positioning spectrum calculation unit 804 is specifically used for: Based on the target speech signal, calculate the covariance matrix among the signals collected by the multiple microphones; Based on the covariance matrix and the steering vector corresponding to each target scanning point, the matching degree between the target speech signal and each target scanning point is calculated. The localization spectrum is composed of the matching degree corresponding to each scanning point.

[0085] Optionally, the positioning spectrum calculation unit 804 is specifically used for: The speech masking code of the target speech signal is obtained based on the multi-channel audio signal. The speech masking code is used to characterize the weight of each time-frequency point in the multi-channel audio signal belonging to the target speech signal. The weighted covariance matrix is ​​calculated based on the multi-channel audio signal and the speech masking code, and is used as the covariance matrix.

[0086] Optionally, the positioning spectrum calculation unit 804 is specifically used for: Blind source separation is performed on the multi-channel audio signal to obtain a multi-channel separated audio signal; Determine the target separated audio signal that meets the preset audio conditions from the multi-channel separated audio signals; Based on the target-separated audio signal, the voice masking code is generated.

[0087] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0088] Based on the same concept as the methods described above, this specification also provides a vehicle, including: a processor and a memory for storing processor-executable instructions; wherein the processor executes the executable instructions to implement the steps of the method as described in any of the above embodiments.

[0089] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0090] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0091] The embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0092] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0093] Suitable computers for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a GPS receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0094] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0095] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0096] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0097] Therefore, specific embodiments of the subject matter have been described. Furthermore, the processes depicted in the figures are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0098] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

Claims

1. A method for vehicle sound zone calibration, characterized in that, The vehicle's interior is divided into multiple pitch-fixing zones; the method includes: Acquire a sound source scanning grid and a calibration audio sample set for the vehicle. The scanning points of the sound source scanning grid cover the spatial orientation inside the cabin. Each calibration audio sample in the calibration audio sample set is associated with the real sound region where it was collected. For each calibrated audio sample, the localization spectrum of the audio signal it contains is calculated. The localization spectrum is used to characterize the degree of matching between the audio signal and the spatial orientation corresponding to each scanning point. The algorithm iterates through multiple different combinations of scan points, where each combination consists of scan points selected from the sound source scanning grid for each calibrated audio region. For each traversed combination of scan points, the following operations are performed: based on the localization spectrum of each calibrated audio sample, the algorithm determines the maximum matching degree of each calibrated audio sample at each scan point in the combination, and uses the audio region corresponding to the maximum matching degree as the predicted audio region of the corresponding calibrated audio sample; based on the consistency between the predicted audio region of each calibrated audio sample and the corresponding real audio region, the algorithm determines the localization accuracy of the scan point combination. The combination of scanning points with the highest accuracy in locating the vocal range is selected from the traversed combinations of scanning points as the calibration result, and each scanning point in the calibration result is used as the target scanning point for locating the corresponding vocal range to be calibrated.

2. The method according to claim 1, characterized in that, The step of obtaining the sound source scanning grid for the vehicle includes: Obtain the geometric model of the interior of the cabin; Based on the geometric model, the sound source scanning grid is constructed according to a preset spatial interval. Each scanning point in the sound source scanning grid corresponds to a steering vector. The steering vector is used to characterize the phase difference relationship between the sound wave and the multiple microphones of the vehicle from the spatial orientation of the corresponding scanning point.

3. The method according to claim 2, characterized in that, The step of calculating the localization spectrum of the audio signal contained in each calibrated audio sample includes: Based on the audio signal of the calibrated audio sample, calculate the covariance matrix among the signals collected by the multiple microphones; Based on the covariance matrix and the steering vector corresponding to each scan point in the sound source scanning grid, the matching degree between the audio signal and each scan point is calculated. The localization spectrum is composed of the matching degree corresponding to each scanning point.

4. The method according to claim 1, characterized in that, The method further includes: Based on the currently determined target scanning point, multiple different combinations of new scanning points are traversed. The new scanning point combination consists of other scanning points selected from the sound source scanning grid for each pitch range to be calibrated, excluding the target scanning point. For each newly added scan point combination encountered, it is merged with the target scan point of the corresponding vocal range to form an extended scan point combination, and the vocal range localization accuracy of the extended scan point combination is calculated. Select the new combination of scanning points that achieves the highest accuracy in locating the sound region, and add the corresponding scanning points to the calibration results.

5. The method according to claim 1, characterized in that, The process of constructing the scan point combination is subject to at least one of the following constraints: Based on the symmetrical layout of the multiple to-be-calibrated pitch zones in the vehicle cabin, the process of selecting scanning points for symmetrically arranged pitch zones is subject to correlation constraints. Based on the positional order of the multiple to be calibrated pitch zones in the front-rear direction of the vehicle, the coordinate relationship of the scanning points selected for different pitch zones in the front-rear direction of the vehicle is constrained sequentially.

6. The method according to claim 5, characterized in that, When the constraint includes an associative constraint, the associative constraint is such that the scan point selected for the left-right symmetrically arranged sound zones is symmetrical in the horizontal direction with respect to the longitudinal center plane of the vehicle. When the constraint includes a sequence constraint, the sequence constraint is such that the coordinates of the scan point selected for the rear audio zone in the vehicle's longitudinal direction do not exceed the corresponding scan point selected for the front audio zone.

7. The method according to claim 1, characterized in that, The method further includes: If the sound region localization result based on the calibration result is abnormal, the corresponding erroneous audio samples are collected and the corresponding actual sound region is recorded to form an erroneous sample set. The erroneous sample set is incorporated into the calibration audio sample set to serve as an extended calibration sample set; Based on the extended calibration sample set, multiple different combinations of scan points are re-traversed to update the calibration results.

8. A method for locating the sound zone of a vehicle, characterized in that, The interior of the vehicle is divided into multiple sound zones, each sound zone being equipped with at least one target scanning point obtained by the sound zone calibration method as described in any one of claims 1-7; the method includes: Acquire a multi-channel audio signal composed of multiple microphones of the vehicle, wherein the multi-channel audio signal contains the target voice signal to be located; The corresponding localization spectrum is calculated based on the multi-channel audio signal. The localization spectrum is used to characterize the degree of matching between the target speech signal and the spatial orientation corresponding to each target scanning point. The target voice region to which the target speech signal belongs is determined based on the target scanning point corresponding to the highest matching degree.

9. The method according to claim 8, characterized in that, The calculation of the corresponding positioning spectrum includes: Based on the target speech signal, calculate the covariance matrix among the signals collected by the multiple microphones; Based on the covariance matrix and the steering vector corresponding to each target scanning point, the matching degree between the target speech signal and each target scanning point is calculated. The localization spectrum is composed of the matching degree corresponding to each scanning point.

10. The method according to claim 9, characterized in that, The calculation of the covariance matrix among the signals acquired by the multiple microphones includes: The speech masking code of the target speech signal is obtained based on the multi-channel audio signal. The speech masking code is used to characterize the weight of each time-frequency point in the multi-channel audio signal belonging to the target speech signal. The weighted covariance matrix is ​​calculated based on the multi-channel audio signal and the speech masking code, and is used as the covariance matrix.

11. The method according to claim 10, characterized in that, The step of obtaining the speech masking code of the target speech signal based on the multi-channel audio signal includes: Blind source separation is performed on the multi-channel audio signal to obtain a multi-channel separated audio signal; Determine the target separated audio signal that meets the preset audio conditions from the multi-channel separated audio signals; Based on the target-separated audio signal, the voice masking code is generated.

12. A vehicle, characterized in that, include: A processor; a memory for storing processor-executable instructions; wherein the processor implements the steps of the method as described in any one of claims 1 to 11 by executing the executable instructions.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 11.

14. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 11.

Images