Flat diaphragm car audio

Abstract

The invention provides a flat diaphragm vehicle-mounted sound system, which realizes vehicle-mounted active acoustic enhancement based on multi-source data fusion, and comprises a road acoustic attribute preloading unit for acquiring road surface material coefficient, curvature and gradient data from a navigation map; the real-time road surface sensing unit is used for acquiring roughness and fluctuation characteristics through a vision and three-dimensional sensor; the sound field dynamic compensation engine fuses the road data, the real-time road condition and the vehicle speed signal to generate sound image offset, low-frequency gain and voice definition parameters; and the audio output control module is used for processing the signals in different frequency bands and reconstructing a three-dimensional sound field. The technical key lies in establishing a road feature-acoustic response mapping model, and realizing curvature / vehicle speed-based sound image stability control, material / roughness-driven low-frequency compensation and gradient-adaptive voice optimization. And the limitations of response lag and insufficient scene adaptability of the traditional noise reduction technology are broken through.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle-mounted audio systems, and relates to a flat diaphragm automobile audio system, in particular to a flat diaphragm automobile audio system based on multi-source data fusion for acoustic enhancement and a method thereof. BACKGROUND

[0002] The core of the flat diaphragm automobile audio system is a large-area, extremely thin and high-rigidity flat diaphragm (commonly made of composite honeycomb or fiber material), the surface of which is directly etched with voice coil wires, and the flat diaphragm is suspended in a strong magnetic field generated by a high-efficiency neodymium magnet array (the magnetic flux density can reach 1.5 Tesla). When an electric current passes through the voice coil, the entire diaphragm is uniformly stressed under the action of the magnetic field and pushes the air to produce sound, thus breaking away from the segmented vibration of the traditional cone-piston, achieving extremely low distortion, superb transient response and natural and transparent sound quality, and being particularly good at restoring delicate mid-high frequencies. Its ultra-thin physical form (the thickness of a single module can be controlled within 5 cm) and the 180° wide directivity radiation characteristics of the nearly plane wave make it an ideal choice for building a loudspeaker array.

[0003] With the development of intelligent automobile technology, the vehicle-mounted acoustic system faces the adaptability challenge under complex road environment. The traditional active noise reduction (ANC) technology mainly relies on the feedback of the in-vehicle microphone, and has problems such as response lag (>100 ms) and inability to cope with transient road noise. The existing noise reduction schemes based on cameras or single sensors can identify some road features, but lack deep coupling with navigation data and vehicle dynamics, resulting in insufficient sound field compensation accuracy. Especially when driving on a curve, the sound image shift caused by the centrifugal force of the vehicle will interfere with the auditory localization, and the wideband noise generated by the rough road will significantly reduce the speech intelligibility.

[0004] For example, CN118782008A discloses a vehicle interior noise active control method and system based on road roughness preview, which acquires the current preview road roughness at the current time and the previous preview road roughness at the previous time, and when the difference between the current preview road roughness and the previous preview road roughness is greater than a preset threshold, the vibration acceleration signal corresponding to the current preview road roughness is acquired; the vibration acceleration signal is filtered to obtain a secondary anti-noise signal; the signal output delay time and the noise source signal are acquired, and when the signal output delay time is reached, the secondary anti-noise signal and the noise source signal are acoustically superimposed to cancel the noise source signal.

[0005] CN117037765A discloses a road noise active control variable step length system based on road classification, provides a road noise active control variable step length system based on road classification, carries out road noise active control variable step length based on image recognition, classifies the road state (roughness) by analyzing the obtained road information, realizes the update of the step length of the road noise active control system, achieves the purpose of variable step length, and can also use the random pulse on the road for preview to dynamically adjust the step length of the control system, thereby improving the adaptability of the road noise active control system to different driving roads and the noise reduction effect.

[0006] It can be seen that the current solution focuses on noise suppression and ignores active optimization of the acoustic environment, which is difficult to meet the demand of intelligent cockpit for immersive auditory experience. There is an urgent need for a sound field enhancement technology that integrates multi-source data prediction and has environmental adaptability.

[0007] In the car, multiple flat panel units can be seamlessly embedded in narrow spaces such as doors, ceilings or instrument tables, forming a large area of continuous sound source. Through independent phase and delay control of each unit in the array by advanced DSP algorithm, precise "beam forming" can be realized - dynamically focusing sound energy on the heads of passengers in different seats, compensating for the sound field shift caused by changes in vehicle speed, noise or passenger position, and significantly improving multi-seat listening consistency. SUMMARY

[0008] To solve the above problems, the present application provides a flat diaphragm car audio system based on multi-source data fusion to realize acoustic enhancement, which has a flat diaphragm loudspeaker array and an audio processing module, and further comprises:

[0009] Road acoustic property preloading unit: connected to the vehicle navigation system, configured to read the pre-stored road feature data in the navigation map, the road feature data including road surface material acoustic transfer coefficient, road curvature geometric parameter and slope angle parameter;

[0010] Real-time road surface perception unit: including visual sensors and three-dimensional scanning sensors installed at the front end of the vehicle, configured to obtain road micro-roughness features through texture feature extraction algorithm, and obtain road macro-undulation features through point cloud processing;

[0011] Sound field dynamic compensation engine: configured to receive the road feature data and real-time road feature data, and perform spatio-temporal synchronous fusion with the vehicle bus transmitted vehicle speed signal, generate a set of dynamic sound field compensation parameters based on the coupling relationship of multi-source data, the parameter set including sound image positioning horizontal offset, low-frequency propagation gain compensation coefficient and speech frequency band filter transfer function;

[0012] An audio output control module configured to perform frequency domain reshaping and spatial reconstruction on an input audio signal according to the dynamic sound field compensation parameter set, and drive a distributed flat diaphragm loudspeaker array to output an environment-adaptive sound field.

[0013] The road acoustic property preloading unit:

[0014] Obtain road acoustic fingerprint data packets from a cloud server through a vehicle networking communication protocol, wherein the data packets contain a material-acoustic response mapping relationship constructed based on historical vehicle cluster collection data;

[0015] Start data preloading when the vehicle is at a preset distance threshold from the target road section, and establish a binding relationship between road feature data and the vehicle coordinate system.

[0016] The real-time road surface perception unit:

[0017] The visual sensor is configured to perform multi-scale texture analysis and output a roughness parameter related to the sound absorption characteristics of the road surface;

[0018] The three-dimensional scanning sensor is configured to generate a road surface height variation gradient vector through laser time-of-flight difference calculation;

[0019] The unit contains a data alignment module that converts sensor data to the vehicle body coordinate system through a coordinate transformation matrix.

[0020] The sound field dynamic compensation engine includes:

[0021] A sound image stability calculation module that calculates a horizontal offset compensation amount ΔX based on the road curvature radius R, the vehicle speed v, and the steering system state parameters through a sound field deflection model, wherein ΔX increases as the curvature radius R decreases and is directly proportional to the square of the vehicle speed v;

[0022] A low-frequency response adaptation module that linearly combines the road surface material acoustic transfer coefficient μ and the real-time roughness parameter ρ to generate a low-frequency gain compensation coefficient G L , wherein the product of G L and (μ·ρ) is positively correlated;

[0023] A speech intelligibility optimization module that dynamically adjusts the quality factor Q V of the bandpass filter according to the slope angle θ and the vehicle acceleration state, wherein Q V increases as the absolute value of the slope increases.

[0024] The curve sound image stability module:

[0025] When the detected steering angular velocity exceeds the dynamic stability threshold, an anti-interference enhancement mode is activated:

[0026] Introducing the derivative of steering angle velocity ω in horizontal offset compensation amount ΔX

[0027] Applying a smoothing window function to the output sound field suppresses transient mutations.

[0028] The audio output control module:

[0029] Using a frequency division processing architecture to decompose the audio signal into three independent subbands:

[0030] The low-frequency subband signal applies the gain compensation generated by the low-frequency response adaptation module

[0031] The mid-frequency speech subband signal passes through the time-varying filter configured by the speech intelligibility optimization module

[0032] The high-frequency subband signal remains in its original state.

[0033] Applying a sound image offset matrix to the compensated subband signal for sound field spatial synthesis, the matrix contains a rotation component determined by the horizontal offset compensation amount ΔX.

[0034] Also contains a dynamic calibration mechanism:

[0035] Collect sound field response signals through in-vehicle microphone arrays;

[0036] Calculate the residual vector of the actual sound pressure spectrum and the target compensation spectrum;

[0037] When the Euclidean norm of the residual vector exceeds the adaptive threshold, trigger the inverse parameter iterative update:

[0038] H new(f) = H old(f) + η·FFT[e(t)];

[0039] Where η is the convergence factor, e(t) is the residual time domain signal, and the update delay is controlled within the system response period.

[0040] A vehicle-mounted active acoustic enhancement method based on the system, comprising the following steps:

[0041] (1) Preloading phase:

[0042] According to the navigation path planning, load the road feature data packet before entering the target acoustic influence area, and establish the mapping relationship from the road coordinate system to the vehicle coordinate system;

[0043] (2) Multi-source fusion phase:

[0044] Receive road feature data, real-time road surface features and vehicle speed signals through a space-time synchronization interface;

[0045] The road curvature parameter is converted to a vehicle motion coordinate system, and the material characteristic parameter is fused with the real-time roughness;

[0046] (3) Parameter generation stage:

[0047] The sound image shift matrix parameter is calculated based on the curvature-vehicle speed coupling relationship;

[0048] The low-frequency gain coefficient is determined by linear combination of material characteristics and roughness;

[0049] The speech filter bandwidth is adjusted according to the slope angle and vehicle posture;

[0050] (4) Sound field reconstruction stage:

[0051] The audio signal is sub-band decomposed, and the compensation parameter is independently applied to each sub-band;

[0052] The sound image shift matrix is applied for multi-channel spatial rendering;

[0053] (5) Dynamic optimization stage:

[0054] The compensation deviation is detected by residual spectrum analysis;

[0055] When the deviation exceeds the adaptive tolerance, the inverse phase parameter is updated and injected into the processing pipeline.

[0056] In the sound field parameter generation stage (3):

[0057] The calculation of the sound image shift matrix parameter uses a kinematic model:

[0058]

[0059] k is the system calibration coefficient, and φ is the vehicle yaw angle;

[0060] When the steering angular velocity ω exceeds the threshold value, ΔX is corrected as:

[0061]

[0062] The generation of the low-frequency gain coefficient uses a feature weighting model:

[0063] G L = α·μ + β·ρ;

[0064] α, β are material characteristic weight factors.

[0065] In the dynamic optimization stage (5):

[0066] A double trigger condition is set to control parameter updating:

[0067] a. When the residual energy in the speech frequency band continuously exceeds the background noise level set by a certain multiple,

[0068] b. When the vehicle motion state mutation causes the acoustic environment change index to exceed the critical value;

[0069] The parameter updating process adopts a least mean square adaptive algorithm, and a convergence factor η decreases with the increase of vehicle speed.

[0070] The beneficial effects of the present application are:

[0071] The present application realizes the advanced prediction and active adaptation of the in-vehicle acoustic environment through the deep fusion of the preloaded road acoustic properties of the navigation system and real-time sensor data. Compared with traditional passive noise reduction technology, the system can start sound field optimization 500 milliseconds before the vehicle enters a special road section, effectively eliminating the sound image deviation and low-frequency resonance problems caused by sudden changes in road conditions. By dynamically adjusting the three-dimensional sound field parameters, passengers can experience stable sound source positioning during continuous curve driving and maintain voice call clarity on rough road surfaces, creating an immersive and disturbance-free driving acoustic environment.

[0072] A mapping model of road physical characteristics and acoustic compensation parameters is established, which converts the traditionally independent navigation data, vehicle motion state and environmental perception information into quantifiable sound field control instructions. The system can intelligently distinguish different road types and driving conditions, automatically optimize low-frequency response characteristics to suppress road vibration noise, and dynamically improve the signal-to-noise ratio of the voice frequency band.

[0073] The concept of "active sound field enhancement" is introduced into the field of vehicle audio, and through the frequency band processing architecture and spatial sound field reconstruction technology, the function transition from "noise suppression" to "environment optimization" is realized. Not only compensates for the influence of road environment on sound quality, but also combines vehicle motion state for sound image stability control, making the audio output and driving scene form a deep synergy. It breaks through the lag compensation mode relying on microphone feedback and establishes a "prediction-adaptation-optimization" acoustic control paradigm, setting a new standard for the auditory experience of intelligent cockpit.

[0074] The sound image stability algorithm significantly reduces the auditory directional error during curve driving, avoiding the distraction of the driver caused by sound source positioning deviation. The closed-loop optimization mechanism continuously monitors the deviation between the actual sound field and the target parameters, and completes dynamic calibration within 20 milliseconds, ensuring the stability of the compensation effect under various working conditions. When the system detects extreme driving conditions, it automatically triggers the safety protection mode, preventing the sound field from mutating and causing interference to the passengers through smooth control strategy, providing redundant protection for driving safety from the acoustic dimension.

[0075] The sound field compensation engine, based on multi-source data fusion, forms an open technology platform compatible with future high-precision maps, vehicle-infrastructure collaboration, and other emerging data sources. Parametric modeling methods allow for continuous optimization of the material-acoustic response database via cloud-based updates, enabling the system to evolve continuously. This framework provides a unified acoustic optimization foundation for applications such as in-car entertainment systems, navigation voice prompts, and hazard warning tones, driving the development of intelligent automotive acoustic systems towards environmental awareness and intelligent decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0077] Attachment Figure 1 This is a schematic diagram of the overall architecture of the present invention;

[0078] Attachment Figure 2 Schematic diagram of the optimization principle of the present invention.

[0079] Attachment Figure 3 Schematic diagram of the sensor arrangement of the present invention. DETAILED DESCRIPTION

[0080] Example 1:

[0081] See attached Figure 1 To the attached Figure 3 A flat-panel diaphragm car audio system that achieves acoustic enhancement based on multi-source data fusion, with a flat-panel diaphragm speaker array and an audio processing module, including:

[0082] A road acoustic property preloading unit is connected to the vehicle navigation system and configured to read road characteristic data pre-stored in the navigation map, wherein the road characteristic data includes an acoustic transmission coefficient of a road surface material, a road curvature geometric parameter, and a slope angle parameter;

[0083] Real-time road surface perception unit: This unit includes a visual sensor and a 3D scanning sensor mounted on the front of the vehicle. It is configured to obtain microscopic road surface roughness characteristics through texture feature extraction algorithms and macroscopic road surface undulation characteristics through point cloud processing.

[0084] A sound field dynamic compensation engine is configured to receive the road feature data and real-time road surface feature data, perform spatiotemporal synchronization fusion with the vehicle speed signal transmitted by the vehicle bus, and generate a dynamic sound field compensation parameter set based on the multi-source data coupling relationship, wherein the parameter set includes a sound image positioning horizontal offset, a low-frequency propagation gain compensation coefficient, and a voice band filter transfer function;

[0085] The audio output control module is configured to perform frequency domain reshaping and spatial reconstruction on the input audio signal according to the dynamic sound field compensation parameter set, and drive the distributed flat diaphragm loudspeaker array to output an environment-adaptive sound field.

[0086] The road acoustic property preloading unit:

[0087] Obtain the road acoustic fingerprint data packet from the cloud server through the Internet of Vehicles communication protocol, and the data packet contains a material-acoustic response mapping relationship constructed based on historical vehicle cluster collection data;

[0088] Start data preloading when the vehicle is at a preset distance threshold from the target road section, and establish a binding relationship between road feature data and the vehicle coordinate system.

[0089] The real-time road surface perception unit:

[0090] The visual sensor is configured to perform multi-scale texture analysis and output a roughness parameter related to the sound absorption characteristics of the road surface;

[0091] The three-dimensional scanning sensor is configured to generate a road surface height change gradient vector through laser time-of-flight difference calculation;

[0092] The unit includes a data alignment module that converts sensor data to the vehicle body coordinate system through a coordinate transformation matrix.

[0093] The sound field dynamic compensation engine includes:

[0094] The sound image stability calculation module calculates the horizontal offset compensation amount ΔX based on the road curvature radius R, the vehicle speed v, and the steering system state parameters through a sound field deflection model, wherein ΔX increases as the curvature radius R decreases, and is proportional to the square of the vehicle speed v;

[0095] The low-frequency response adaptation module linearly combines the road surface material acoustic transfer coefficient μ and the real-time roughness parameter ρ to generate a low-frequency gain compensation coefficient G L , wherein G L is positively correlated with the product of (μ·ρ);

[0096] The speech intelligibility optimization module dynamically adjusts the quality factor Q of the bandpass filter according to the slope angle θ and the vehicle acceleration state, wherein Q V increases as the absolute value of the slope increases. V

[0097] The curve sound image stability module:

[0098] When the detected steering angular velocity exceeds the dynamic stability threshold, the anti-interference enhancement mode is activated: ​

[0099] Introducing the derivative term of steering angular velocity ω in horizontal offset compensation amount ΔX

[0100] Applying a smoothing window function to the output sound field suppresses transient mutations.

[0101] The audio output control module:

[0102] The audio signal is decomposed into three independent subbands using a frequency division processing architecture:

[0103] The low-frequency subband signal applies the gain compensation generated by the low-frequency response adaptation module

[0104] The mid-frequency speech subband signal passes through the time-varying filter configured by the speech intelligibility optimization module

[0105] The high-frequency subband signal remains in its original state.

[0106] The compensated subband signals are subjected to sound field spatial synthesis by applying a sound image shift matrix, which contains a rotation component determined by the horizontal offset compensation amount ΔX.

[0107] It also contains a dynamic calibration mechanism:

[0108] The sound field response signal is collected by an in-vehicle microphone array.

[0109] The residual vector of the actual sound pressure spectrum and the target compensation spectrum is calculated.

[0110] When the Euclidean norm of the residual vector exceeds the adaptive threshold, the inverse parameter iterative update is triggered:

[0111] H new(f) = H old(f) + η·FFT[e(t)];

[0112] Where η is the convergence factor, e(t) is the residual time-domain signal, and the update delay is controlled within the system response period.

[0113] Embodiment 2:

[0114] A vehicle-mounted active acoustic enhancement method based on the system, comprising the following steps:

[0115] (1) Pre-loading stage:

[0116] According to the navigation path planning, load the road feature data packet before entering the target acoustic influence area, and establish the mapping relationship from the road coordinate system to the vehicle coordinate system;

[0117] Specifically in this step:

[0118] Road acoustic property acquisition, when the vehicle navigation sets the route, the system requests the road acoustic label from the cloud database through the vehicle networking protocol: trigger condition: start preloading when the distance D from the target section is less than 2km.

[0119] Data binding: establish the conversion matrix of the road coordinate system to the vehicle coordinate system.

[0120] (2) Multi-source fusion stage:

[0121] Receive road feature data, real-time road surface features and vehicle speed signals through the space-time synchronization interface;

[0122] Convert the road curvature parameter to the vehicle motion coordinate system, and perform feature fusion of the material characteristic parameter and the real-time roughness;

[0123] Specifically, in this step:

[0124] The visual sensor (front-view camera) performs multi-scale texture analysis:

[0125]

[0126] Output roughness level: p∈[0,1].

[0127] Wherein, the texture feature extraction:

[0128] Sobel edge density calculation:

[0129]

[0130] Wherein, I is the gray image matrix, and N is the total number of pixels.

[0131] Gray level co-occurrence matrix (GLCM) contrast calculation:

[0132]

[0133] P(i,j) is the joint probability distribution of gray levels with a distance of d pixels.

[0134] Roughness index: p = w1*EdgeDensity + w2*Contrast;

[0135] Three-dimensional scanning sensor (laser radar):

[0136] Calculate the height change gradient:

[0137] Space-time alignment: fuse sensor data through extended Kalman filtering:

[0138] X k = A·X {k-1} +B·uk +w k

[0139] Z k =H·X k +v k

[0140] Where A is the state transfer matrix and H is the observation matrix.

[0141] (3) Parameter generation stage:

[0142] Calculate the parameters of the sound image offset matrix based on the curvature-vehicle speed coupling relationship;

[0143] The low-frequency gain coefficient is determined by a linear combination of material properties and roughness;

[0144] Adjust the voice filter bandwidth according to the slope angle and vehicle posture;

[0145] Sound field compensation:

[0146] Pan offset:

[0147] v: vehicle speed (m / s), R: road curvature radius (m), φ: vehicle yaw angle (rad), k: system calibration coefficient (default 0.1);

[0148] Emergency Compensation:

[0149] λ = 0.05 (angular acceleration gain), η = 0.1 (lateral acceleration gain), a y : lateral acceleration (g).

[0150] Low Frequency Gain:

[0151] μ: pavement material coefficient (asphalt = 0.2, cement = 0.35);

[0152] ρ: fusion roughness;

[0153] height gradient;

[0154] Weight coefficients: α=1.2, β=0.8, γ=0.5.

[0155] Speed ​​adjustment:

[0156] κ=0.15 (vehicle speed sensitivity coefficient), v ref =60km / h (reference speed);

[0157] Voice Q value: Q V =Q0+γ Q*|θ|+δ Q *a x

[0158] Center frequency: f c =1500+50*sign(a x )*|a x | {0.5} , parameter coupling:

[0159] (4) Sound field reconstruction stage:

[0160] Sub-band decomposition is applied to the audio signal, and compensation parameters are independently applied to each sub-band. Sound image shift matrix is applied for multi-channel spatial rendering.

[0161] (5) Dynamic optimization stage:

[0162] Compensation deviation is detected by residual spectrum analysis.

[0163] When the deviation exceeds the adaptive tolerance, update the inverse phase parameter and inject it into the processing pipeline. In the sound field parameter generation stage (3):

[0164] The calculation of the sound image shift matrix parameter uses a kinematic model:

[0165]

[0166] k is the system calibration coefficient, and φ is the vehicle yaw angle.

[0167] When the steering angular velocity ω exceeds the threshold value, ΔX is corrected as:

[0168]

[0169] The generation of the low-frequency gain coefficient uses a feature weighting model:

[0170] G L =α·μ+β·ρ;

[0171] α, β are material characteristic weight factors.

[0172] In the dynamic optimization stage (5):

[0173] Set double trigger condition control parameter update:

[0174] a. When the residual energy in the speech frequency band exceeds the background noise level set multiple times,

[0175] b. When the vehicle motion state mutation causes the acoustic environment change index to exceed the critical value;

[0176] The parameter update process uses a least mean square adaptive algorithm, and the convergence factor η decreases with increasing vehicle speed.

[0177] Pre-loading phase detailed process

[0178] The pre-loading phase is a critical process for the system to prepare data and initialize configuration before the vehicle enters the target road section. When the driver sets the travel route through the vehicle-mounted navigation system, the system immediately starts the route analysis algorithm to scan special road feature points in the entire path. These feature points include sharp turning sections (curvature radius less than 300 meters), special pavement, and slopes with a gradient change of more than 5%. The system connects to the cloud road feature database through the 4G / 5G vehicle networking module, indexed by road section ID and geographic location coordinates, to request the download of acoustic property data packets for these special road sections.

[0179] During the data packet download process, the system simultaneously starts the vehicle position real-time tracking module. By fusing GNSS satellite positioning signals, vehicle-mounted inertial measurement unit (IMU) data, and wheel speed sensor information, the vehicle's distance relative to the target road section is continuously calculated with an accuracy of 0.1 meters. When the vehicle enters the dynamic calculation range from the target road section, the system triggers the data pre-loading instruction. At this time, the cloud data packet is transmitted to the vehicle local cache area through a secure encryption channel, containing three core road attributes: road surface material acoustic transmission coefficient, road curvature geometric parameters recording the radius and direction angle of the curve, and slope angle parameters recording the inclination change of the slope.

[0180] After the data loading is completed, the system starts the coordinate mapping engine. This engine first establishes a three-dimensional coordinate system for the road: taking the starting point of the target road section as the origin, the road forward direction as the X-axis, the vertical road upward as the Y-axis, and the horizontal right as the Z-axis. At the same time, a three-dimensional coordinate system for the vehicle is established: taking the front axle center as the origin, the vehicle head direction as the X-axis, pointing to the left side door as the Y-axis, and vertically upward as the Z-axis. The system aligns the two coordinate systems through a space transformation algorithm, calculates the conversion parameters of each feature point in the road coordinate system to the vehicle coordinate system, including position offset, direction rotation angle, and slope compensation coefficient. This process generates a dynamic coordinate conversion matrix, ensuring that all subsequent road feature data can be mapped to the current vehicle position in real time.

[0181] Multi-source fusion phase detailed process

[0182] The multi-source fusion phase is the core link of the system for real-time collection, alignment, and integration of environmental information. When the vehicle approaches the target road section about 500 meters, the system activates the forward perception sensor array. The 2 million pixel high-definition camera installed in the vehicle's front bumper area collects front road surface images at a speed of 30 frames per second, and the vision processing unit executes a multi-scale texture analysis algorithm to segment the road surface into several grid areas, calculates the edge density and texture contrast features of each area, and finally outputs a comprehensive roughness index between 0 and 1, with a larger value indicating a rougher road surface.

[0183] Meanwhile, the 128-line laser radar in the center of the vehicle grille starts a three-dimensional scan, emitting an array of laser pulses to probe the microscopic morphology of the road surface. The three-dimensional coordinates of each laser point in space are accurately recorded, and the system calculates the lateral and longitudinal height variation gradients of the road surface through point cloud processing algorithms to identify special terrain features such as speed bumps and potholes. These data are stored in vector form, containing height difference values and direction information.

[0184] At the same time as the sensor data is collected, the system initiates a space-time alignment program. The positioning synchronization module receives time stamp data from the GNSS receiver, IMU unit, and wheel speed meter, and unifies the time base of all sensors to nanosecond-level precision through the Precision Time Protocol. The space alignment engine then uses the coordinate conversion matrix generated during the preloading phase to convert the two-dimensional image feature points recognized by the camera, the three-dimensional point cloud data obtained by the laser radar, and the road feature information provided by the navigation map into the vehicle body coordinate system. This process specifically handles the small spatial offsets of different sensor data, such as accurately mapping the road features 10 meters ahead recognized by the camera to the expected wheel track position when the vehicle reaches that location.

[0185] The data integration stage adopts a hierarchical fusion strategy. The bottom layer fusion first combines the real-time data of the visual sensor and the laser radar: the weighted average of the roughness index of the vision and the height variation gradient of the laser generates a comprehensive road condition assessment. The middle layer fusion matches the real-time perception data with the preloaded road feature data, such as comparing the confidence level of the gravel texture recognized by the camera with the "gravel road section" label in the navigation data. The high-level fusion finally introduces the vehicle dynamic parameters: through the CAN bus, it obtains real-time vehicle speed, steering angle, acceleration, and other information to evaluate the influence weight of the current driving state on the acoustic environment.

[0186] The fusion process continuously conducts credibility evaluation, and the system marks data conflict areas (such as laser radar detecting potholes but navigation data showing flat road surface) and automatically selects the data source with high confidence as the dominant. Finally, a multi-dimensional fusion data package is output, containing spatially accurate road feature descriptions, environment-vehicle coupling influence coefficients, and data quality evaluation indicators.

[0187] Finally, the system performs pre-compensation parameter calculation. Based on the types of road features to be encountered, it generates an initial sound field compensation scheme in advance: for sharp turns, it calculates the sound image offset compensation; for rough roads, it estimates the low-frequency gain value; for steep slopes, it presets the voice frequency optimization parameters. These pre-calculated parameters are sent to the cache area to ensure that when the vehicle actually enters the target road section, the system can instantly retrieve the optimal compensation scheme.

[0188] Parameter generation stage detailed process:

[0189] The parameter generation stage converts the fused multi-source data into executable sound field control instructions. The system first processes the sound image stability parameter: based on the road curvature radius and real-time vehicle speed, the sound image offset caused by the centrifugal effect when the vehicle turns is calculated. This calculation takes into account the lateral offset of the vehicle's current trajectory from the road centerline and the instantaneous steering angle feedback from the steering system. For sharp turning conditions, the system additionally introduces steering angular velocity change rate as a dynamic compensation factor to prevent auditory discomfort caused by sudden changes in sound image. Finally, a sound image horizontal offset angle value is output, which is quantized as a left-right channel balance adjustment coefficient.

[0190] Next, the low-frequency response parameter is generated. The system analyzes the road surface material characteristics (such as asphalt or cement) and real-time roughness data to calculate the degree of influence of road surface vibration on in-vehicle low-frequency sound waves. This calculation considers three dimensions: the absorption characteristics of the material itself to sound waves, the vibration intensity determined by roughness, and the impact frequency reflected by the height change gradient. Real-time speed parameters are also fused, as higher speed increases the influence of road surface vibration on the acoustic environment. The output result is a low-frequency gain compensation coefficient, which will be directly applied to the low-frequency equalizer of the sound system.

[0191] The speech intelligibility optimization parameter generation module focuses on the impact of slope changes. The system dynamically adjusts the speech frequency band processing strategy based on the slope angle and vehicle acceleration state: when climbing uphill, the engine load increases, causing low-frequency noise to increase, so the system appropriately increases the mid-high frequency gain; when descending, wind noise may occur, so the system narrows the speech frequency band bandwidth to improve the signal-to-noise ratio. Combined with real-time vehicle cabin noise monitoring data, the center frequency of the speech frequency band is fine-tuned to ensure that the human voice is always in the best audible area.

[0192] Finally, all parameters are co-optimized. The system establishes a parameter coupling matrix to analyze the mutual influence relationship between each compensation item: for example, sound image offset compensation may affect low-frequency response, and slope optimization may change the sound field balance. After decoupling algorithm to eliminate parameter conflicts, the system applies physical constraint conditions: sound image offset is limited within ±30 degrees to prevent excessive compensation, and gain change is set to a safety threshold of ±6dB. All parameters are processed by a smoothing filter to ensure smooth transition without sudden changes.

[0193] The parameter generation module finally performs real-time optimization: the calculation tasks are distributed to different cores of the multi-core processor, and the sound image stability calculation is distributed to high-priority threads to ensure completion within 2ms, and the speech optimization is distributed to medium-priority threads. When the calculation is timed out in extreme cases, the system automatically enables the cache parameters of the last period to ensure continuity. The final generated parameter set contains a timestamp and a quality flag, which is transmitted to the audio output module through a high-speed bus.

[0194] The above description of the embodiments has been provided for the purpose of illustration and description. It is not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Individual elements or features of a particular embodiment are generally not limited to that particular embodiment, but are interchangeable with other embodiments in accordance with the disclosure, to the extent not doing so would render this disclosure unsupportable or unenforceable. In many aspects, the same element or feature can be changed or modified without departing from this disclosure. Such alterations are intended to be included within the scope of this disclosure, and all such modifications are intended to be within the scope of the disclosure.

[0195] The example embodiments are provided so that this disclosure will be thorough, and will fully convey the scope to those who are skilled in the art. Numerous specific details are set forth such as examples of specific parts, devices, and methods, to provide a thorough understanding of the embodiments of the present disclosure. It will be apparent to those skilled in the art that specific details need not be employed, that example embodiments can be practiced in many different

[0196] Herein, professional terms are used only for the purpose of describing particular example embodiments, and are not intended to be limiting. The singular forms "a," "an," and "the" used herein are intended to mean "one or more" unless the context clearly indicates otherwise. The terms "comprises," "comprising," "includes," and "including" are inclusive and therefore specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. The method steps, processes, and operations described herein are not to be construed as necessarily requiring their performance in the particular order in which they are described. It is also to be understood that additional or alternative steps can be employed.

Claims

1. A flat-panel diaphragm car audio system capable of achieving acoustic enhancement based on multi-source data fusion, comprising a flat-panel diaphragm speaker array and an audio processing module, characterized in that Include: A road acoustic property preloading unit is connected to the vehicle navigation system and configured to read road characteristic data pre-stored in the navigation map, wherein the road characteristic data includes an acoustic transmission coefficient of a road surface material, a road curvature geometric parameter, and a slope angle parameter; Real-time road surface perception unit: This unit includes a visual sensor and a 3D scanning sensor mounted on the front of the vehicle. It is configured to obtain microscopic road surface roughness characteristics through texture feature extraction algorithms and macroscopic road surface undulation characteristics through point cloud processing. A sound field dynamic compensation engine is configured to receive the road feature data and real-time road surface feature data, perform spatiotemporal synchronization fusion with the vehicle speed signal transmitted by the vehicle bus, and generate a dynamic sound field compensation parameter set based on the multi-source data coupling relationship, wherein the parameter set includes a sound image positioning horizontal offset, a low-frequency propagation gain compensation coefficient, and a voice band filter transfer function; Audio output control module: configured to perform frequency domain reshaping and spatial reconstruction on the input audio signal according to the dynamic sound field compensation parameter set, and drive the distributed flat-panel diaphragm speaker array to output an environment-adaptive sound field.

2. The system according to claim 1, characterized in that The road acoustic property preloading unit: Obtaining a road acoustic fingerprint data package from a cloud server via an Internet of Vehicles communication protocol, the data package containing a material-acoustic response mapping relationship constructed based on historical vehicle cluster collection data; When the vehicle reaches a preset distance threshold from the target road section, data preloading is initiated to establish a binding relationship between the road feature data and the vehicle coordinate system.

3. The system according to claim 1, characterized in that The real-time road surface perception unit: The visual sensor is configured to perform multi-scale texture analysis and output a roughness parameter related to the sound absorption characteristics of the road surface; The three-dimensional scanning sensor is configured to generate a road surface height change gradient vector by laser flight time difference calculation; The unit includes a data alignment module that converts sensor data into a vehicle body coordinate system through a coordinate transformation matrix.

4. The system according to claim 1, wherein The sound field dynamic compensation engine includes: Sound and image stabilization calculation module: Based on the road curvature radius R, vehicle speed v, and steering system state parameters, the sound field deflection model is used to calculate the horizontal offset compensation ΔX, where ΔX increases as the curvature radius R decreases and is proportional to the square of the vehicle speed v; Low-frequency response adaptation module: linearly combines the acoustic transmission coefficient μ of the pavement material with the real-time roughness parameter ρ to generate the low-frequency gain compensation coefficient G L , where G L It is positively correlated with the product of (μ·ρ); Speech clarity optimization module: Dynamically adjusts the quality factor Q of the bandpass filter according to the slope angle θ and vehicle acceleration status V , where Q V It increases with the absolute value of the slope.

5. The system according to claim 4, characterized in that The cornering sound and image stabilization module: When the steering angle velocity is detected to exceed the dynamic stability threshold, the enhanced anti-disturbance mode is activated: Introducing the differential term of the steering angular velocity ω into the horizontal offset compensation ΔX A smoothing window function is applied to the output sound field to suppress transient mutations.

6. The system according to claim 1, wherein The audio output control module: A frequency division processing architecture is used to decompose the audio signal into three independent sub-bands: The low frequency subband signal is given a gain compensation generated by the low frequency response adaptation module. The IF speech subband signal passes through the time-varying filter configured by the speech intelligibility optimization module The high-frequency sub-band signal remains in its original state; The sound field space synthesis is performed by applying an acoustic image shift matrix to the compensated sub-band signal, wherein the matrix includes a rotation component determined by the horizontal shift compensation amount ΔX.

7. The system according to claim 1, wherein Also includes a dynamic calibration mechanism: Acquire sound field response signals through the in-car microphone array; Calculate the residual vector between the actual sound pressure spectrum and the target compensation spectrum; When the Euclidean norm of the residual vector exceeds the adaptive threshold, the inversion parameter iterative update is triggered: H new(f) =H old(f) +η·FFT[e(t)]; Where η is the convergence factor, e(t) is the residual time domain signal, and the update delay is controlled within the system response period.

8. A vehicle-mounted active acoustic enhancement method based on the system according to claims 1-7, characterized in that The following steps are involved: (1) Preloading stage: According to the navigation path planning, the road feature data package is loaded before entering the target acoustic influence area, and the mapping relationship between the road coordinate system and the vehicle coordinate system is established; (2) Multi-source fusion stage: Receive road feature data, real-time road surface characteristics and vehicle speed signals through a time-space synchronization interface; Convert road curvature parameters to the vehicle motion coordinate system and fuse material characteristic parameters with real-time roughness; (3) Parameter generation stage: Calculate the parameters of the sound image offset matrix based on the curvature-vehicle speed coupling relationship; The low-frequency gain coefficient is determined by a linear combination of material properties and roughness; Adjust the voice filter bandwidth according to the slope angle and vehicle posture; (4) Sound field reconstruction stage: Decompose the audio signal into sub-bands and apply compensation parameters to each sub-band independently; Applying pan-and-pan offset matrices for multi-channel spatial rendering; (5) Dynamic optimization stage: Detect compensation deviation through residual spectrum analysis; When the deviation exceeds the adaptive tolerance, the inversion parameters are updated and injected into the processing pipeline.

9. The method according to claim 8, wherein In the sound field parameter generation stage (3): The calculation of the sound image offset matrix parameters adopts the kinematic model: k is the system calibration coefficient, φ is the vehicle yaw angle; When the steering angular velocity ω exceeds the threshold, ΔX is corrected to: The low-frequency gain coefficient is generated using a feature-weighted model: G L =a·m+b·r; α and β are material property weight factors.

10. The method according to claim 8, characterized in that In the dynamic optimization stage (5): Set dual trigger conditions to control parameter updates: a. When the residual energy in the speech band continues to exceed the background noise level by a set multiple, b. When the vehicle's motion state suddenly changes, causing the acoustic environment change index to exceed the critical value; The parameter updating process adopts the least mean square adaptive algorithm, and the convergence factor η decreases with the increase of vehicle speed.

Images