An adaptive audio control method and system for in-vehicle scenarios

Abstract

This invention relates to the field of audio control technology, and more particularly to an adaptive audio control method and system for in-vehicle scenarios. The method includes the following steps: acquiring in-vehicle audio signal streams based on an in-vehicle microphone array, performing dynamic range compression and frequency response balance optimization to construct an in-vehicle reference audio stream; analyzing the acoustic characteristics at multiple locations based on the in-vehicle reference audio stream, and performing acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the vehicle cabin; extracting multi-dimensional driving behavior data, performing dynamic driving preference modeling, and constructing a driving behavior profile; calculating audio sensitivity based on the driving behavior profile, and dynamically adaptively adjusting it based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters; applying safety optimization constraints and adaptive sound quality gain to the dynamic audio adjustment parameters, and outputting the final in-vehicle audio. This invention adjusts audio parameters in real time based on the vehicle scenario, improving driving safety and auditory experience.

Description

Technical Field

[0001] This invention relates to the field of audio control technology, and in particular to an adaptive audio control method and system for vehicle-mounted scenarios. Background Technology

[0002] Compared to traditional home or office environments, the in-vehicle environment is characterized by limited space, complex noise levels, diverse sound sources, and interference from driving tasks. In practical applications, the auditory experience of drivers and passengers is affected not only by physical factors such as road conditions, vehicle speed, and window opening / closing, but also by subjective factors such as driving behavior, passenger position, and individual preferences. These complex factors collectively dictate that in-vehicle audio systems must possess dynamic adjustment capabilities to provide optimal sound output in different situations. However, most current in-vehicle audio systems still employ static configurations or simple preset modes, failing to optimize in real time based on environmental changes and user status, thus struggling to meet increasingly complex interactive needs.

[0003] Traditional in-vehicle audio control methods primarily rely on equalizer presets, automatic volume control (AVL), and driver-related control. While these methods improve the audio experience to some extent, their adjustment strategies are mostly based on fixed rules, lacking the ability to comprehensively analyze and adaptively adjust multi-source information, and thus failing to achieve precise control based on user behavior and the acoustic environment. Furthermore, existing audio systems generally lack in-depth modeling of the cabin's spatial structure and the sound field characteristics at multiple locations, resulting in significant differences in the audio experience for occupants in different seats, impacting overall auditory comfort. In addition, the lack of in-depth learning of driver behavior and preferences also limits the personalized service capabilities of in-vehicle audio systems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an adaptive audio control method and system for in-vehicle scenarios, thereby resolving at least one of the aforementioned technical issues.

[0005] To achieve the above objectives, the present invention provides an adaptive audio control method for in-vehicle scenarios, comprising the following steps:

[0006] Step S1: Collect in-vehicle audio signal stream based on the vehicle microphone array, perform dynamic range compression and frequency response balance optimization, and construct in-vehicle reference audio stream;

[0007] Step S2: Analyze the acoustic characteristics at multiple locations based on the in-vehicle reference audio stream, and perform acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the vehicle cabin;

[0008] Step S3: Extract multi-dimensional driving behavior data, perform dynamic driving preference modeling, and construct a driving behavior profile;

[0009] Step S4: Calculate audio sensitivity based on driving behavior profile, and dynamically adapt and adjust based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters;

[0010] Step S5: Perform safety optimization constraints and adaptive sound quality gain on the dynamic audio adjustment parameters, and output the final in-vehicle audio.

[0011] This specification provides an adaptive audio control system for in-vehicle scenarios, used to execute the adaptive audio control method for in-vehicle scenarios as described above, including:

[0012] The audio acquisition module is used to acquire in-vehicle audio signal streams based on the vehicle microphone array, perform dynamic range compression and frequency response balance optimization, and construct an in-vehicle reference audio stream.

[0013] The cabin sound field module is used to analyze the acoustic characteristics of multiple locations based on the in-vehicle reference audio stream, and to perform acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the cabin.

[0014] The behavior profiling module is used to extract multi-dimensional driving behavior data, perform dynamic driving preference modeling, and construct driving behavior profiles.

[0015] The audio adjustment module is used to calculate audio sensitivity based on driving behavior profiles and dynamically and adaptively adjust based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters.

[0016] The audio quality gain module is used to perform safety optimization constraints and adaptive audio quality gain on dynamic audio adjustment parameters, and output the final in-vehicle audio.

[0017] The beneficial effects of this invention are as follows: Through an in-vehicle microphone array, it can collect real-world ambient sound and played audio from multiple points within the vehicle, achieving a comprehensive perception of the current sound field state. It effectively controls sudden volume spikes or large dynamic range differences, preventing road noise or sudden noise from masking important audio (such as navigation commands and voice dialogue), thus improving voice clarity. It corrects the uneven frequency response caused by speaker position and material reflection within the cabin, making the audio smoother and more natural, providing an accurate reference audio benchmark for subsequent modeling. Through multi-position acoustic characteristic analysis, combined with path calculations such as reflection, diffraction, and absorption, a three-dimensional model of sound propagation within the cabin is constructed, providing a spatial reference for sound direction control. It provides underlying support for personalized sound source localization (e.g., enhancing navigation audio in the driver's seat and weakening advertising audio in the passenger seat). Through three-dimensional modeling, differentiated audio output can be customized for each seat (e.g., playing fairytale audio in children's seats and maintaining clear broadcasts in the driver's seat). Different drivers exhibit variations in driving style (aggressive / smooth), emotional state (anxious / relaxed), and environmental adaptation. Behavioral data (such as steering frequency, acceleration / deceleration, and driving routes) can indirectly reflect their preferences for audio rhythm, volume, and frequency response. When drivers are fatigued or their attention is diminished, background audio can be adjusted to maintain their alertness (e.g., playing more rhythmic music). This forms a long-term learning loop between the audio preference model and behavioral characteristics, supporting subsequent AI learning and optimization. By combining driver preferences and real-time location sound field characteristics, volume, EQ, and reverberation parameters are dynamically adjusted to achieve an audio effect where "the sound moves with the driver in the car." Audio clarity is enhanced in noisy environments or when driver attention is low, while sound effects are softened in comfortable scenarios to avoid auditory fatigue. Dynamic adjustment allows for differentiated presentation of the same song for different drivers or in different driving situations, creating a "personalized" auditory experience. Volume peaks are limited to prevent excessively loud audio from interfering with driver attention; background noise is automatically reduced during important alert sounds (such as ADAS warnings and navigation directions) to ensure priority for alert sounds. It utilizes audio quality gain algorithms to compensate for audio detail loss (such as loss due to dynamic compression), improving low-frequency fullness and high-frequency resolution; it balances multi-source output (media audio, telephone audio, voice assistant audio) to avoid audio conflicts. It supports automatic optimization of sound effect parameters based on content type (music / telephone / navigation) to enhance the overall listening experience. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the steps of an adaptive audio control method for vehicle-mounted scenarios according to the present invention.

[0019] Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1.

[0020] Figure 3 This is a flowchart illustrating the detailed implementation steps of step S2. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0022] This application provides an adaptive audio control method and system for in-vehicle scenarios. The execution entities of the adaptive audio control method and system for in-vehicle scenarios include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be considered general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0023] Please see Figures 1 to 3 This invention provides an adaptive audio control method for in-vehicle scenarios, comprising the following steps:

[0024] Step S1: Collect in-vehicle audio signal stream based on the vehicle microphone array, perform dynamic range compression and frequency response balance optimization, and construct in-vehicle reference audio stream;

[0025] Step S2: Analyze the acoustic characteristics at multiple locations based on the in-vehicle reference audio stream, and perform acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the vehicle cabin;

[0026] Step S3: Extract multi-dimensional driving behavior data, perform dynamic driving preference modeling, and construct a driving behavior profile;

[0027] Step S4: Calculate audio sensitivity based on driving behavior profile, and dynamically adapt and adjust based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters;

[0028] Step S5: Perform safety optimization constraints and adaptive sound quality gain on the dynamic audio adjustment parameters, and output the final in-vehicle audio.

[0029] In the embodiments of the present invention, see Figure 1 This is a flowchart illustrating the steps of an adaptive audio control method for in-vehicle scenarios according to the present invention. In this example, the steps of the adaptive audio control method for in-vehicle scenarios include:

[0030] Step S1: Collect in-vehicle audio signal stream based on the vehicle microphone array, perform dynamic range compression and frequency response balance optimization, and construct in-vehicle reference audio stream;

[0031] In this embodiment, a multi-point distributed microphone array is used to collect full-frequency audio signals within the vehicle in real time. The microphone array employs eight omnidirectional condenser microphones, distributed in key locations such as the roof, dashboard, headrests, and doors. The sampling frequency is set to 48kHz, and the quantization precision is 24 bits to ensure high-fidelity audio signal acquisition. The acquired raw audio signal stream undergoes preprocessing and analysis. First, a high-pass filter eliminates low-frequency interference below 20Hz. Then, an adaptive noise cancellation algorithm is used to identify and suppress background noise such as engine noise and road friction noise, achieving a noise suppression depth of 25-30dB. Next, dynamic range compression processing is applied to the audio signal, controlling the signal dynamic range within 60dB. The compression ratio is set to 4:1 to 8:1, ensuring the audibility of weak signals while avoiding peak clipping distortion of strong signals. Finally, frequency response balance optimization was performed. A 31-band graphic equalizer was used to finely adjust the entire frequency band from 20Hz to 20kHz, with the gain range of each band controlled within ±12dB. This focused on compensating for the frequency response non-uniformity caused by the acoustic environment of the vehicle cabin, and constructing a standardized in-vehicle reference audio signal stream to provide a reliable data foundation for subsequent acoustic analysis.

[0032] Step S2: Analyze the acoustic characteristics at multiple locations based on the in-vehicle reference audio stream, and perform acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the vehicle cabin;

[0033] In this embodiment, a comprehensive analysis of the acoustic environment of the vehicle cabin is performed based on the in-vehicle reference audio signal stream. First, standard test signals (including pink noise, swept frequency signals, and pulse signals) are played at different locations within the cabin, and response signals from each listening position are simultaneously acquired using a microphone array. Impulse response measurements are performed on the acquired response signals, and the acoustic reflection coefficients of various surface materials within the cabin are calculated. Measurement results show that the reflection coefficient of the door trim panels is 0.15-0.25, the sound absorption coefficient of the seat fabric reaches 0.65-0.80, and the reflection coefficient of the roof material is approximately 0.35-0.45. The propagation path of sound waves within the cabin is analyzed using a time delay estimation algorithm, and the arrival time difference of early reflected sound is measured, with a typical value of 2-15 ms. The reverberation time RT60 is controlled within the range of 0.3-0.6 seconds. Based on the ray tracing principle, a hybrid modeling method combining geometric acoustics and wave acoustics is used to construct a three-dimensional sound field propagation model of the cabin, with an accuracy within ±2 dB. The acoustic simulation of the vehicle cabin geometry was carried out using the finite element method. The cabin was divided into more than 10,000 acoustic units, and the sound pressure level distribution and frequency response characteristics at each location were calculated. Finally, a three-dimensional sound field distribution database covering the entire frequency band of 20Hz-20kHz was established, providing an accurate acoustic environment reference for the spatialization of audio.

[0034] Step S3: Extract multi-dimensional driving behavior data, perform dynamic driving preference modeling, and construct a driving behavior profile;

[0035] In this embodiment, multi-dimensional driving behavior data is acquired in real time via the vehicle's CAN bus. The data acquisition frequency is set to 100Hz to ensure accurate capture of behavioral characteristics. Key monitoring parameters include steering wheel angular velocity (range ±540° / s), accelerator pedal opening change rate (0-100%, response time <50ms), brake pedal pressure (0-2000N), and vehicle longitudinal acceleration (±8m / s). ) and lateral acceleration (±12m / Key indicators such as driving behavior data were analyzed and pattern recognition was performed on the collected driving behavior data. A sliding window technique was used to extract behavioral features within a 30-second time window, calculating the driver's aggressiveness index, stability coefficient, and reaction sensitivity parameters. A personalized driver behavior model was established using statistical learning methods. Analysis results showed that aggressive drivers operated the steering wheel 40-60% more frequently than stable drivers, and their braking frequency differed by 2-3 times. A driving behavior profile database was constructed based on long-term data accumulation, including driving style classification (conservative, balanced, aggressive), attention concentration assessment (high, medium, low levels), and audio sensitivity preference (preference for low, medium, or high frequencies), among other multi-dimensional feature parameters. The profile accuracy rate exceeded 85%, providing a reliable behavioral basis for personalized audio adjustment.

[0036] Step S4: Calculate audio sensitivity based on driving behavior profile, and dynamically adapt and adjust based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters;

[0037] In this embodiment, the driver's audio sensitivity is quantitatively calculated and analyzed based on driving behavior profile data, establishing a mapping model between behavioral characteristics and audio preferences. For aggressive drivers, a higher sensitivity to mid-to-high frequency audio (2-8kHz) is detected, with a volume preference 3-6dB higher than the baseline value; conservative drivers are more sensitive to low-to-mid frequency audio (100Hz-2kHz), with a preferred volume 2-4dB lower than the baseline value. Audio spatialization processing parameters are calculated based on the vehicle cabin's three-dimensional sound field model, optimizing the sound field distribution for the driver's position. Precise sound localization is achieved through delay adjustment (range 0-15ms) and phase correction (±180°) of the eight speaker units. Audio parameters are dynamically adjusted according to real-time changes in driving behavior. When emergency braking or sharp turns are detected, the volume is automatically reduced by 5-10dB and mid-frequency clarity is enhanced to ensure driving safety. Through adaptive digital signal processing algorithms, combined with the driver's personalized characteristics and the vehicle cabin acoustic environment, the optimal audio gain allocation, frequency response curve adjustment, and stereo imaging localization parameters are calculated in real-time. The dynamic adjustment parameter update frequency is set to 10 times per second, and the response latency is controlled within 100ms. A complete set of dynamic audio adjustment parameters, including 31-band equalizer settings, multi-channel gain control, and spatial sound effect processing, is built to achieve a truly personalized in-vehicle audio experience.

[0038] Step S5: Perform safety optimization constraints and adaptive sound quality gain on the dynamic audio adjustment parameters, and output the final in-vehicle audio.

[0039] In this embodiment, multi-level safety optimization constraints are applied to the dynamic audio adjustment parameters. First, a volume limiting strategy based on vehicle speed is established. When the vehicle speed exceeds 80 km / h, the volume limit is automatically set to 70% of the baseline value, and further limited to 60% when the speed exceeds 120 km / h, ensuring that the driver can perceive ambient sounds in a timely manner while driving at high speeds. Complex driving scenarios, such as curves, slopes, and construction areas, are identified through GPS and camera fusion perception technology. In these scenarios, the low-frequency enhancement effect is automatically reduced by 2-4 dB, while the mid-to-high frequency clarity is improved by 3-5 dB, helping the driver maintain focus. An adaptive sound quality gain optimization algorithm is implemented, dynamically adjusting the signal-to-noise ratio of the audio signal according to the cabin noise level. When the ambient noise exceeds 65 dB, the intelligent volume compensation function is automatically activated, with the gain range controlled between 0-8 dB. The final audio rendering is performed through a multi-channel audio processing engine. The power distribution of the eight speaker units is adjusted in real time according to the optimized parameters, with the power range of a single speaker from 5-25 W and the total harmonic distortion controlled below 0.1%. An integrated real-time audio quality monitoring and feedback mechanism continuously monitors the output audio quality through a reference microphone, automatically fine-tuning parameters when distortion or abnormalities are detected. The final output in-vehicle audio signal achieves a frequency response flatness of ±3dB, stereo separation greater than 40dB, and a dynamic range exceeding 90dB, realizing high-fidelity, personalized, and safety-oriented intelligent in-vehicle audio control.

[0040] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0041] When vehicle audio playback is detected, the in-vehicle audio signal stream is collected based on the vehicle microphone array.

[0042] Spectrum analysis and external environmental noise identification are performed on the in-vehicle audio signal stream to generate external environmental noise sources, including road friction noise, wind noise, external vehicle noise and external environmental interference sources.

[0043] Calculate the noise frequency of the external environmental noise source and perform dynamic spatial distribution analysis to generate a dynamic noise distribution map;

[0044] Calculate the noise source intensity attenuation coefficient and frequency domain masking threshold based on the dynamic noise distribution map;

[0045] Adaptive hierarchical noise suppression is performed based on the noise source intensity attenuation coefficient and the frequency domain masking threshold to obtain a noise-suppressed audio stream.

[0046] Dynamic range compression and frequency response balance optimization are performed on the noise-suppressed audio stream to construct an in-vehicle reference audio stream.

[0047] In this embodiment, when vehicle audio playback is triggered, the audio signal stream is first acquired in real time through a microphone array deployed inside the vehicle. The microphone array typically consists of 4 to 8 high-sensitivity pickup units, distributed across the roof and front and rear seat areas to cover the entire cabin space and form a spatial sampling network. The acquisition process uses a 48kHz sampling rate and 24-bit quantization precision to ensure the capture of sound details. To avoid direct sound interference from the in-vehicle speakers, the array incorporates beamforming technology to filter the direction of direct sound, highlighting noise signals from outside the vehicle. The goal of this stage is to simultaneously acquire a mixed signal containing the played audio and ambient noise, while retaining sufficient spatial information for subsequent noise modeling and separation. Through multi-channel synchronous acquisition, direct and reflected sound can be effectively captured, forming an audio input stream with spatial characteristics, thus providing reliable data for subsequent analysis. The acquired audio signal stream enters the spectrum analysis stage, where the time-domain signal is decomposed into a time-frequency feature matrix using a short-time Fourier transform (STFT). The analysis window length was set to 1024 points, with a frame shift of 256 points, thus achieving a balance between temporal and frequency resolution. In the spectrogram, different external vehicle noises exhibit specific energy distributions: road friction noise is concentrated in the 200Hz to 1kHz range, wind noise is mainly concentrated in the 1kHz to 4kHz range, external vehicle noise covers the 200Hz to 5kHz range, while horn honking and other environmental interference may extend above 5kHz. By extracting spectrogram features using convolutional neural networks or deep neural networks, noise types can be classified and labeled, thereby generating labels for external environmental noise sources.

[0048] After noise identification, the noise frequency is further calculated, and a dynamic noise distribution map is constructed by combining the spatial positioning results. Using microphone array sound source localization technology, the noise direction is determined using a delay-sum algorithm or the generalized cross-correlation method (GCC-PHAT), with directional accuracy controlled within ±5°. The noise direction is mapped to its corresponding spectral energy and projected onto the three-dimensional spatial model of the vehicle compartment, thus obtaining the spatial distribution of the noise. This distribution map presents the noise intensity changes in different frequency bands and directions in a multi-dimensional form, with a refresh rate maintained above 20Hz to ensure real-time reflection of the dynamic fluctuations of the noise field under high-speed driving conditions. After generating the dynamic noise distribution map, the noise source needs to be modeled, and its intensity attenuation coefficient and frequency domain masking threshold need to be calculated. The calculation of the intensity attenuation coefficient comprehensively considers the distance from the sound source to the microphone, the sound absorption characteristics of the vehicle compartment materials, and the reflection of different frequency bands. For example, fabrics have strong absorption for frequencies below 500Hz, while glass surfaces reflect frequencies above 3kHz more strongly; therefore, the attenuation curves differ across frequency bands. Meanwhile, the frequency domain masking threshold is calculated based on the auditory masking effect of the human ear. In the spectral energy distribution, the human ear is particularly sensitive to the mid-frequency band from 500Hz to 2kHz, so the threshold in this region needs to be set more strictly to ensure that music or speech content can still be clearly perceived in noise.

[0049] Based on the noise attenuation coefficient and frequency domain masking threshold, the adaptive hierarchical noise suppression stage begins. This method employs differentiated suppression strategies according to the frequency distribution and spatial characteristics of different noise sources. Low-frequency road friction noise is suppressed using spectral subtraction to ensure that low-frequency energy is weakened without compromising overall sound quality; mid-frequency external vehicle noise is dynamically adjusted using Wiener filtering to maintain the clarity of speech and music content; high-frequency wind noise and environmental interference are weakened in real time by a deep learning enhancement model. Suppression parameters are dynamically adjusted according to vehicle speed, environment, and noise type to ensure optimal suppression effects in different scenarios. After noise suppression, the audio stream needs further dynamic range compression and frequency response balance optimization to construct an in-vehicle baseline audio stream. Dynamic range compression balances the energy of different volume components, keeping weak sounds clear in environmental noise while avoiding excessive emphasis on strong sounds. The compression ratio can be set to 3:1, the threshold to -20dBFS, the attack time to 10ms, and the release time to 100ms to ensure that the output sound is both natural and clear. Frequency response balance optimization compensates for the audio based on the acoustic characteristics of the vehicle cabin. In the low-frequency region, excessive amplification due to spatial standing waves is achieved, so reduction is applied around 80Hz; while in the high-frequency region, energy excess due to reflection is corrected in the 4kHz to 6kHz range. The frequency response curve is continuously adjusted through an adaptive filter to achieve a balanced and harmonious effect in the in-car environment, constructing a stable and comfortable reference audio stream.

[0050] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0051] Identify the original output audio parameters of the in-vehicle audio, calculate the audio parameter differences of the in-vehicle reference audio stream, and obtain the in-vehicle audio distortion parameters;

[0052] The in-vehicle audio distortion parameters include frequency response difference, dynamic range difference, and phase difference;

[0053] Reverse engineering analysis is performed based on the in-vehicle audio distortion parameters, and the acoustic reflection, absorption and scattering effects of various materials inside the vehicle cabin are identified to generate the propagation attenuation law of the vehicle cabin.

[0054] Based on the propagation attenuation law of the vehicle cabin, the acoustic characteristics at multiple locations are analyzed, and the acoustic transmission path distribution evolution is carried out to construct a three-dimensional sound field model of the vehicle cabin.

[0055] In this embodiment, the original output audio parameters are accurately identified. The original output audio parameters refer to the digital or analog audio signal directly output to the speakers by the vehicle's head unit or amplifier, characterized by features such as frequency response curves, dynamic range indices, and phase response characteristics. Simultaneously, the in-vehicle reference audio stream, acquired by a microphone array and processed by the pre-amplifier, serves as a reference signal. By comparing the differences between the original audio parameters and the in-vehicle reference stream, the distortion effect of the cabin acoustic environment on the audio signal can be quantified. In actual processing, amplitude-frequency response analysis is used to compare the two signals, obtaining the amplitude difference within the 20Hz to 20kHz frequency band; simultaneously, dynamic range measurement is used to calculate the compression or expansion of the signal at different sound pressure levels; and combined with phase response detection, the phase shift caused by spatial propagation is identified. After identifying and comparing the original output with the in-vehicle reference stream, the in-vehicle audio distortion parameters are further obtained. The distortion parameters mainly include three aspects: frequency response difference, dynamic range difference, and phase difference. Frequency response difference is obtained by calculating the energy difference between the original signal and the signal acquired inside the vehicle at the same frequency point. Commonly, this manifests as increased energy in the low-frequency range and attenuation in the high-frequency range. Dynamic range difference is obtained by measuring the amplitude changes of the two signals under the same input sound pressure level, indicating compression of the dynamic range. This typically manifests as weak frequencies being masked by ambient noise or strong frequencies being amplified by spatial reflections. Phase difference is calculated based on group delay or phase response curves to identify the phase shift of the signal at different frequency bands during propagation.

[0056] After obtaining the in-vehicle audio distortion parameters, the reverse engineering analysis phase begins. This step involves reverse-engineering the distortion parameters to identify the influence of various materials within the vehicle cabin on sound propagation. The interior structure of a vehicle cabin is typically composed of various materials such as glass, fabric, leather, plastic, and metal. Different materials exhibit varying reflection, absorption, and scattering characteristics at different frequency bands. Frequency response differences can be used to deduce the absorption rate of materials at specific frequency bands; phase differences can be used to infer the reflection path length and delay; and dynamic range differences reveal the contribution of different materials to sound energy attenuation and diffusion. For example, glass has strong reflection of mid-to-high frequencies (above 3kHz), fabric has strong absorption of low frequencies (<500Hz), and metal surfaces produce significant scattering effects on broadband sound. By synthesizing these differences, a database of material acoustic properties is established, and in-vehicle cabin propagation attenuation patterns are generated. Based on these patterns, further multi-location acoustic characteristic analysis is performed. By comparing microphone signals collected at different locations within the cabin (such as the driver's seat, front passenger seat, and rear seats), the acoustic response characteristics at different seating points can be obtained. The analysis includes the proportion of direct sound energy, early reflection distribution, reverberation time, and sound energy attenuation rates at various frequency bands. Subsequently, the sound propagation path in the vehicle cabin is modeled based on the distribution and evolution of the acoustic transmission path. Specifically, the cabin space is abstracted as a three-dimensional acoustic scene, divided into several mesh units, each assigned different material properties and attenuation parameters. Using ray tracing or finite element acoustic modeling methods, the propagation process of sound from the loudspeaker to different receiving points is simulated, generating a three-dimensional sound field model of the cabin. This model not only visually displays the sound distribution in the cabin but also provides precise spatial acoustic data support for subsequent audio optimization, active compensation, and adaptive control, thereby achieving comprehensive tuning and optimization of in-vehicle audio.

[0057] In this embodiment, step S3 includes the following steps:

[0058] Acquire multi-dimensional driving behavior data transmitted in real time via the vehicle's CAN bus, including vehicle speed change rate, acceleration vector, steering wheel angle frequency, brake pedal pressure, and accelerator pedal opening changes;

[0059] Time-series pattern mining is performed on the multi-dimensional driving behavior data to generate time-series driving patterns;

[0060] Based on time-series driving patterns, personalized driving behavior habits are mined, and the degree of aggressiveness, preference for stability, and concentration are quantified to generate personalized driving characteristic parameters.

[0061] Dynamic driving preference modeling is performed based on personalized driving feature parameters to construct a driving behavior profile.

[0062] In this embodiment, multi-dimensional driving behavior data is acquired in real time via the CAN (Controller Area Network) bus. This data includes key indicators such as vehicle speed change rate, acceleration vector, steering wheel angle frequency, brake pedal pressure, and accelerator pedal opening change. The vehicle speed change rate is obtained by differentiating the speed signal and can be used to reflect the acceleration or deceleration trend of the vehicle per unit time. The acceleration vector consists of three-dimensional acceleration components (longitudinal, lateral, and vertical), reflecting the dynamic behavior of the vehicle under straight-line driving, turning, and bumpy road conditions. The steering wheel angle frequency characterizes the driver's operation frequency and steering habits by statistically analyzing the number of steering wheel angle changes per unit time. Changes in brake pedal pressure and accelerator pedal opening directly characterize the driver's acceleration and deceleration control intensity and rhythm. The acquisition process requires a high sampling rate (e.g., above 100Hz) to capture instantaneous changes in driving behavior. Data filtering and denoising methods are used to remove electrical interference and abnormal signals, ensuring the accuracy and continuity of the data input. The time-series data is segmented, and a sliding window mechanism is used to extract driving feature segments within different time periods. Temporal features include speed fluctuation curves, acceleration vector change trends, steering wheel angle change frequency curves, and pedal control curves. Using methods such as cluster analysis, Hidden Markov Models (HMMs), or Long Short-Term Memory Networks (LSTMs), behavioral patterns under different driving conditions can be identified. For example, rapid acceleration accompanied by frequent steering wheel corrections can form an "aggressive overtaking" mode, while gradual speed changes accompanied by low-frequency steering and stable throttle constitute a "smooth cruising" mode. By calculating the frequency of occurrence, duration, and switching patterns of different modes, a complete set of temporal driving modes is generated.

[0063] Further personalized analysis and quantification of driving behavior habits are conducted. The analysis primarily includes driving aggression, stability preference, and attention concentration. Aggression can be quantified by statistically analyzing the frequency and amplitude of high-dynamic operations such as rapid acceleration, deceleration, and sharp turns. Stability preference is described by assessing speed fluctuations, steering wheel smoothness, and the stability of throttle opening changes. Attention concentration is assessed by combining operational continuity, delayed response, and sudden actions; frequent lags or compensatory actions during driver operation result in a lower concentration score. Finally, these indicators are transformed into a set of standardized driving characteristic parameters, such as using a range of 0 to 1 to represent the aggression index, stability index, and concentration index. These parameters clearly reflect the driver's individual habits and preferences, forming quantified personalized driving characteristics. Dynamic driving preference modeling is then performed to ultimately create a driving behavior profile. The modeling method employs multi-dimensional feature fusion and weighted analysis, mapping indicators such as aggression, stability preference, and attention concentration into a multi-dimensional feature space. Through a dynamic weight adjustment mechanism, the model can be updated in real time based on the driver's performance under different operating conditions. For example, the weight of the aggressiveness index increases when driving on highways, while the weight of the stability parameter increases under urban congestion conditions. The modeling results are presented in the form of a driving behavior profile, which includes the driver's operating style, behavioral tendencies, and their changing patterns. The driving behavior profile is not only a set of static indicators, but also a dynamic description that evolves with driving data. Combined with in-vehicle adaptive audio control, the driving behavior profile can be used to adjust music playback style, noise suppression strategies, or volume dynamic range to achieve adaptive matching between driver preferences and the vehicle's acoustic environment, thereby improving the overall driving experience.

[0064] In this embodiment, step S4 includes the following steps:

[0065] Stereo audio processing is performed based on the three-dimensional sound field model of the vehicle cabin, and the delay compensation and phase adjustment parameters of each speaker unit are calculated to generate stereo sound field audio parameters.

[0066] Based on the driver's behavior profile, the auditory comfort level of the driver is calculated, and an auditory comfort assessment value is generated.

[0067] The audio sensitivity requirement is calculated based on the auditory comfort assessment value to obtain the audio sensitivity coefficient;

[0068] Dynamic audio adjustment parameters are constructed by dynamically and adaptively adjusting the audio parameters of the stereo sound field based on the audio sensitivity coefficient.

[0069] The dynamic adaptive adjustment includes adjusting the audio frequency response curve, dynamic range, and spatial positioning accuracy.

[0070] In this embodiment, based on the actual positions of the speakers within the vehicle (such as the front doors, center console, rear side panels, and subwoofer), the input stereo signal is mapped onto the vehicle cabin model using an acoustic transfer function, calculating the delay, attenuation, and phase changes of the sound during spatial propagation. Subsequently, combining the reflection and absorption characteristics obtained from the modeling, delay compensation and phase adjustment are applied to different speaker units. For example, the signals from the rear speakers, which are farther from the driver's seat, require time delay compensation in advance to ensure that the sound waves received by the driver's ears are in phase with the sound waves from the front speakers, thereby avoiding stereo field imbalance. The phase adjustment parameters are calculated by analyzing the group delay offset of each frequency band to ensure that sound energy can be correctly superimposed across the entire frequency range, avoiding phase cancellation or overlap distortion. The final output stereo field audio parameters include the delay compensation value, phase correction coefficient, and sound energy distribution ratio for each speaker unit. These parameters together ensure a balanced, stable, and spatially accurate stereo field within the vehicle cabin. Auditory comfort is calculated in conjunction with driving conditions. Driving behavior profiles include characteristics such as the driver's level of aggression, preference for smoothness, and level of concentration. These characteristics directly influence the driver's auditory perception needs in different scenarios. For example, when a driver exhibits a high level of aggression, they are more receptive to audio with a strong rhythm and a wide dynamic range, while under smooth driving conditions, the driver prefers a balanced and comfortable audio environment. These behavioral characteristics are combined with real-time in-vehicle audio parameters, and an auditory comfort assessment value is calculated using a multi-dimensional weighted function. This assessment value is a quantitative index, typically set between 0 and 100, used to represent the driver's subjective comfort with audio output in the current environment. Factors considered in the calculation include volume level, frequency response curve smoothness, spatial positioning stability, and the degree to which the dynamic range matches the driving rhythm.

[0071] The audio sensitivity coefficient is calculated. This coefficient characterizes the driver's sensitivity to changes in audio output, i.e., the ability to adapt to frequency, volume, and spatial distribution at different comfort levels. The calculation method is based on a nonlinear mapping function, matching the comfort assessment value with the adjustment requirements of the audio parameters. When the comfort assessment value is low, the sensitivity coefficient is increased, allowing the audio output parameters to respond quickly and make larger adjustments; when the comfort assessment value is high, the sensitivity coefficient is decreased, requiring only fine-tuning to maintain a comfortable state. The sensitivity coefficient output can be divided into three parts: frequency response sensitivity, dynamic range sensitivity, and spatial positioning sensitivity. For example, frequency response sensitivity mainly affects the compensation amplitude for low and high frequencies; dynamic range sensitivity controls the degree of compression and expansion; and spatial positioning sensitivity adjusts the accuracy of the sound image in the vehicle interior. The stereo audio parameters are dynamically and adaptively adjusted to generate the final dynamic audio adjustment parameters. Adaptive adjustment mainly includes three aspects: First, adjusting the audio frequency response curve by correcting the energy distribution of low, mid, and high frequencies based on frequency response sensitivity to ensure clarity of music and speech even in noisy environments. Second, optimizing the dynamic range by controlling audio compression and expansion using a dynamic range sensitivity coefficient, ensuring volume changes match the rhythm of driving without causing excessive fatigue or distraction. Third, improving spatial positioning accuracy by adjusting speaker delay compensation and phase correction through spatial positioning sensitivity parameters to ensure accurate sound image positioning in the driver's seat, enhancing immersion and spatial awareness. In actual operation, these parameters are adjusted in real time at a millisecond refresh rate, keeping the entire audio output synchronized with driving behavior and the acoustic environment. The final dynamic audio adjustment parameters not only ensure natural and realistic sound but also achieve adaptive control of in-vehicle audio under different driving conditions, providing the driver with the best listening experience.

[0072] In this embodiment, step S5 includes the following steps:

[0073] The system collects images from the dashcam, performs adaptive volume constraint processing, and obtains the maximum audio output volume.

[0074] Based on the maximum audio output volume, the dynamic audio adjustment parameters are subjected to safety optimization constraints to obtain volume-constrained audio.

[0075] Multi-channel spatial rendering is performed on volume-constrained audio to obtain spatially optimized audio.

[0076] The spatial rendering optimizes the audio with adaptive sound quality gain and outputs the final in-vehicle audio.

[0077] In this embodiment, monitoring images from a dashcam are acquired and analyzed in conjunction with the driving environment to adaptively constrain the volume of the in-vehicle audio output. The dashcam is typically installed on the windshield, capturing real-time road scenes and lighting changes ahead. Image processing extracts risk indicators of the driving environment, such as traffic flow density, pedestrian or vehicle proximity, traffic light status, and sudden obstacles. Image recognition algorithms, such as object detection (YOLO, SSD, etc.) and optical flow analysis, are used to identify the complexity of the scene and calculate the driver's required attention load. In complex traffic scenarios or sudden risk situations, the maximum audio output volume is automatically reduced to avoid excessive volume interfering with the driver's auditory perception and concentration; while in stable road conditions, the maximum output volume can be moderately increased to ensure entertainment and comfort. After obtaining the maximum audio output volume, this constraint is applied to previously generated dynamic audio adjustment parameters for safety optimization. Dynamic audio adjustment parameters include frequency response curves, dynamic range, and spatial positioning compensation, while the maximum volume constraint provides a safety upper limit to ensure that the output does not exceed the driver's tolerance under any circumstances. The specific approach involves comparing the instantaneous energy peak in the dynamic audio stream with the maximum volume parameter. When the peak exceeds a set threshold, the gain coefficient is automatically adjusted or the dynamic range is reduced to avoid excessive loudness. To ensure the naturalness of the adjustment, a buffer and smooth transition curve are set during the gain correction process to avoid abruptness caused by sudden volume reduction. At the same time, the maximum volume constraint not only affects the overall signal energy but also distributes it to different channels, keeping the left and right channels and the front and rear channels balanced, and preventing excessive volume on one side from affecting the stability of the sound field.

[0078] The process then moves to multi-channel spatial rendering. A car cabin typically houses multiple speaker units, including front door speakers, center console speakers, rear speakers, and subwoofers. To create an immersive stereo sound field, volume-constrained audio is spatially rendered based on a 3D acoustic model. Specifically, this is achieved by using acoustic transfer functions and speaker position parameters to differentiate the audio signals in time and phase, allowing different channel signals to form precise spatial sound image localization within the cabin. For example, in the driver's seat, the voice signal can be located at the center console, while the accompaniment sound creates an surround effect through the front and rear channels, enhancing immersion. Simultaneously, to avoid sound image shifts caused by cabin reflections and absorption characteristics, delay compensation and energy distribution for each channel are dynamically adjusted to ensure a balanced sound field. The resulting spatially rendered optimized audio not only maintains the spatial feel of stereo but also possesses high fidelity and directional accuracy, allowing the driver to enjoy rich auditory layers while maintaining focus. Adaptive sound quality gain processing is then applied to further enhance sound clarity and detail. Audio quality enhancement encompasses three aspects: frequency response enhancement, harmonic correction, and dynamic range fine-tuning. Firstly, regarding frequency response, an adaptive equalization algorithm finely corrects low, mid, and high frequencies, ensuring full low frequencies in low-noise environments and highlighting mid and high frequencies in high-noise environments such as high-speed driving, thereby improving the audibility of speech and music. Secondly, for harmonic correction, non-linear processing compensates for energy loss caused by the vehicle's cabin structure, resulting in a fuller and more natural sound. Finally, regarding dynamic range, considering driver comfort needs, weak frequencies are moderately enhanced while controlling strong frequencies to avoid distortion. The entire gain process is adaptive, meaning parameters are adjusted in real-time based on vehicle speed, driving behavior profile, and ambient noise. The final output in-vehicle audio achieves a dynamic balance between safety, spatiality, and sound quality, providing the driver with an immersive and comfortable listening experience.

[0079] In this embodiment, the specific steps for acquiring images monitored by the dashcam, performing adaptive volume constraint processing, and obtaining the maximum audio output volume are as follows:

[0080] Collect images from the vehicle's dashcam;

[0081] Based on the images monitored by the dashcam, depth image recognition is performed, and environmental perception calculations are carried out to extract the road curvature coefficient and road surface smoothness.

[0082] Road type identification and traffic speed limit data extraction are performed based on images monitored by dashcams.

[0083] The road complexity is obtained by evaluating the road curvature coefficient and road surface smoothness.

[0084] Dynamic analysis of road scenarios is performed based on road complexity and traffic speed limit data to generate road scenario features;

[0085] Adaptive volume constraint processing is performed based on road scene features to obtain the maximum audio output volume.

[0086] In this embodiment, a dashcam continuously captures images of the road environment ahead. The dashcam is typically mounted above the windshield, equipped with a wide-angle lens and high-definition resolution (such as 1080P or 4K), capable of capturing road signs, lane lines, road conditions, and information about vehicles and pedestrians ahead. The frame rate is generally set above 30fps to ensure the ability to capture rapidly changing scenes, while incorporating high dynamic range (HDR) technology to adapt to both strong and low-light environments. During acquisition, image data is transmitted in real-time to the onboard computing unit via a CAN bus or dedicated data interface, and undergoes preliminary preprocessing, including distortion correction, brightness equalization, and noise suppression, thus providing a clear and stable input signal for subsequent image recognition and environmental perception calculations. A convolutional neural network (CNN) is used to identify lane lines, road surface texture, and curve morphology, thereby calculating the road's curvature coefficient. The curvature coefficient measures the degree of road curvature; a higher value indicates a sharper curve and a higher driving task complexity. Next, optical flow analysis and texture feature calculations are used to identify road surface smoothness. For example, the presence of continuous, fine undulations or irregular reflections indicates the presence of bumps or potholes in the road section. To ensure accuracy, a sliding window statistical method is used to fuse the feature results of multiple consecutive frames, reducing single-frame errors. The final output feature data includes two key parameters: the curvature coefficient of the road surface and the road surface smoothness, reflecting the road geometry and surface quality respectively, providing a quantitative basis for road complexity assessment.

[0087] Road type identification is primarily achieved through the detection of lane width, number of lanes, and roadside facilities. For example, highways typically have multiple lanes and are equipped with a central median; urban roads have narrower lanes and are accompanied by traffic lights and pedestrian crossings; rural roads may lack clear signage. To enhance robustness, semantic segmentation algorithms are used to divide the image into regions, thereby accurately distinguishing lanes, shoulders, and other ancillary facilities. For speed limit data extraction, object detection methods are used to identify road speed limit signs, and optical character recognition (OCR) technology is used to parse the speed limit values. If image recognition is incomplete, data from high-precision maps or in-vehicle navigation systems can be used for auxiliary correction. The complexity calculation method employs a weighted comprehensive model, mapping curvature coefficients and smoothness indices to a unified complexity metric. For example, roads with low curvature and good smoothness have lower complexity scores; while on sections with sharp curves and bumps, the complexity score is significantly higher. To ensure the precision of the evaluation, complexity is divided into multiple levels, such as "low complexity," "medium complexity," and "high complexity," and quantified using a continuous numerical range from 0 to 1. This complexity value directly reflects the operational difficulty and attentional load faced by the driver on that road section. In this way, complex road characteristics can be transformed into intuitive numerical indicators.

[0088] After acquiring road complexity and traffic speed limit data, these two data points are combined for dynamic road scene analysis. A rule engine and dynamic weight allocation method are used to fuse complexity indicators with speed limit information to generate road scene features. If the road complexity is high and the speed limit is low, the road segment is identified as a high-risk scenario requiring high driver concentration; conversely, if the complexity is low and the speed limit is high, the scene features lean towards a low-risk cruising state. The dynamic analysis process considers not only current parameters but also uses time-series prediction algorithms to predict road conditions over the next few seconds to tens of seconds. For example, in a series of sharp curves, a continuously increasing trend in complexity is identified, generating corresponding high-complexity scene features. The final output road scene features contain multi-dimensional information such as road type, risk level, and driver load, serving as a crucial bridge connecting the external environment and audio control strategies. By combining road scene features with driving safety requirements, the maximum audio output volume is calculated. In high-complexity scenarios (such as sharp curves, bumpy roads, or low-speed-limit areas), the maximum output volume is reduced to avoid distracting the driver and ensure they can fully perceive road noise and environmental cues. In low-complexity scenarios (such as straight highways), the maximum output volume is allowed to be moderately increased to enhance entertainment and comfort. The volume constraint calculation uses a piecewise function and a smooth adjustment algorithm to make volume changes natural and gradual, avoiding discomfort caused by sudden jumps. The final maximum audio output volume will be used as the control parameter for the in-vehicle audio system, combined with the previous dynamic audio adjustment module, to achieve true adaptive closed-loop control of "road scenario—driving behavior—audio output".

[0089] In this embodiment, an adaptive audio control system for in-vehicle scenarios is provided, for executing the adaptive audio control method for in-vehicle scenarios as described above, including:

[0090] The audio acquisition module is used to acquire in-vehicle audio signal streams based on the vehicle microphone array, perform dynamic range compression and frequency response balance optimization, and construct an in-vehicle reference audio stream.

[0091] The cabin sound field module is used to analyze the acoustic characteristics of multiple locations based on the in-vehicle reference audio stream, and to perform acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the cabin.

[0092] The behavior profiling module is used to extract multi-dimensional driving behavior data, perform dynamic driving preference modeling, and construct driving behavior profiles.

[0093] The audio adjustment module is used to calculate audio sensitivity based on driving behavior profiles and dynamically and adaptively adjust based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters.

[0094] The audio quality gain module is used to perform safety optimization constraints and adaptive audio quality gain on dynamic audio adjustment parameters, and output the final in-vehicle audio.

[0095] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application be incorporated into the invention.

[0096] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. An adaptive audio control method for in-vehicle scenarios, characterized in that, Includes the following steps: Step S1: Collect in-vehicle audio signal stream based on the vehicle microphone array, perform dynamic range compression and frequency response balance optimization, and construct in-vehicle reference audio stream; Step S2: Analyze the acoustic characteristics at multiple locations based on the in-vehicle reference audio stream, and perform acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the vehicle cabin. Step S3: Extract multi-dimensional driving behavior data, perform dynamic driving preference modeling, and construct a driving behavior profile; Step S4: Calculate audio sensitivity based on driving behavior profile, and dynamically adapt and adjust based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters; Step S5: Perform safety optimization constraints and adaptive sound quality gain on the dynamic audio adjustment parameters, and output the final in-vehicle audio. The specific steps of step S1 are as follows: When vehicle audio playback is detected, the in-vehicle audio signal stream is collected based on the vehicle microphone array. Spectrum analysis and external environmental noise identification are performed on the in-vehicle audio signal stream to generate external environmental noise sources, including road friction noise, wind noise, external vehicle noise and external environmental interference sources. Calculate the noise frequency of the external environmental noise source and perform dynamic spatial distribution analysis to generate a dynamic noise distribution map; Calculate the noise source intensity attenuation coefficient and frequency domain masking threshold based on the dynamic noise distribution map; Adaptive hierarchical noise suppression is performed based on the noise source intensity attenuation coefficient and the frequency domain masking threshold to obtain a noise-suppressed audio stream. Dynamic range compression and frequency response balance optimization are performed on the noise-suppressed audio stream to construct an in-vehicle reference audio stream.

2. The adaptive audio control method for vehicle-mounted scenarios according to claim 1, characterized in that, The specific steps of step S2 are as follows: Identify the original output audio parameters of the vehicle audio system, calculate the audio parameter differences of the in-vehicle reference audio stream, and obtain the in-vehicle audio distortion parameters. The in-vehicle audio distortion parameters include frequency response difference, dynamic range difference, and phase difference; Reverse engineering analysis is performed based on the in-vehicle audio distortion parameters, and the acoustic reflection, absorption and scattering effects of various materials inside the vehicle cabin are identified to generate the propagation attenuation law of the vehicle cabin. Based on the propagation attenuation law of the vehicle cabin, the acoustic characteristics at multiple locations are analyzed, and the acoustic transmission path distribution evolution is carried out to construct a three-dimensional sound field model of the vehicle cabin.

3. The adaptive audio control method for in-vehicle scenarios according to claim 1, characterized in that, Step S3 is as follows: Acquire multi-dimensional driving behavior data transmitted in real time via the vehicle's CAN bus, including vehicle speed change rate, acceleration vector, steering wheel angle frequency, brake pedal pressure, and accelerator pedal opening changes; Time-series pattern mining is performed on the multi-dimensional driving behavior data to generate time-series driving patterns; Based on time-series driving patterns, personalized driving behavior habits are mined, and the degree of aggressiveness, preference for stability, and concentration are quantified to generate personalized driving characteristic parameters. Dynamic driving preference modeling is performed based on personalized driving feature parameters to construct a driving behavior profile.

4. The adaptive audio control method for in-vehicle scenarios according to claim 1, characterized in that, The specific steps of step S4 are as follows: Stereo audio processing is performed based on the three-dimensional sound field model of the vehicle cabin, and the delay compensation and phase adjustment parameters of each speaker unit are calculated to generate stereo sound field audio parameters. Based on the driver's behavior profile, the auditory comfort level of the driver is calculated, and an auditory comfort assessment value is generated. The audio sensitivity requirement is calculated based on the auditory comfort assessment value to obtain the audio sensitivity coefficient; Dynamic audio adjustment parameters are constructed by dynamically and adaptively adjusting the audio parameters of the stereo sound field based on the audio sensitivity coefficient. The dynamic adaptive adjustment includes adjusting the audio frequency response curve, dynamic range, and spatial positioning accuracy.

5. The adaptive audio control method for vehicle-mounted scenarios according to claim 1, characterized in that, The specific steps of step S5 are as follows: The system collects images from the dashcam, performs adaptive volume constraint processing, and obtains the maximum audio output volume. The dynamic audio adjustment parameters are optimized and constrained for safety based on the maximum audio output volume to obtain volume-constrained audio. Multi-channel spatial rendering is performed on volume-constrained audio to obtain spatially optimized audio. The spatial rendering optimizes the audio with adaptive sound quality gain and outputs the final in-vehicle audio.

6. The adaptive audio control method for vehicle-mounted scenarios according to claim 5, characterized in that, The specific steps for acquiring images monitored by the dashcam, performing adaptive volume constraint processing, and obtaining the maximum audio output volume are as follows: Collect images from the vehicle's dashcam; Based on the images monitored by the dashcam, depth image recognition is performed, and environmental perception calculations are carried out to extract the road curvature coefficient and road surface smoothness. Road type identification and traffic speed limit data extraction are performed based on images monitored by dashcams. The road complexity is obtained by evaluating the road curvature coefficient and road surface smoothness. Dynamic analysis of road scenarios is performed based on road complexity and traffic speed limit data to generate road scenario features; Adaptive volume constraint processing is performed based on road scene features to obtain the maximum audio output volume.

7. An adaptive audio control system for vehicle-mounted applications, characterized in that, The method for performing adaptive audio control in a vehicle-mounted scenario as described in claim 1 includes: The audio acquisition module is used to acquire in-vehicle audio signal streams based on the vehicle microphone array, perform dynamic range compression and frequency response balance optimization, and construct an in-vehicle reference audio stream. The cabin sound field module is used to analyze the acoustic characteristics of multiple locations based on the in-vehicle reference audio stream, and to perform acoustic transmission path distribution evolution to construct a three-dimensional sound field model of the cabin. The behavior profiling module is used to extract multi-dimensional driving behavior data, perform dynamic driving preference modeling, and construct driving behavior profiles. The audio adjustment module is used to calculate audio sensitivity based on driving behavior profiles and dynamically and adaptively adjust based on the three-dimensional sound field model of the vehicle cabin to construct dynamic audio adjustment parameters. The audio quality gain module is used to perform safety optimization constraints and adaptive audio quality gain on dynamic audio adjustment parameters, and output the final in-vehicle audio.

Images