Intelligent cockpit acoustic prompt perception data processing method and system

Abstract

The application relates to the technical field of intelligent cockpit acoustic testing, and discloses an intelligent cockpit acoustic prompt perception data processing method and system, which comprises the following steps: acquiring real-vehicle cockpit ambient sound, vehicle operation, driving tasks, acoustic prompts and driver response data, synchronously segmenting and constructing a matching scene group of cross combinations of noise states and driving load states; reproducing the scene in a test environment, correcting parameters according to ear position frequency band sound pressure level error and driving load reproduction error; playing acoustic prompts according to multiple output gains, determining coupling influence and correcting subjective and objective mapping relationships according to index differences of cross scenes; determining the lower limit and the upper limit of the output gain according to the corrected mapping relationship, forming a safe gain range and evaluating the to-be-tested acoustic prompt. The application separates the coupling influence of noise and driving load, and improves the accuracy and repeatability of perception evaluation and gain threshold value determination in a composite scene.

Description

Technical Field

[0001] This invention relates to the field of smart cockpit acoustic testing technology, and more specifically, to a method and system for processing acoustic cues perception data in smart cockpits. Background Technology

[0002] As smart cockpit functions become increasingly sophisticated, acoustic cues are widely used to deliver collision warnings, vehicle status anomalies, and operational feedback to drivers. Whether a driver can perceive these acoustic cues promptly depends not only on the sound pressure level, frequency distribution, and duration of the cue tone, but also on the combined effects of ambient cabin noise and driver workload. In noisy environments with similar overall sound pressure levels but different spectral distributions, the degree of frequency band masking experienced by the cue tone may vary. Furthermore, even within the same noisy environment, differences in the complexity of the driving task can cause variations in cue recognition rates and response times.

[0003] Chinese invention patent application CN115019831A discloses an automated testing method and system for an intelligent cockpit voice interaction system. It records noise during vehicle driving, plays back the background noise in a laboratory using a multi-channel speaker and equalizer, and obtains wake-up rate, wake-up time, and speech recognition rate under test conditions with set background noise, output sound pressure, and playback intervals. While this method can reproduce in-vehicle noise fields under different operating conditions in a laboratory, it primarily tests the recognition performance of the voice interaction system and does not consider driving workload and driver subjective feelings to evaluate the perceptibility and interference level of acoustic cues. Chinese invention patent application CN116935891A discloses a method, design method, device, and electronic equipment for evaluating automotive cockpit sound. It determines the subjective evaluation dimensions of cockpit sound and the associated objective quantitative parameters, collects subjective evaluation data and objective quantitative data of sound samples, and establishes a correlation model between the two. For warning sounds, the above method uses urgency and annoyance as subjective evaluation dimensions, which can evaluate the cabin sound from a combination of subjective and objective perspectives. However, it does not construct a cross-test scenario that combines environmental noise and driving load, nor does it separate the perceptual changes caused by the coupling effect of the two.

[0004] Existing technologies involve both cabin noise playback and subjective / objective evaluation of cabin sound. However, in complex driving scenarios, it remains difficult to distinguish whether changes in perceived alerts are caused by changes in the noise spectrum, changes in driving load, or the coupling effect of the two. If a mapping relationship is directly established using objective acoustic indicators and driver response data from different scenarios, scenario reproduction bias and coupling factors may affect the evaluation results. Furthermore, existing methods often focus on whether alert sounds can be recognized, lacking a unified determination of the lower limit of perception and the upper limit of acoustic interference, making it difficult to establish a safe gain range applicable to different noise and driving load conditions. Summary of the Invention

[0005] In view of this, the present invention proposes a method and system for processing acoustic cues perception data in an intelligent cockpit, aiming to improve the repeatability of composite scenarios and the reliability of acoustic cues evaluation results.

[0006] To achieve the above objectives, the present invention provides the following solution: On one hand, the present invention provides a method for processing acoustic cues perception data in an intelligent cockpit, comprising: Acquire cabin ambient sound data, vehicle operation data, driving task data, acoustic prompt data, and driver response data from real-person vehicle tests, and simultaneously segment and extract noise spectrum characteristics and driving load characteristics; Based on the noise spectrum characteristics and driving load characteristics, a set of matching scenarios is constructed that includes cross-combinations of different noise states and different driving load states; The matching scenario group was reproduced in the test environment. The noise playback parameters were corrected according to the driver's ear position noise reproduction error, and the driving task parameters were corrected according to the driving load reproduction error to obtain an effective test scenario. In each effective test scenario, acoustic prompts are played according to multiple output gains, objective acoustic indicators are measured and calculated, driver response data is collected and perceived response indicators are calculated, the coupling effect of noise and driving load is determined according to the difference between the indicators between the cross combinations, and the mapping relationship between objective acoustic indicators and perceived response indicators is corrected using the coupling effect. Based on the corrected mapping relationship, the lower limit of the output gain that meets the requirements of prompt recognition rate and response time, and the upper limit of the output gain that meets the requirements of acoustic interference scoring are determined to form the scene safety gain range. The evaluation result is generated based on the comparison between the output gain of the acoustic prompt under test and the scene safety gain range.

[0007] Furthermore, the synchronization segmentation includes: The cabin ambient sound data, vehicle operation data, and driving task data are aligned according to a unified time scale; the changes in sound pressure level and driving load characteristics in each frequency band are calculated within a continuous time window, and the moment when any change exceeds the corresponding change threshold is determined as a candidate change point; candidate change points with a time interval less than a preset interval are merged, and the data is divided according to the merged candidate change points; the acoustic prompt data and driver response data are associated with the corresponding segments according to the output time of the acoustic prompt and the driver response time.

[0008] Furthermore, the noise spectrum characteristics include sound pressure level in each frequency band, sound pressure fluctuation in each frequency band, and dominant noise frequency band; the driving load characteristics include target speed change frequency, road curvature change frequency, number of specified operations, frequency of occurrence of secondary tasks, and duration of tasks; the noise spectrum characteristics and driving load characteristics are respectively processed into dimensionless values, and the noise spectrum characteristic distance and driving load characteristic distance between different segments are calculated based on the dimensionless values.

[0009] Furthermore, the matching scenario group includes at least: The first scenario formed by the combination of the first noise state and the first driving load state, the second scenario formed by the combination of the second noise state and the first driving load state, the third scenario formed by the combination of the first noise state and the second driving load state, and the fourth scenario formed by the combination of the second noise state and the second driving load state. Among them, the noise spectrum feature distance between scenes corresponding to the same noise state is less than the first matching threshold, and the driving load feature distance between scenes corresponding to the same driving load state is less than the second matching threshold; the noise spectrum feature distance between the first noise state and the second noise state reaches the first distinguishing threshold, and the driving load feature distance between the first driving load state and the second driving load state reaches the second distinguishing threshold. The first matching threshold is the maximum value of the noise spectrum feature distance under the same noise state, and the second matching threshold is the maximum value of the driving load feature distance under the same driving load state; the first distinguishing threshold is the minimum value of the noise spectrum feature distance between different noise states, and the second distinguishing threshold is the minimum value of the driving load feature distance between different driving load states. The first noise state is when the total sound pressure level at the driver's ear is in the first sound pressure level range and the dominant noise frequency band is in the preset frequency band range; the second noise state is when the total sound pressure level at the driver's ear is in the higher second sound pressure level range and the dominant noise frequency band is in the preset frequency band range. The first driving load state is a state in which the target vehicle speed change frequency and the secondary task occurrence frequency are both in the first frequency range, and the second driving load state is a state in which the target vehicle speed change frequency and the secondary task occurrence frequency are both in the higher second frequency range.

[0010] Furthermore, reproducing the matching scenario group includes: The cabin ambient sound is played back through a multi-channel speaker, and a driving task is loaded through a driving simulation device. The driver's ear position noise reproduction error includes the sound pressure level error and frequency band sound pressure fluctuation error between the real vehicle scene and the reproduced scene. The frequency band gain of the corresponding playback channel is corrected according to the direction of the driver's ear position noise reproduction error. The driving load reproduction error includes the reproduction error of the target vehicle speed change frequency, road curvature change frequency, specified number of operations, frequency of occurrence of sub-tasks, and duration of tasks. The driving task parameters are corrected according to the driving load reproduction error. The correction is repeated until both types of reproduction errors are less than or equal to the corresponding reproduction thresholds, and the reproduction scene that meets the corresponding reproduction threshold is determined as the valid test scene.

[0011] Furthermore, the objective acoustic indicators include the difference between the sound pressure level of each frequency band of the acoustic cues at the driver's ear position and the corresponding effective masking threshold, the proportion of effective cues frequency bands, loudness, duration, and modulation depth; the perceptual response indicators include cues recognition rate, response time, clarity score, and acoustic interference score. Using a scenario with the same driving task and output gain, and a noise sound pressure level lower than a preset value as a quiet reference scenario, calculate the changes in objective acoustic indicators and perceptual response indicators of each effective test scenario relative to the quiet reference scenario, and use the changes to determine the coupling effect. Retain the original objective acoustic indicators and original perceptual response indicators of each effective test scenario, and use the coupling effect to correct the mapping relationship established between the original objective acoustic indicators and the original perceptual response indicators.

[0012] Further, determining the coupling effect includes: Under the first driving load condition, calculate the change value of the first index under the second noise condition relative to the first noise condition; under the second driving load condition, calculate the change value of the second index under the second noise condition relative to the first noise condition; and determine the difference between the change value of the second index and the change value of the first index as the coupling influence of the corresponding index. The original objective acoustic indicators of the four cross-scenes are input into the basic mapping relationship to obtain the predicted perception response indicators of the four cross-scenes. The predicted coupling influence is determined based on the difference between the predicted perception response indicators and the cross-scenes combinations. The difference between the measured coupling influence and the predicted coupling influence of the perception response indicators is determined as the interaction correction term. The interaction correction term is added to the output of the basic mapping relationship to obtain the corrected mapping relationship.

[0013] Furthermore, the method also includes: Based on the mapping relationship corrected by the interaction correction term, the mapping direction, mapping residual and repeated test dispersion between each objective acoustic index and the prompt recognition rate, response time and acoustic interference score are determined respectively. Objective acoustic indicators that have a preset mapping direction with both prompt recognition rate and response time are identified as candidate perception indicators, and objective acoustic indicators that have a preset mapping direction with acoustic interference score are identified as candidate interference indicators. The mapping residuals and repeated test dispersions of each candidate indicator are normalized and weighted, and the candidate perception indicator with the smallest weighted result is selected as the core perception indicator, and the candidate interference indicator with the smallest weighted result is selected as the core interference indicator.

[0014] Furthermore, determining the security gain range for the scenario includes: Multiple output gains are set within a preset initial gain test range. Based on the core perception index and the corresponding corrected mapping relationship, the original prompt recognition rate and original response time under each output gain are predicted, and the minimum output gain that meets the requirements for both prompt recognition rate and response time is determined. Based on the core interference index and the corresponding corrected mapping relationship, the original acoustic interference score under each output gain is predicted, and the maximum output gain that meets the requirements for acoustic interference score is determined. Insert a new output gain between adjacent output gains and repeat the test until the difference between adjacent output gains is less than the gain resolution. The minimum output gain obtained is taken as the lower limit of the output gain, and the maximum output gain obtained is taken as the upper limit of the output gain. When the lower limit of the output gain is less than or equal to the upper limit of the output gain, the interval between the two is determined as the scene safety gain range.

[0015] On the other hand, the present invention provides an intelligent cockpit acoustic cues perception data processing system, comprising: The data acquisition module is used to collect cabin ambient sound data, vehicle operation data, driving task data, and acoustic prompt data during real-person and real-vehicle testing. The scene processing module is used to synchronously segment the data, extract noise spectrum features and driving load features, and construct matching scene groups that include cross combinations of different noise states and different driving load states. A noise playback module is used to reproduce the cabin ambient sound in the matched scene group through a multi-channel speaker; The driving task loading module is used to reproduce the driving tasks in the matching scenario group through a driving simulation device; The prompt output module is used to play acoustic prompts according to multiple output gains; The response acquisition module is used to collect driver response data; The data processing module is used to perform reproduction parameter correction based on the driver's ear position noise reproduction error and driving load reproduction error, calculate the coupling effect of noise and driving load based on the index difference between the objective acoustic index and the perceived response index in the cross combination, correct the mapping relationship between the objective acoustic index and the perceived response index using the coupling effect, determine the scene safety gain range based on the corrected mapping relationship, and generate the evaluation result of the acoustic prompt to be tested.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By combining real vehicle test data into matching scenario groups with different noise and driving load states, and determining the coupling influence based on the index difference between the cross scenarios, it is possible to distinguish the impact of noise changes, driving load changes, and their coupling effect on the prompt perception results, thereby reducing the interference of scenario factors on the subjective-objective mapping relationship.

[0017] 2. The noise playback parameters are corrected based on the sound pressure level error in the driver's ear position frequency band, and the driving task parameters are corrected based on the driving load reproduction error formed by the target vehicle speed change frequency, road curvature change frequency, specified number of operations and frequency of occurrence of the sub-task, so that the noise state and driving load state meet the reproduction conditions and reduce the impact of reproduction scene deviation on the test results.

[0018] 3. The mapping relationship between objective acoustic indicators and perceptual response indicators is corrected by using the coupling influence quantity, and perceptual core indicators and interference core indicators are screened separately so that the evaluation indicators can not only characterize the prompt recognition and response, but also reflect the acoustic interference caused by the prompt sound.

[0019] 4. Using the acoustic prompt output gain as a unified adjustment parameter, the lower limit of the output gain that meets the perception condition and the upper limit of the output gain that meets the acoustic interference condition are searched separately to form a scene safety gain range. This can avoid the problem of excessive interference of the prompt sound caused by only setting the perceptible lower limit, and provide a quantitative basis for the acoustic prompt test evaluation under different noise conditions and driving load conditions. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of a smart cockpit acoustic cues perception data processing method provided in an embodiment of the present invention; Figure 2This is a functional framework diagram of an intelligent cockpit acoustic prompting perception data processing system provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] The embodiments of this application use a real vehicle with an intelligent cockpit, a noise test environment, and driving simulation equipment to complete data acquisition and reproduction tests.

[0025] See Figure 1 As shown in the embodiment of this application, a smart cockpit acoustic cues perception data processing method specifically includes: S100: Acquires cabin ambient sound data, vehicle operation data, driving task data, acoustic prompt data, and driver response data from real-person vehicle tests. Aligns various types of data according to a unified time scale, synchronously segments the aligned data, and extracts noise spectrum characteristics and driving load characteristics.

[0026] S200: Based on the noise spectrum characteristics and driving load characteristics, segmented data are grouped into matching scenario groups that include cross-combinations of different noise states and different driving load states.

[0027] S300: Reproduce the matching scenario group in the noise test environment, correct the noise playback parameters according to the driver's ear position noise reproduction error, correct the driving task parameters according to the driving load reproduction error, and determine the scenario that meets the corresponding reproduction conditions for both types of reproduction errors as the valid test scenario.

[0028] S400: Play acoustic cues according to multiple output gains in each valid test scenario, measure and calculate objective acoustic indicators, collect driver response data and calculate perceived response indicators; perform baseline correction on objective acoustic indicators and perceived response indicators, determine the coupling effect of noise and driving load based on the indicator difference between each cross combination within the matching scenario group, and use the coupling effect to correct the mapping relationship between objective acoustic indicators and perceived response indicators.

[0029] S500: Based on the corrected mapping relationship, filter the core perception indicators and core interference indicators, determine the lower limit of output gain that meets the requirements of prompt recognition rate and response time, and the upper limit of output gain that meets the requirements of acoustic interference scoring; when the lower limit of output gain is less than or equal to the upper limit of output gain, form the scene safety gain range, and generate the evaluation result based on the comparison result between the output gain of the acoustic prompt to be tested and the scene safety gain range.

[0030] The data acquisition module includes driver ear microphones, vehicle bus data acquisition equipment, driving task recording equipment, and acoustic prompt recording interface. The driver ear microphones are positioned near the driver's left and right ears, with the distance between the microphone and the ear canal entrance controlled within 5cm. They are used to collect ambient cabin sound and acoustic signals after acoustic prompts reach the driver's ears. The microphones are calibrated using a sound level calibrator before testing, with an acoustic signal sampling frequency of 48kHz and a quantization bit depth of 24 bits.

[0031] The onboard bus data acquisition equipment connects to the vehicle bus and acquires vehicle speed, steering wheel angle, accelerator pedal opening, brake pedal opening, and yaw rate at a sampling frequency of no less than 50Hz. The driving task recording equipment records target speed change commands, steering tasks, prescribed operations, sub-tasks, and the start and end times of each task. The acoustic prompt recording interface records the prompt type, prompt output time, reference sound signal, output gain, duration, and repetition interval.

[0032] Driver response data is collected via steering wheel buttons, touch input devices, or voice input devices. The start time of the response timing is the moment the acoustic prompt begins to be output, and the end time is the moment the driver makes the first valid confirmation operation. If no valid confirmation operation is detected within 5 seconds after the prompt is output, it is recorded as no response; if the confirmation result is inconsistent with the prompt type, it is recorded as a misidentification.

[0033] The noise test environment is equipped with no fewer than four playback channels. The front left and right channels are used to play back powertrain noise and wind noise, while the rear left and right channels are used to play back road noise and rear ambient noise. Each playback channel is equipped with independently adjustable bandwidth gain. The driving simulation equipment is used to load target vehicle speed changes, road curvature changes, prescribed operations, and sub-tasks.

[0034] All data acquisition devices are connected to a unified clock. Devices that cannot be directly connected to a unified clock receive a synchronization trigger signal at the beginning and end of each test, and correct the device clock deviation based on the synchronization trigger signal to ensure that the time alignment error of various data is less than 10ms.

[0035] It should be noted that the "bandwidth gain" of the noise playback channel is used to correct for ambient sound reproduction errors; the "output gain" of the acoustic cue is the digital gain applied to the reference cue signal, which is used to change the output intensity of the acoustic cue. The two are different control parameters.

[0036] In real-world vehicle testing, the vehicle is driven along a pre-defined route, and the same type of acoustic cues are output under different road conditions, speeds, and driving tasks. Each test condition is repeated at least three times.

[0037] The cabin ambient sound data was analyzed using one-third octave bands, with an analysis frequency range of 200Hz to 8000Hz. The continuous sound signal was divided into analysis windows of 1 second length with 50% overlap between adjacent time windows. The equivalent sound pressure level and frequency band sound pressure fluctuation were calculated for each frequency band within each analysis window. In this embodiment, the standard deviation of the frequency band sound pressure level within the analysis window was used as the frequency band sound pressure fluctuation.

[0038] For adjacent analysis windows, the absolute value of the sound pressure level difference in each frequency band is calculated, and the maximum value among the differences in each frequency band is taken as the sound pressure level change at the current moment. The change in driving load characteristics is determined based on the changes in the target vehicle speed change frequency, road curvature change frequency, prescribed number of operations, frequency of occurrence of sub-tasks, and duration of tasks within adjacent analysis windows.

[0039] When the sound pressure level change exceeds 3 dB, or the change in any driving load characteristic exceeds the corresponding change threshold, the start time of the corresponding analysis window is determined as a candidate change point. When the time interval between adjacent candidate change points is less than 2 seconds, the adjacent candidate change points are merged, and the candidate change point with the larger change amplitude is retained. When the duration of the data segment defined by adjacent change points is less than 10 seconds, the data segment is merged with the adjacent data segments with smaller feature distances.

[0040] The change threshold can also be determined by repeatedly collecting samples under stable driving conditions. Specifically, the 95th percentile value of the change in adjacent time windows of each feature under stable driving conditions is calculated, and this percentile value is used as the corresponding change threshold to reduce missegmentation caused by measurement fluctuations.

[0041] After segmentation, the acoustic cues are associated with the data segment containing the output time based on the output time of the acoustic cues; similarly, the driver response data is associated with the data segment containing the corresponding acoustic cues based on the driver response time. When a data segment contains multiple acoustic cues, a separate association record is established between each acoustic cue and the driver response data.

[0042] In the extraction of noise spectrum characteristics and driving load characteristics, for each segment, the following noise spectrum characteristics are calculated: the equivalent sound pressure level of each one-third octave band; the standard deviation of the sound pressure level of each band over the duration of the segment, as the band sound pressure fluctuation; and the three bands with the largest sound pressure levels, as the dominant noise bands.

[0043] For each segment, driving load characteristics are calculated based on a uniform statistical duration of one minute. Specifically, the target speed change frequency is the number of times the target speed changes effectively; the road curvature change frequency is the number of times the road curvature crosses the boundary of a set interval; the prescribed operation count is the number of times the driver performs lane changes, steering, acceleration / deceleration, or in-vehicle control operations according to task requirements; the sub-task occurrence frequency is the number of times information confirmation or task selection occurs during driving; and the task duration is the cumulative duration during which the driving task is effectively executed.

[0044] Each feature is normalized using range. For positive features, the difference between the feature value and the minimum sample value is divided by the difference between the maximum and minimum sample values; if the maximum and minimum sample values ​​are the same, the normalization result is set to zero. Feature values ​​outside the sample range are truncated to the range of zero to one.

[0045] The noise spectrum characteristic distance between two segments is calculated using a weighted average of the normalized sound pressure level of each frequency band, the sound pressure fluctuation of each frequency band, and the frequency band consistency of the dominant noise. In this embodiment, the weight of the sound pressure level of each frequency band is 0.6, the weight of the sound pressure fluctuation of each frequency band is 0.3, and the weight of the frequency band consistency of the dominant noise is 0.1.

[0046] The driving load characteristic distance between two segments is calculated using the square root of the sum of the squares of the normalized differences of each driving load characteristic. The weights for the target speed change frequency, road curvature change frequency, prescribed number of operations, sub-task occurrence frequency, and task duration can each be set to 0.2. In this embodiment, the weights for the target speed change frequency, road curvature change frequency, prescribed number of operations, sub-task occurrence frequency, and task duration are all set to 0.2. In other embodiments, the corresponding weights can be determined based on the test repeatability of each driving task parameter. These weights are determined before the start of the same batch of tests and are not adjusted based on driver response data.

[0047] Based on the noise spectrum characteristics and driving load characteristics, the actual vehicle segment is divided into a first noise state, a second noise state, a first driving load state, and a second driving load state.

[0048] In this embodiment, the first matching threshold is the maximum value of the noise spectrum feature distance under the same noise state, and the second matching threshold is the maximum value of the driving load feature distance under the same driving load state; the first distinguishing threshold is the minimum value of the noise spectrum feature distance between different noise states, and the second distinguishing threshold is the minimum value of the driving load feature distance between different driving load states. The first noise state is when the total sound pressure level at the driver's ear is in the first sound pressure level range and the dominant noise frequency band is in the preset frequency band range; the second noise state is when the total sound pressure level at the driver's ear is in the higher second sound pressure level range and the dominant noise frequency band is in the preset frequency band range. The first driving load state is when both the target vehicle speed change frequency and the secondary task occurrence frequency are in the first frequency range, and the second driving load state is when both the target vehicle speed change frequency and the secondary task occurrence frequency are in the higher second frequency range.

[0049] In this embodiment, the first noise state is when the total sound pressure level at the driver's ear is between 55dB(A) and 62dB(A), and the dominant noise frequency band is between 315Hz and 1000Hz; the second noise state is when the total sound pressure level at the driver's ear is between 68dB(A) and 75dB(A), and the dominant noise frequency band is between 315Hz and 1000Hz.

[0050] Under the first driving load condition, the target vehicle speed changes at a frequency of less than or equal to 2 times per minute, and no secondary tasks are set; under the second driving load condition, the target vehicle speed changes at a frequency of 4 to 6 times per minute, and the secondary tasks occur at a frequency of 3 to 5 times per minute, and include the specified steering or information confirmation operations.

[0051] Filter the following four scenarios from the segmented data: Scenario 1: First noise level and first driving load level.

[0052] Second scenario: Second noise state and first driving load state.

[0053] Third scenario: First noise state and second driving load state.

[0054] Fourth scenario: Second noise state and second driving load state.

[0055] Between scenarios belonging to the same noise state, the noise spectrum characteristic distance should be less than a first matching threshold; between scenarios belonging to the same driving load state, the driving load characteristic distance should be less than a second matching threshold. The noise spectrum characteristic distance between the first noise state and the second noise state should reach a first distinguishing threshold, and the driving load characteristic distance between the first driving load state and the second driving load state should reach a second distinguishing threshold.

[0056] In this embodiment, both the first matching threshold and the second matching threshold are set to 0.15, and both the first discrimination threshold and the second discrimination threshold are set to 0.35. These thresholds can also be determined based on the feature distance distribution of repeatedly collected samples under the same test conditions: the 95th percentile of the feature distance under the same conditions is used as the matching threshold, and the 5th percentile of the feature distance between different conditions is used as the discrimination threshold.

[0057] Within the same matching scenario group, the vehicle model, driver position, acoustic cue type, reference sound signal, cue duration, and cue output position must remain consistent. If a complete four-scenario combination cannot be formed, the segment with the second smallest feature distance is first selected from the candidate segments; if the matching threshold is still not met, the missing combination is re-acquired, and the coupling influence is not calculated using the incomplete scenario group.

[0058] The cabin ambient sound from the four scenarios was input into a multi-channel loudspeaker. Before playback, the transmission characteristics from each playback channel to the driver's ear were measured using two microphones located at the driver's ear, and frequency band equalization was used to compensate for the response differences between the loudspeakers and the test space.

[0059] For each reproduced scene, continuous playback should last no less than 30 seconds. Calculate the sound pressure level error and frequency band sound pressure fluctuation error for each frequency band in both the real vehicle scene and the reproduced scene. Calculate the errors for the left and right ears separately, and take the larger absolute value as the error for the corresponding frequency band.

[0060] When the sound pressure level of a certain frequency band in the reproduced scene is lower than that in the actual vehicle scene, increase the gain of the playback channel that contributes the most to that frequency band; when it is higher than that in the actual vehicle scene, decrease the gain of the corresponding frequency band. The initial correction step size is half the absolute value of the error and is limited to 0.5dB to 2dB; if the error direction changes between two adjacent corrections, the correction step size is halved. The cumulative frequency band gain correction for a single channel is limited to ±12dB.

[0061] When the frequency band sound pressure fluctuation error exceeds the threshold, the amplitude envelope of the playback signal is adjusted according to the envelope change of the corresponding frequency band of the actual vehicle sound signal; when the fluctuation of the reproduced scene is insufficient, the amplitude of the envelope change is increased; when the fluctuation of the reproduced scene is too large, the amplitude of the envelope change is decreased.

[0062] This embodiment uses a sound pressure level error of less than or equal to 1.5 dB and a sound pressure fluctuation error of less than or equal to 1 dB for each frequency band as noise reproduction conditions. The noise reproduction threshold is jointly calibrated based on the impact of the test system's repeatability deviation and the noise reproduction error on the perception response index. Specifically, under the condition that the cabin ambient sound and playback parameters remain unchanged, the playback is repeated at least five times, and the repeatability deviation of the sound pressure level and sound pressure fluctuation of each frequency band is calculated. The 95th percentile of each repeatability deviation is used as the lower limit of the threshold. During the repeated playback, the frequency band gain or amplitude envelope of the corresponding playback channel is changed in 0.5 dB increments to determine the minimum error at which the prompt recognition rate, response time, or acoustic interference score begins to show stable changes. This minimum error is used as the upper limit of the threshold. The noise reproduction threshold is determined between the lower limit and the upper limit of the threshold. After the above calibration, this embodiment uses a sound pressure level error of less than or equal to 1.5 dB and a sound pressure fluctuation error of less than or equal to 1 dB for each frequency band as noise reproduction conditions.

[0063] During driving task reproduction, the target vehicle speed change times, target vehicle speed values, road curvature sequences, specified operation events, sub-task events, and task durations in the real vehicle segments are converted into task scripts executable by the driving simulation equipment. The reproduction errors of each driving load characteristic in the real vehicle scenario and the reproduced scenario are calculated separately. When the driving load characteristic value in the real vehicle scenario is non-zero, the relative error is obtained by dividing the absolute value of the difference between the two by the characteristic value in the real vehicle scenario. When the driving load characteristic value in the real vehicle scenario is zero, it is checked whether the corresponding characteristic value in the reproduced scenario is also zero. If both are zero, the corresponding reproduction error is recorded as zero; if the corresponding characteristic value in the reproduced scenario is non-zero, the corresponding characteristic is determined to not meet the reproduction conditions.

[0064] When the relative error of any driving load characteristic exceeds 5%, the corresponding driving task parameters are corrected. For example, if the target vehicle speed change frequency is insufficient, the time interval between adjacent target vehicle speed change commands is shortened; if the road curvature change frequency is insufficient, curvature change nodes are added; if the frequency of secondary tasks is too high, the time interval between adjacent secondary tasks is extended.

[0065] After correcting the noise playback parameters and driving task parameters, the scenario is reproduced and measured again. The maximum number of corrections is set to five. If, after five corrections, both the driver's ear position noise reproduction error and the driving load reproduction error meet the corresponding reproduction conditions, the reproduced scenario is determined as a valid test scenario; otherwise, the valid test qualification of the corresponding scenario is cancelled, and a real vehicle segment is reselected for reproduction.

[0066] The driving load reproduction error is calculated only based on the objective parameters of the driving task, without modifying the driving task based on the driver's recognition rate, response time, or subjective rating, thus avoiding changes in the scenario reproduction conditions due to individual differences among different drivers.

[0067] The reference acoustic cues can be a composite cues containing frequency components of 500Hz, 1000Hz, and 2000Hz, with a duration of 500ms and a minimum interval of 15s between adjacent cues. The reference acoustic cues are played through a fixed output channel, and the zero-dB output gain is calibrated to 65dB(A) at the driver's ear position under quiet test conditions.

[0068] The output gain of the acoustic cue represents the digital gain relative to the reference acoustic cue. For example, with an output gain of 3dB, the target ear position sound pressure level is approximately 68dB(A); the actual ear position sound pressure level is still measured via the microphone and is not directly replaced by the digital gain in the acoustic measurement.

[0069] The initial gain test range can be determined through pre-testing. The lowest output gain that results in correct identification is extended downwards by 3dB as the lower end of the range, and the output gain at which the acoustic interference score first exceeds the limit is extended upwards by 3dB as the upper end of the range, both limited to the safe range allowed by the output device. In this embodiment, -6dB to 12dB is used as the initial gain test range, and the output gain is initially set in 3dB intervals.

[0070] Each output gain in each valid test scenario is played at least five times. The execution order of the four scenarios and the playback order of each output gain are determined by randomization; the same output gain is not played twice consecutively. The driver rests for at least 3 minutes after completing each scenario to reduce learning and fatigue effects.

[0071] After the prompt is displayed, the driver has 5 seconds to select the prompt type using the steering wheel buttons. A correct selection is recorded as correct recognition, an incorrect selection as misrecognition, and no selection as no response. The response time is the time from the start of the prompt output to the first valid button press. Misrecognition and no-response samples are not included in the calculation of the median response time, but are retained in the calculation of the prompt recognition rate.

[0072] After each acoustic cues are recognized, the driver provides a clarity score and an acoustic interference score based on a seven-level scale. These scores are then associated with the scene identifier, output gain, and playback sequence number of the cues. A recovery interval is set before the next cues are tested to prevent the scoring process from altering the driving load conditions for the next cues test. In the clarity score, 1 indicates completely inaudible, and 7 indicates completely clear. In the acoustic interference score, 1 indicates no interference, and 7 indicates unacceptable severe interference.

[0073] For each acoustic cue, the digital reference signal from the cue output module and the driver's ear position sound signal are recorded synchronously, and the ambient noise sound pressure level for each frequency band is calculated from the ambient sound at least 2 seconds before the cue output. The transmission response from the cue output channel to the driver's ear position is determined based on the cross-spectrum between the digital reference signal and the ear position sound signal, as well as the auto-spectrum of the digital reference signal. The cue sound component coherent with the digital reference signal is extracted from the ear position sound signal. The sound pressure level, loudness, duration, and modulation depth for each frequency band are calculated based on the cue sound component; ambient noise is only used to determine the effective masking threshold. If the coherence between the digital reference signal and the ear position sound signal is lower than the test system's calibration lower limit, the measurement is marked as invalid and replayed.

[0074] The effective masking threshold is determined based on the ambient noise sound pressure level and masking correction amount for the corresponding frequency band. The masking correction amount is obtained through auditory calibration in a quiet environment: the test prompt signal is progressively increased in the ambient noise of each frequency band, and the sound pressure level difference between the prompt signal and the ambient noise is recorded when the test personnel reach the specified recognition rate. This sound pressure level difference is used as the masking correction amount for the corresponding frequency band.

[0075] Subtracting the effective masking threshold of the k-th frequency band from the sound pressure level of the acoustic cue in that band yields the cue perception margin for that band. Frequency bands with a cue perception margin of 3 dB are defined as effective cue bands. Dividing the number of effective cue bands by the total number of analyzed frequency bands included in the acoustic cue yields the proportion of effective cue bands.

[0076] Loudness was determined using the Zwicker loudness calculation method. The driver's ear position sound signal was converted to a one-third octave band sound pressure level, and then outer ear and middle ear transmission corrections, critical band switching, masking corrections, and characteristic loudness integration were performed sequentially to obtain the total loudness in sonos. The same loudness calculation method and parameters were used for the same batch of tests. Duration was the time between the first time the cue signal envelope exceeded the peak value by 10% and the last time it fell below the peak value by 10%. Modulation depth was determined based on the ratio of the difference between the maximum and minimum values ​​of the signal envelope during the cue duration to their sum.

[0077] The perception response index is calculated as follows: The prompt recognition rate is the ratio of the number of correct recognitions to the total number of prompt playbacks; the response time is the median response time of the correctly recognized samples; the sharpness score is the average sharpness score under the same scene and the same output gain; and the acoustic interference score is the average acoustic interference score under the same scene and the same output gain.

[0078] When the single response time is less than 150ms, the corresponding data will be identified as data processed in advance and removed. Invalid samples caused by device disconnection or missing data timestamps will not be included in the total number of playbacks, and the reason for removal will be recorded.

[0079] For each driving load state and each output gain, a quiet reference scenario is set up with identical driving tasks and acoustic cues. In the quiet reference scenario, the background noise at the driver's ear position is controlled below 35 dB(A).

[0080] For the same driver, the same driving task, and the same output gain, the difference between the objective acoustic metrics of the effective test scenario and the objective acoustic metrics of the quiet reference scenario is calculated as the baseline correction for the objective acoustic metrics. Similarly, the difference between the perceived response metrics of the effective test scenario and the perceived response metrics of the quiet reference scenario is calculated as the baseline correction for the perceived response metrics. Simultaneously, the original objective acoustic metrics and original perceived response metrics from each effective test scenario are retained. The original metrics are used to establish the basic mapping relationship and determine the gain boundary.

[0081] The baseline correction for the recognition rate is calculated by subtracting the recognition rate of the quiet reference scene from the recognition rate of the effective test scene; the baseline corrections for response time, sharpness score, and acoustic interference score are calculated using the same interpolation method. Before calculating the coupling effect, the baseline corrections for different dimensions are divided by the range of variation of the corresponding indicator across all test samples; indicators with a range of zero variation are not included in the calculation of the coupling effect or the selection of core indicators.

[0082] Baseline correction is used to reduce the interference of individual driver differences and test equipment deviations on the calculation of coupling effect. After baseline correction is completed, the coupling effect is calculated using the baseline correction values ​​of four valid test scenarios; the original objective acoustic indicators and original perception response indicators are still used when establishing mapping relationships and determining gain boundaries to ensure that the prediction results use the same units as the prompt recognition rate, response time, and scoring threshold.

[0083] For any objective acoustic index or perceptual response index, under the same acoustic cue type and the same output gain, the index values ​​of the first to fourth scenes are recorded as the first scene index value, the second scene index value, the third scene index value, and the fourth scene index value, respectively.

[0084] Under the first driving load condition, the change value of the first indicator is obtained by subtracting the first indicator value from the second indicator value; under the second driving load condition, the change value of the second indicator is obtained by subtracting the third indicator value from the fourth indicator value; the coupling influence of the corresponding indicator is obtained by subtracting the change value of the first indicator from the change value of the second indicator. C=(Q 22 -Q 12 )-(Q 21 -Q 11 ); Among them, Q 11 Q represents the index value under the first noise state and the first driving load state. 21Q represents the index value under the second noise state and the first driving load state. 12 Q represents the index value under the first noise state and the second driving load state. 22 The index values ​​represent the second noise state and the second driving load state, and C represents the coupling effect.

[0085] For example, at a certain output gain, the cue recognition rates for the four scenarios are 0.92, 0.80, 0.86, and 0.60, respectively. The change in recognition rate due to noise variation under the first driving load condition is -0.12, and the change under the second driving load condition is -0.26, with a coupled effect of -0.14 on the cue recognition rate. This indicates that when the second noise condition and the second driving load condition coexist, the cue recognition rate exhibits a decrease that cannot be simply explained by the sum of the effects of the two factors individually.

[0086] Driving load primarily alters the driver's perceptual response to acoustic cues, but does not necessarily change the physical sound field at the ear position. Therefore, the coupling effect of objective acoustic indicators can also be used to identify scene reproduction residuals, while the coupling effect of perceptual response indicators is used to characterize the additional impact of complex scenes on the driver's perception.

[0087] Using the original objective acoustic metrics of each valid test scenario as input, and the original cue recognition rate, original response time, original sharpness score, or original acoustic interference score as output, basic mapping relationships are established respectively. The cue recognition rate adopts a logistic regression relationship, while the response time, sharpness score, and acoustic interference score adopt a linear regression relationship. When the metrics do not satisfy linear change, a piecewise monotonic mapping relationship is established according to the ranking results of the objective acoustic metrics.

[0088] For the p-th objective acoustic index and the q-th perceived response index, the original objective acoustic indices for the four scenarios are input into the basic mapping relationship to obtain the predicted perceived response indices for the four scenarios. The predicted coupling influence of the predicted perceived response index is calculated according to the same cross-difference order as the measured coupling influence. The measured coupling influence of the q-th perceived response index is subtracted from the predicted coupling influence to obtain the interaction correction term between the two. The interaction correction term represents the additional perceived influence caused by the combined effect of noise and driving load that the basic mapping relationship fails to explain.

[0089] Cross-scene markers are set in the corrected mapping relationship. The cross-scene markers for the first, second, and third scenes are zero, and the cross-scene marker for the fourth scene is one. The product of the interaction correction term and the cross-scene marker is added to the output of the basic mapping relationship. The correction result for the cue recognition rate is limited to between zero and one, the correction result for the response time is not less than zero, and the correction results for the sharpness score and acoustic interference score are limited to between one and seven. This yields a corrected mapping relationship with the same dimensions as the original perceptual response index.

[0090] The mapping parameters are obtained using no less than two-thirds of the test data, with the remaining test data used for validation. The test data is divided by driver to ensure that data from the same driver is not simultaneously included in the parameter calculation and validation data. The prediction error in the validation data is used as the mapping residual.

[0091] For each objective acoustic indicator, the mapping direction between the indicator after interactive correction and the prompt recognition rate, response time, and acoustic interference score is calculated.

[0092] When objective acoustic metrics increase, the recognition rate should show an upward trend, and the response time should show a downward trend. Objective acoustic metrics that satisfy these two mapping directions are identified as candidate perceptual metrics. Objective acoustic metrics that show an upward trend in acoustic interference scores when objective acoustic metrics increase are identified as candidate interference metrics.

[0093] This embodiment uses the rank correlation coefficient to determine the mapping direction. When the rank correlation coefficient between the objective acoustic indicator and the cue recognition rate is greater than zero, and the rank correlation coefficient between the objective acoustic indicator and the response time is less than zero, the corresponding objective acoustic indicator is considered to satisfy the mapping direction of the perceptual candidate indicator. When the rank correlation coefficient between the objective acoustic indicator and the acoustic interference score is greater than zero, the corresponding objective acoustic indicator is considered to satisfy the mapping direction of the interference candidate indicator. When the absolute value of the rank correlation coefficient is less than 0.3, the mapping relationship is considered unstable, and the corresponding candidate indicator is removed. This embodiment determines the association stability threshold through a permutation test. Specifically, while keeping the objective acoustic indicators unchanged, the perceptual response indicators are randomly shuffled within samples of the same acoustic cue type and the same output gain. The rank correlation coefficient between the shuffled objective acoustic indicators and the perceptual response indicators is calculated. This process is repeated at least 1000 times, and the 95th percentile of the obtained rank correlation coefficient is rounded up to 0.1 as the association stability threshold. The obtained association stability threshold is used to exclude weak correlations that may still occur due to random sample arrangement. In this embodiment, the association stability threshold is set to 0.3.

[0094] The mapping residual is calculated using the root mean square of the difference between the predicted and measured values ​​of the validation samples. The repeated test dispersion is calculated using the standard deviation of the repeated test results under the same scenario and the same output gain. The mapping residual and the repeated test dispersion are normalized by dividing by the maximum value among the corresponding candidate indicators; when the corresponding maximum value is zero, the corresponding normalization result is recorded as zero.

[0095] In this embodiment, the weight of the normalized mapping residual is set to 0.7, and the weight of the normalized repeated test dispersion is set to 0.3. The weighted results of each candidate index are calculated. The perceptual candidate index with the smallest weighted result is determined as the perceptual core index, and the interference candidate index with the smallest weighted result is determined as the interference core index.

[0096] For example, if the effective cue frequency band ratio simultaneously meets the requirements of positive change with cue recognition rate and negative change with response time, and has the smallest weighted result, then the effective cue frequency band ratio is determined as the core perception indicator; if loudness and acoustic interference score show a stable positive change, and has the smallest weighted result among interference candidate indicators, then loudness is determined as the core interference indicator.

[0097] The core indicators of perception and the core indicators of interference are screened separately. It is not required that the same objective acoustic indicator simultaneously characterize the perceptibility and acoustic interference.

[0098] In this embodiment, a recognition rate of no less than 90% and a response time of no more than 2 seconds are used as perception conditions, and an acoustic interference score of no more than 3 points under a seven-level scale is used as acoustic interference conditions. These conditions can be determined based on product testing specifications, driving safety requirements, or previous human factors testing results, and should remain unchanged within the same batch of tests.

[0099] First, within the initial gain test range of -6dB to 12dB, each output gain is tested at 3dB intervals. Based on the core perception indicators and the corrected mapping relationship, the original prompt recognition rate and original response time under each output gain are predicted, and output gains that meet both requirements are selected. When two adjacent output gains respectively fail to meet the perception conditions and meet them, these two are determined as the perception boundary search interval. If all perception conditions are not met within the initial gain test interval, the test interval is expanded upwards within the allowable range of the prompt output device. If no satisfactory result is found, an evaluation result of "perception conditions cannot be met" is generated.

[0100] Set a new output gain at the midpoint of the sensing boundary search interval and repeat the test. If the new output gain meets the sensing conditions, adjust the upper end of the search interval to the new output gain; if it does not meet the sensing conditions, adjust the lower end of the search interval to the new output gain. Repeat the above process until the difference between adjacent output gains is less than 0.5dB, and determine the gain on the side that meets the sensing conditions as the lower limit of the output gain.

[0101] Based on the core interference indicators and the corrected mapping relationship, the original acoustic interference score for each output gain is predicted, and output gains that meet the acoustic interference conditions are selected. When two adjacent output gains meet and do not meet the acoustic interference conditions respectively, they are defined as the interference boundary search interval. When all output gains within the initial gain test interval meet the acoustic interference conditions, the maximum allowable output gain of the output device is used as the upper limit of the output gain. When all output gains do not meet the acoustic interference conditions, an evaluation result of "acoustic interference conditions not met" is generated. The test is repeated within the interference boundary search interval using the midpoint insertion method until the difference between adjacent output gains is less than 0.5dB. The gain on the side that meets the acoustic interference conditions is determined as the upper limit of the output gain.

[0102] When the lower limit of output gain is less than or equal to the upper limit of output gain, the interval between the two is defined as the scene's safe gain range. For example, if the lower limit of output gain is 2.5dB and the upper limit of output gain is 7dB, then the scene's safe gain range is 2.5dB to 7dB.

[0103] When the lower limit of the output gain is greater than the upper limit of the output gain, it indicates that the current acoustic cues cannot simultaneously meet the perception conditions and acoustic interference conditions in the corresponding scenario. An evaluation result of "No scene safety gain range exists" is generated, and a prompt is given to adjust the frequency band composition, duration, or modulation method of the acoustic cues and retest. If no core perception or interference indicators are selected, or if the corresponding boundary search interval cannot be obtained, no scene safety gain range is output, and a corresponding insufficient data prompt is generated.

[0104] In the vehicle or cockpit prototype under evaluation, the output gain of the acoustic cue under test relative to the reference acoustic cue is read from the cue output module, and the actual sound pressure level of the acoustic cue under test is measured through the driver's ear microphone to verify the calibration relationship between the output gain and the ear sound pressure level. The noise state and driving load state of the current test scenario are determined, and the corresponding scenario safety gain range is invoked. The acoustic cue under test and the reference acoustic signal used to establish the scenario safety gain range have the same frequency band composition, duration, modulation method, and cue output position; if any condition is inconsistent, the gain range test is re-executed.

[0105] When the output gain of the acoustic prompt under test is lower than the lower limit of the output gain, an evaluation result of "insufficient perception" is generated; when it is within the safe gain range of the scene, an evaluation result of "effective" is generated; when it is higher than the upper limit of the output gain, an evaluation result of "acoustic interference exceeds the limit" is generated.

[0106] The evaluation results should include at least the test scenario identification, noise spectrum characteristics, driving load characteristics, driver ear position noise reproduction error, driving load reproduction error, perception core indicators, interference core indicators, lower limit of output gain, upper limit of output gain, output gain of the acoustic prompt under test, and evaluation category.

[0107] See Figure 2 As shown, the data acquisition module sends cabin ambient sound data, vehicle operation data, driving task data, and acoustic prompt data to the scene processing module. The scene processing module completes synchronous segmentation, feature extraction, feature distance calculation, and matching scene group construction, and sends the noise data of each scene to the noise playback module and the driving task script to the driving task loading module.

[0108] The noise playback module plays back the cabin ambient sound according to the frequency band gain given by the data processing module; the driving task loading module loads the driving task according to the driving task parameters; the prompt output module plays acoustic prompts according to each output gain; and the response acquisition module sends the driver's response data to the data processing module.

[0109] The data processing module sequentially performs driver ear position noise reproduction error calculation, driving load reproduction error calculation, reproduction parameter correction, objective acoustic index calculation, perception response index calculation, baseline correction, coupling effect calculation, mapping relationship correction, core index screening, and scene safety gain range determination, and generates the evaluation results of the acoustic prompt to be tested.

[0110] Through the above implementation methods, real vehicle data acquisition, four-way cross-scenario construction, test environment reproduction, subjective and objective index acquisition, coupling effect separation and output gain boundary determination can be completed according to a unified data format and judgment conditions, so that different testers can repeat the present invention according to the same process.

[0111] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the present invention.

Claims

1. A method for processing acoustic cues perception data in an intelligent cockpit, characterized in that, include: Acquire cabin ambient sound data, vehicle operation data, driving task data, acoustic prompt data, and driver response data from real-person vehicle tests, and simultaneously segment and extract noise spectrum characteristics and driving load characteristics; Based on the noise spectrum characteristics and driving load characteristics, a set of matching scenarios is constructed, which includes cross-combinations of different noise states and different driving load states; The matching scenario group was reproduced in the test environment. The noise playback parameters were corrected according to the driver's ear position noise reproduction error, and the driving task parameters were corrected according to the driving load reproduction error to obtain an effective test scenario. In each effective test scenario, acoustic prompts are played according to multiple output gains, objective acoustic indicators are measured and calculated, driver response data is collected and perceived response indicators are calculated, the coupling effect of noise and driving load is determined according to the difference between the indicators between the cross combinations, and the mapping relationship between objective acoustic indicators and perceived response indicators is corrected using the coupling effect. Based on the corrected mapping relationship, the lower limit of the output gain that meets the requirements of prompt recognition rate and response time, and the upper limit of the output gain that meets the requirements of acoustic interference scoring are determined to form the scene safety gain range. The evaluation result is generated based on the comparison between the output gain of the acoustic prompt under test and the scene safety gain range.

2. The intelligent cockpit acoustic cues perception data processing method according to claim 1, characterized in that, The synchronization segments include: The cabin ambient sound data, vehicle operation data, and driving task data are aligned according to a unified time scale; the changes in sound pressure level and driving load characteristics in each frequency band are calculated within a continuous time window, and the moment when any change exceeds the corresponding change threshold is determined as a candidate change point; candidate change points with a time interval less than a preset interval are merged, and the data is divided according to the merged candidate change points; the acoustic prompt data and driver response data are associated with the corresponding segments according to the output time of the acoustic prompt and the driver response time.

3. The intelligent cockpit acoustic cues perception data processing method according to claim 2, characterized in that, The noise spectrum characteristics include sound pressure level in each frequency band, sound pressure fluctuation in each frequency band, and dominant noise frequency band; the driving load characteristics include target vehicle speed change frequency, road curvature change frequency, number of specified operations, frequency of sub-task occurrence, and task duration. The noise spectrum characteristics and driving load characteristics are respectively processed into dimensionless values, and the distance between the noise spectrum characteristics and driving load characteristics between different segments is calculated based on the dimensionless values.

4. The intelligent cockpit acoustic cues perception data processing method according to claim 3, characterized in that, The matching scenario group includes at least: The first scenario formed by the combination of the first noise state and the first driving load state, the second scenario formed by the combination of the second noise state and the first driving load state, the third scenario formed by the combination of the first noise state and the second driving load state, and the fourth scenario formed by the combination of the second noise state and the second driving load state. Among them, the noise spectrum feature distance between scenes corresponding to the same noise state is less than the first matching threshold, and the driving load feature distance between scenes corresponding to the same driving load state is less than the second matching threshold; the noise spectrum feature distance between the first noise state and the second noise state reaches the first distinguishing threshold, and the driving load feature distance between the first driving load state and the second driving load state reaches the second distinguishing threshold. The first matching threshold is the maximum value of the noise spectrum feature distance under the same noise state, and the second matching threshold is the maximum value of the driving load feature distance under the same driving load state; the first distinguishing threshold is the minimum value of the noise spectrum feature distance between different noise states, and the second distinguishing threshold is the minimum value of the driving load feature distance between different driving load states. The first noise state is when the total sound pressure level at the driver's ear is in the first sound pressure level range and the dominant noise frequency band is in the preset frequency band range; the second noise state is when the total sound pressure level at the driver's ear is in the higher second sound pressure level range and the dominant noise frequency band is in the preset frequency band range. The first driving load state is a state in which the target vehicle speed change frequency and the secondary task occurrence frequency are both in the first frequency range, and the second driving load state is a state in which the target vehicle speed change frequency and the secondary task occurrence frequency are both in the higher second frequency range.

5. The intelligent cockpit acoustic cues perception data processing method according to claim 4, characterized in that, The matching scenario group to be reproduced includes: The cabin ambient sound is played back through a multi-channel speaker, and a driving task is loaded through a driving simulation device. The driver's ear position noise reproduction error includes the sound pressure level error and frequency band sound pressure fluctuation error between the real vehicle scene and the reproduced scene. The frequency band gain of the corresponding playback channel is corrected according to the direction of the driver's ear position noise reproduction error. The driving load reproduction error includes the reproduction error of the target vehicle speed change frequency, road curvature change frequency, specified number of operations, frequency of occurrence of sub-tasks, and duration of tasks. The driving task parameters are corrected according to the driving load reproduction error. The correction is repeated until both types of reproduction errors are less than or equal to the corresponding reproduction thresholds, and the reproduction scene that meets the corresponding reproduction threshold is determined as the valid test scene.

6. The intelligent cockpit acoustic cues perception data processing method according to claim 5, characterized in that, The objective acoustic indicators include the difference between the sound pressure level of each frequency band of the acoustic cues at the driver's ear position and the corresponding effective masking threshold, the proportion of effective cues frequency bands, loudness, duration, and modulation depth; the perceptual response indicators include cues recognition rate, response time, clarity score, and acoustic interference score. Using a scenario with the same driving task and output gain, and a noise sound pressure level lower than a preset value as a quiet reference scenario, calculate the changes in objective acoustic indicators and perceptual response indicators of each effective test scenario relative to the quiet reference scenario, and use the changes to determine the coupling effect. Retain the original objective acoustic indicators and original perceptual response indicators of each effective test scenario, and use the coupling effect to correct the mapping relationship established between the original objective acoustic indicators and the original perceptual response indicators.

7. The intelligent cockpit acoustic cues perception data processing method according to claim 6, characterized in that, Determining the coupling effect includes: Under the first driving load condition, calculate the change value of the first index under the second noise condition relative to the first noise condition; under the second driving load condition, calculate the change value of the second index under the second noise condition relative to the first noise condition; and determine the difference between the change value of the second index and the change value of the first index as the coupling influence of the corresponding index. The original objective acoustic indicators of the four cross-scenes are input into the basic mapping relationship to obtain the predicted perception response indicators of the four cross-scenes. The predicted coupling influence is determined based on the difference between the predicted perception response indicators and the cross-scenes combinations. The difference between the measured coupling influence and the predicted coupling influence of the perception response indicators is determined as the interaction correction term. The interaction correction term is added to the output of the basic mapping relationship to obtain the corrected mapping relationship.

8. The intelligent cockpit acoustic cues perception data processing method according to claim 7, characterized in that, The method further includes: Based on the mapping relationship corrected by the interaction correction term, the mapping direction, mapping residual and repeated test dispersion between each objective acoustic index and the prompt recognition rate, response time and acoustic interference score are determined respectively. Objective acoustic indicators that have a preset mapping direction with both prompt recognition rate and response time are identified as candidate perception indicators, and objective acoustic indicators that have a preset mapping direction with acoustic interference score are identified as candidate interference indicators. The mapping residuals and repeated test dispersions of each candidate indicator are normalized and weighted, and the candidate perception indicator with the smallest weighted result is selected as the core perception indicator, and the candidate interference indicator with the smallest weighted result is selected as the core interference indicator.

9. The intelligent cockpit acoustic cues perception data processing method according to claim 8, characterized in that, Determining the security gain range for the scenario includes: Multiple output gains are set within a preset initial gain test range. Based on the core perception index and the corresponding corrected mapping relationship, the original prompt recognition rate and original response time under each output gain are predicted, and the minimum output gain that meets the requirements for both prompt recognition rate and response time is determined. Based on the core interference index and the corresponding corrected mapping relationship, the original acoustic interference score under each output gain is predicted, and the maximum output gain that meets the requirements for acoustic interference score is determined. Insert a new output gain between adjacent output gains and repeat the test until the difference between adjacent output gains is less than the gain resolution. The minimum output gain obtained is taken as the lower limit of the output gain, and the maximum output gain obtained is taken as the upper limit of the output gain. When the lower limit of the output gain is less than or equal to the upper limit of the output gain, the interval between the two is determined as the scene safety gain range.

10. A smart cockpit acoustic cues perception data processing system, used to implement the smart cockpit acoustic cues perception data processing method according to any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect cabin ambient sound data, vehicle operation data, driving task data, and acoustic prompt data during real-person and real-vehicle testing. The scene processing module is used to synchronously segment the data, extract noise spectrum features and driving load features, and construct matching scene groups that include cross combinations of different noise states and different driving load states. A noise playback module is used to reproduce the cabin ambient sound in the matched scene group through a multi-channel speaker; The driving task loading module is used to reproduce the driving tasks in the matching scenario group through a driving simulation device; The prompt output module is used to play acoustic prompts according to multiple output gains; The response acquisition module is used to collect driver response data; The data processing module is used to perform reproduction parameter correction based on the driver's ear position noise reproduction error and driving load reproduction error, calculate the coupling effect of noise and driving load based on the index difference between the objective acoustic index and the perceived response index in the cross combination, correct the mapping relationship between the objective acoustic index and the perceived response index using the coupling effect, determine the scene safety gain range based on the corrected mapping relationship, and generate the evaluation result of the acoustic prompt to be tested.

Images