In-vehicle environment adjusting method, vehicle and medium

Abstract

The invention discloses an in-vehicle environment adjusting method, a vehicle and a medium, and relates to the technical field of intelligent cabins. The method comprises the steps of obtaining an original pressure signal collected by a pressure sensor in a vehicle air conditioner air duct, and preprocessing the original pressure signal to extract an effective pressure fluctuation signal related to passenger activity; wherein the sampling frequency of the pressure sensor is configured to be capable of capturing pressure fluctuation caused by the activity of the passenger; performing scene probability calculation based on the effective pressure fluctuation signal to determine an identification result of the current state of the passenger in the vehicle; and based on the recognition result, generating a control signal for controlling at least one execution mechanism in the vehicle so as to drive the execution mechanism to execute environment control based on the passenger state. According to the invention, on the premise of avoiding infringement of passenger privacy and not increasing system hardware cost, multi-dimensional passenger state accurate identification from static existence to dynamic activity is realized, so that a more intelligent and efficient adaptive control scheme is provided for the in-vehicle environment.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of intelligent cockpit, in particular to an in-vehicle environment regulation method, a vehicle and a medium. BACKGROUND

[0002] In the technical field of intelligent cockpit, in order to realize personalized comfortable experience and efficient energy management, accurate perception of the state of in-vehicle occupants (such as the number, position and activity type) has become a key requirement. At present, the common occupant perception scheme mainly relies on optical sensing systems or seat pressure sensors. Although the optical sensing system has high recognition accuracy, its hardware cost is high, and it continuously collects images in the private space of the vehicle, which has a significant risk of invading the privacy of the occupants, reducing the acceptance of users. The seat pressure sensor can usually only provide a binary judgment of the presence or absence of an occupant, and cannot distinguish the specific activity state, nor can it detect the activity of the occupant in the non-seat area, resulting in a single recognition dimension, limiting the depth and breadth of intelligent control of the system.

[0003] In addition, the above-mentioned prior art solutions all need to install special perception hardware in the vehicle, which not only increases the overall complexity and manufacturing cost of the system, but also limits its application on cost-sensitive vehicle models. These solutions have limitations in the perception dimension and cannot provide rich and continuous occupant state information for higher-order intelligent environment control, resulting in a still relatively extensive control strategy for the execution mechanism such as the air conditioner, making it difficult to achieve true personalization and energy efficiency optimization.

[0004] Therefore, how to avoid invading the privacy of the occupants and significantly reduce the hardware cost of the system, while realizing multi-dimensional accurate identification of the state of the occupants from static existence to dynamic activity, thereby providing more intelligent and efficient adaptive control for the in-vehicle environment, has become a technical problem that needs to be solved by the technical personnel in the field. SUMMARY

[0005] In view of the above problems, the present disclosure provides an in-vehicle environment regulation method, a vehicle and a medium which overcome the above problems or at least partially solve the above problems, and the technical solutions are as follows: An in-vehicle environment regulation method, the method comprising: acquiring an original pressure signal collected by a pressure sensor in an air conditioning duct of a vehicle, and preprocessing the original pressure signal to extract an effective pressure fluctuation signal related to occupant activity; wherein the sampling frequency of the pressure sensor is configured to be able to capture pressure fluctuations caused by occupant activity; performing scene probability calculation based on the effective pressure fluctuation signal to determine the recognition result of the current in-vehicle occupant state; based on the recognition result, generating a control signal for controlling at least one execution mechanism in the vehicle to drive the execution mechanism to perform environment control based on the occupant state.

[0006] The present disclosure provides a vehicle interior environment adjustment method. The method reuses the original air duct pressure sensor of the vehicle air conditioning system and configures the sampling frequency of the sensor to capture the pressure fluctuation caused by the passenger activity, thereby identifying the passenger state. This design eliminates the need for additional dedicated sensors (such as OMS cameras or seat pressure sensors), thereby reducing the system cost and complexity from the hardware root. At the same time, since the airflow pressure, a non-optical and non-contact physical quantity, is used, the privacy leakage risk caused by image acquisition is fundamentally avoided, and the user trust is improved. Further, by linking the identification result with the vehicle interior environment control system, the system is able to realize the leap from simply "sensing the existence" to "understanding the state" and "actively responding", thereby significantly improving the intelligence level and driving comfort of the system.

[0007] Optionally, the original pressure signal is pre-processed to extract an effective pressure fluctuation signal related to the passenger activity, specifically including: performing band-pass filtering on the original pressure signal to filter out signal components with a frequency lower than a first preset threshold and higher than a second preset threshold, to obtain a pressure fluctuation signal in a target frequency band; wherein the target frequency band contains a characteristic pressure fluctuation frequency range caused by human activity; performing noise reduction processing on the pressure fluctuation signal in the target frequency band to obtain a preliminary denoising signal; based on a reference noise signal preset by a vehicle vibration sensor, estimating and eliminating noise components related to vehicle vibration from the preliminary denoising signal by an adaptive filtering algorithm, to output the effective pressure fluctuation signal.

[0008] In this embodiment, the combination of band-pass filtering and adaptive filtering based on the reference noise is used for noise reduction strategy, which can systematically separate the effective pressure fluctuation signal related to the passenger activity from the original pressure signal. The band-pass filtering first limits the signal to the characteristic frequency band of human activity, which preliminarily filters out the noise outside the frequency band (such as the high-frequency noise of the fan); then, the signal of the vehicle vibration sensor is used as a reference to dynamically estimate and eliminate the vibration noise related to the vehicle running state by the adaptive filtering technology. The direct technical effect of this series of cooperative processing is that even in the complex working conditions of air conditioning on and vehicle driving, the system can still extract the effective signal with high signal-to-noise ratio, which lays a reliable signal foundation for the accurate identification of the subsequent mode.

[0009] Optionally, the pressure fluctuation signal in the target frequency band is processed to obtain a preliminary denoising signal, specifically including: performing multi-resolution analysis on the pressure fluctuation signal in the target frequency band by wavelet transform to decompose the pressure fluctuation signal into sub-band signals of different frequency scales; quantizing the wavelet coefficients of each sub-band signal based on a preset threshold rule to remove transient impulse interference, and reconstructing each sub-band signal after the quantization processing to generate the preliminary denoising signal.

[0010] The wavelet transform denoising process further defined in this embodiment is particularly suitable for processing non-stationary pressure fluctuation signals generated by occupant activities. The multi-resolution analysis characteristic of wavelet transform enables it to effectively identify and remove transient pulse interference (such as sudden noise caused by vehicle bumps, door opening and closing) in the signal, which is difficult to completely eliminate by traditional filtering methods. This feature plays a role in the technical solution in that it further purifies the signal after preliminary denoising, avoiding misleading of the subsequent feature extraction and pattern recognition process by transient pulses, thereby improving the robustness and recognition accuracy of the system in real driving environments.

[0011] Optionally, scene probability calculation is performed based on the effective pressure fluctuation signal to determine the recognition result of the current in-vehicle occupant state, specifically including: extracting time domain features and frequency domain features from the effective pressure fluctuation signal to construct a to-be-recognized feature vector representing the current in-vehicle state; wherein the time domain features include the amplitude, mean and variance of the signal waveform, and the frequency domain features include the signal spectrum features obtained by fast Fourier transform; inputting the to-be-recognized feature vector into the machine learning model to perform local feature perception and hierarchical abstraction on the to-be-recognized feature vector, and outputting probability distribution corresponding to different occupant scenes; determining the occupant scene with the highest probability as the recognition result of the current in-vehicle occupant state.

[0012] In this embodiment, the feature vector is constructed by extracting time domain and frequency domain features from the effective pressure fluctuation signal, and the machine learning model is used to output probability distribution, which can completely represent the uniqueness of the occupant activity pattern from different dimensions. This embodiment realizes fine-grained differentiation of occupant scenes, enabling the system to intelligently learn from pressure fluctuation patterns and accurately judge complex activity states (such as conversation, movement, etc.), thereby improving the accuracy and scene adaptability of the recognition result.

[0013] Optionally, the time domain features and frequency domain features are extracted from the effective pressure fluctuation signal, specifically including: performing window sliding sampling on the effective pressure fluctuation signal; in the time domain, calculating the amplitude, mean and variance of the signal waveform in each time window to obtain the statistical distribution characteristics of the effective pressure fluctuation signal in the time domain; in the frequency domain, performing fast Fourier transform on the signal segment in each time window to convert the time domain signal to a frequency energy spectrum, and calculating the energy integral and dominant frequency component of the energy spectrum in the preset human activity frequency band to obtain the energy distribution characteristics of the effective pressure fluctuation signal in the frequency domain.

[0014] The embodiment refines the specific way of feature extraction, adopts window sliding sampling and combines time domain statistical characteristics and frequency energy distribution analysis. The feature ensures the real-time and continuity of feature extraction in the technical scheme, can dynamically capture the timing change rule of the occupant activity, and enhances the robust perception ability of the system to the variable occupant state in the real driving environment.

[0015] Optionally, based on the recognition result, a control signal for controlling at least one execution mechanism in the vehicle is generated, specifically including: analyzing the occupant number and position distribution information in the recognition result to generate independent air volume control instructions for different air outlets of the air conditioning system; wherein the control instruction contains a target air volume value corresponding to the occupant distribution area; and / or, analyzing the occupant activity type information in the recognition result, combining a pre-stored activity type and fresh air demand mapping relationship, generating a mode switching instruction for controlling the air conditioning fresh air mode and air volume to adjust the air circulation state in the vehicle.

[0016] The embodiment limits the generation mechanism of the control signal, generates independent control instructions and fresh air mode instructions for the air outlets of the air conditioning system by analyzing the occupant distribution and activity information in the recognition result, realizes accurate environment regulation based on the actual occupant state, for example, automatically reducing the rear air volume when detecting that the rear row is empty, or increasing fresh air when the number of occupants increases, thereby improving comfort while optimizing energy utilization efficiency.

[0017] Optionally, the occupant activity type information in the recognition result is analyzed to generate a mode switching instruction, specifically including: when the identified activity type is a pre-set high-metabolic activity type, a first control instruction is generated to increase the fresh air volume and switch to an external circulation mode to improve the air exchange efficiency in the vehicle; when the identified activity type is a pre-set low-metabolic activity type, a second control instruction is generated to maintain or reduce the fresh air volume and switch to an internal circulation mode to optimize the energy utilization efficiency of the air conditioning system.

[0018] The embodiment further distinguishes the control strategies corresponding to high-metabolic and low-metabolic activity types, realizes intelligent management of the air quality in the vehicle by dynamically adjusting the fresh air mode and air volume. The feature effectively balances the comfort and energy efficiency requirements in the technical scheme, for example, increasing ventilation to reduce CO2 concentration when high-metabolic activity is identified, and maintaining internal circulation to save energy in low-metabolic state, further enhancing the adaptive ability of the system.

[0019] Optionally, the method further comprises: pre-training the machine learning model, specifically comprising: collecting a plurality of sample pressure signals acquired by the pressure sensor under different known passenger scenarios, and labeling each sample pressure signal to form a training data set containing signal data and corresponding scenario labels; using the training data set to iteratively train the machine learning model, and adjusting the convolution kernel weights and bias parameters in the model through a back propagation algorithm until the scenario recognition accuracy of the model output reaches a preset convergence threshold.

[0020] In this embodiment, the process of pre-training the machine learning model with a large amount of labeled data ensures that the system has strong scenario generalization capability. The model can automatically mine the complex and nonlinear mapping relationship between the "pressure fingerprint" and the passenger scenario through iterative learning, rather than relying on fixed threshold rules set by humans. The core beneficial effect brought by this is that the system not only can accurately identify known typical scenarios, but also has good inference and adaptation capability for complex new scenarios composed of multiple activities that are not strictly defined in advance, thereby making the entire system more intelligent and practical.

[0021] A vehicle, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform: acquiring an original pressure signal collected by a pressure sensor in a vehicle air conditioning air duct, and preprocessing the original pressure signal to extract an effective pressure fluctuation signal related to passenger activity; wherein the sampling frequency of the pressure sensor is configured to be able to capture pressure fluctuations caused by passenger activity; inputting the original pressure signal into a pre-trained machine learning model to output an identification result of a current in-vehicle passenger state; based on the identification result, generating a control signal for controlling at least one execution mechanism in the vehicle to drive the execution mechanism to perform environment control based on the passenger state.

[0022] A computer-readable storage medium storing computer-executable instructions, the computer-executable instructions being configured to: acquire an original pressure signal collected by a pressure sensor in a vehicle air conditioning air duct, and preprocess the original pressure signal to extract an effective pressure fluctuation signal related to passenger activity; wherein the sampling frequency of the pressure sensor is configured to be able to capture pressure fluctuations caused by passenger activity; input the original pressure signal into a pre-trained machine learning model to output an identification result of a current in-vehicle passenger state; based on the identification result, generate a control signal for controlling at least one execution mechanism in the vehicle to drive the execution mechanism to perform environment control based on the passenger state. BRIEF DESCRIPTION OF DRAWINGS

[0023] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings: Figure 1 A flowchart of a method for adjusting an in-vehicle environment is shown. Figure 2 A structural diagram of a vehicle is shown. DETAILED DESCRIPTION

[0024] Example embodiments of the present disclosure will be described herein below with reference to the accompanying drawings. Although example embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0025] With the development of intelligent cockpit technology, higher requirements are put forward for the intelligent, personalized, and non-sensing control of the in-vehicle environment. In the prior art, the key premise for realizing adaptive regulation of the cabin environment is accurate perception of the occupant state, and the current mainstream perception scheme mainly relies on an optical vision system or a seat pressure sensor. However, such schemes have inherent defects: first, the optical system can obtain rich visual information, but has high hardware cost, high computing power demand, and inevitable occupant privacy leakage risk; second, the seat pressure sensor can only perceive the static state of whether the seat is occupied, cannot effectively identify the dynamic activities of the occupant (such as conversation, movement), and cannot detect the presence of the occupant in the non-seat area.

[0026] Specifically, the traditional scheme has the following technical shortcomings: 1. Conflict between perception dimension and privacy protection: the existing technology is difficult to obtain dynamic information that can be used to judge the activity state of the occupant without invading the privacy of the occupant. The optical sensor directly collects biometric information, which touches the privacy red line; and the simple pressure or capacitive sensor cannot distinguish between fine states such as "sitting still" and "micro-motion" due to the single dimension of information, resulting in rough and lack of humanized environment control strategies.

[0027] 2. Conflict between signal perception and system cost: adding special sensors (such as OMS cameras, pressure pads distributed in each seat) to improve perception ability directly increases the system hardware cost and wiring complexity. At the same time, these sensors are single-function, and cannot form effective hardware reuse with the existing vehicle systems (such as air conditioning systems), resulting in resource waste and difficulty in popularization in cost-sensitive vehicle models.

[0028] 3. Disconnection between control strategy and dynamic scenarios: Due to the limitation of perception information to static "existence", the control logic of existing environmental control systems (especially air conditioners) is relatively rigid. For example, it is impossible to intelligently adjust the air volume, air direction or switch the internal and external circulation modes of a specific area according to dynamic scenarios such as "there are children frequently moving in the back row" or "there are many people talking in the front row", resulting in low energy utilization efficiency and difficulty in meeting the real-time comfort needs of passengers.

[0029] To this end, the present application provides an in-vehicle environment regulation method, Figure 1 A flowchart of an in-vehicle environment regulation method is provided for one or more embodiments of the present specification. The method can be applied to different types of vehicles, and the flowchart can be executed by a computing device in the corresponding field (such as a controller, a car machine or a server set in the cloud, etc.), and some input parameters or intermediate results in the flowchart allow manual intervention to adjust to help improve accuracy.

[0030] The implementation of the in-vehicle environment regulation method related to the embodiments of the present disclosure can be a terminal device or a server, and the present application does not make special limitations thereon. For the convenience of understanding and description, the following embodiments are described in detail with the server as an example. It should be noted that the server can be a separate device, or a system composed of multiple devices, i.e., a distributed server, and the present application does not make specific limitations thereon.

[0031] As Figure 1 shown, the present disclosure provides an in-vehicle environment regulation method, comprising: Step 101, acquiring an original pressure signal collected by a pressure sensor in an air conditioning duct of a vehicle, and preprocessing the original pressure signal to extract an effective pressure fluctuation signal related to passenger activity. In the present embodiment, the pressure sensor refers to a duct pressure sensing device originally installed in the vehicle air conditioning system. It can be understood that the pressure sensor is mainly used to monitor the basic pressure level in the duct in the conventional air conditioning control to realize the functions of conventional air volume adjustment or constant air volume maintenance. One of the key innovations of the present application is the reuse of these existing sensors, without the need to add any dedicated passenger detection sensor, thereby significantly reducing the system hardware cost and complexity.

[0032] It can be understood that different activities of the passengers in the vehicle will cause different pressure fluctuations. For example, when the passengers are just sitting, the airflow disturbance caused by their steady breathing is very weak, and the corresponding pressure signal shows low-frequency and low-amplitude steady fluctuations. When the rear passengers are moving or playing, the pressure signal of the children is usually higher than that of the adults due to the more irregular and fast amplitude changes of the children's activities. The overall mode is different from that of the adults and may include higher frequency and more irregular fluctuation sequences with sharp amplitude changes and may also be accompanied by sudden pulses.

[0033] Therefore, in order to effectively capture various activities of the human body, the sampling frequency of the pressure sensor is reconfigured. Specifically, it is upgraded from the original low-frequency sampling state (usually several hertz to several tens of hertz) serving only the basic control of the air conditioner to a level capable of sensitively capturing high-frequency pressure fluctuations caused by passenger activities. Exemplarily, the sampling frequency is set to be not less than 100 Hz. This configuration enables the sensor to perceive the micro-motions of the passengers, such as breathing, limb movement, chest movement during conversation, and clothing friction, which will cause small airflow disturbances in the closed cabin and eventually form specific pressure fluctuation patterns in the air conditioner air duct.

[0034] Preferably, the installation position of the pressure sensor is in the main air duct of the air conditioner or a key branch air duct serving the passenger cabin and is located on the upper half of the straight pipe section with stable airflow velocity. Such a layout helps to more directly and clearly perceive the overall pressure changes caused by the activities of the passengers in different areas (such as the front row and the rear row). It should be noted that the installation should avoid ventilation dead angles and strong vibration sources to reduce signal interference introduced by non-passenger activities.

[0035] Further, the original pressure signal is preprocessed to extract the effective pressure fluctuation signal related to the passenger activity.

[0036] In the present embodiment, the purpose of preprocessing the original pressure signal is to separate the effective pressure fluctuation purely generated by the passenger activity from the original signal containing various interferences. It can be understood that the directly collected original pressure signal is a complex mixture, which not only contains the passenger activity signal that we are interested in, but also mixes various interference components such as vehicle driving vibration, air conditioner fan operation noise, and external environmental impact (such as road bumps). The strength of these interference signals is usually much greater than the weak pressure fluctuations caused by passenger activities, so a delicate signal processing procedure is required to extract the effective feature signal.

[0037] In one specific embodiment of the present disclosure, the original pressure signal is pre-processed to extract an effective pressure fluctuation signal related to occupant activity, specifically including: performing band-pass filtering on the original pressure signal to filter out signal components with frequencies lower than a first preset threshold and higher than a second preset threshold, to obtain a pressure fluctuation signal in a target frequency band; wherein the target frequency band contains a characteristic pressure fluctuation frequency range induced by human activity; performing noise reduction processing on the pressure fluctuation signal in the target frequency band to obtain a preliminary denoising signal; based on a reference noise signal preset by a vehicle vibration sensor, estimating and eliminating noise components related to vehicle vibration from the preliminary denoising signal through an adaptive filtering algorithm, and outputting the effective pressure fluctuation signal.

[0038] In one specific embodiment of the present disclosure, the pressure fluctuation signal in the target frequency band is processed to obtain a preliminary denoising signal, specifically including: using wavelet transform to perform multi-resolution analysis on the pressure fluctuation signal in the target frequency band to decompose the pressure fluctuation signal into sub-band signals of different frequency scales; based on a preset threshold rule, quantizing the wavelet coefficients of each sub-band signal to remove transient impulse interference, and reconstructing each sub-band signal after the quantization processing to generate the preliminary denoising signal.

[0039] It can be understood that the pre-processing process of the present embodiment first performs band-pass filtering on the original pressure signal. Specifically, a target frequency band containing the characteristic frequency range of human activity (e.g. 0.1 Hz to 10 Hz) is set, and signal components with frequencies lower than a first preset threshold and higher than a second preset threshold are filtered out through the band-pass filter. Exemplarily, setting the low-frequency cutoff point at 0.1 Hz can effectively remove the steady-state low-frequency vibration generated when the vehicle is driving at a constant speed, and setting the high-frequency cutoff point at 10 Hz can filter out the high-frequency noise generated by the rotation of the air conditioner fan blades. After this processing, the target frequency band pressure fluctuation signal has preliminarily highlighted the characteristic components related to human activity, such as the periodic fluctuations of about 0.2 Hz-0.3 Hz generated by smooth breathing, or the transient fluctuations of 1 Hz-3 Hz generated by hand movements.

[0040] Further, after the initial band-pass filtering is completed, the system further performs noise reduction processing on the pressure fluctuation signal of the target frequency band, and in this embodiment, the wavelet transform technology is specifically used. It can be understood that the wavelet transform has the characteristic of multi-resolution analysis, and can decompose the signal into sub-band signals of different frequency scales. Specifically, the system performs multi-level wavelet decomposition on the signal to obtain multiple sub-band coefficients from low frequency to high frequency. Exemplarily, when instantaneous pulse interference such as opening the door, bumping over the bump, etc. occurs, these interferences will generate wavelet coefficients with significant amplitude in certain high-frequency sub-bands. Through the preset threshold rule, the coefficients are quantitatively processed, the coefficients representing noise are set to zero or reduced, and the coefficients representing the real passenger activity are reserved, and finally the preliminary de-noising signal is generated through wavelet reconstruction. This processing method can effectively retain important features such as periodic fluctuations caused by conversations, while removing sudden interference.

[0041] It should be noted that, in order to further improve the signal quality, the system also introduces a reference noise signal based on the vehicle vibration sensor, and performs secondary noise reduction through an adaptive filtering algorithm. Specifically, the accelerometer installed on the vehicle body collects the vehicle vibration signal in real time as a reference input, and the adaptive filter estimates and eliminates the noise components related to the vehicle vibration from the preliminary de-noising signal by continuously adjusting its parameters. Exemplarily, when the vehicle is driving on rough road, the vibration sensor will detect mechanical vibration of a certain frequency, and the adaptive filter will find and eliminate the noise of the same frequency component in the pressure signal, thereby outputting the final effective pressure fluctuation signal. This processing method can effectively deal with vibration interference under different speeds and different road conditions, and ensure that the extracted signal truly reflects the passenger activity state.

[0042] Step 102, performing scene probability calculation based on the effective pressure fluctuation signal to determine the recognition result of the current passenger state in the vehicle.

[0043] In this embodiment, it can be understood that the preprocessed effective pressure fluctuation signal has removed most of the interference, but it is still an abstract time series data, which needs to be converted into meaningful passenger state information through special feature extraction and pattern recognition.

[0044] In one specific embodiment of the present disclosure, scene probability calculation is performed based on the effective pressure fluctuation signal to determine the recognition result of the current in-vehicle occupant state, specifically including: extracting time domain features and frequency domain features from the effective pressure fluctuation signal to construct a to-be-recognized feature vector representing the current in-vehicle state; wherein the time domain features include the amplitude, mean and variance of the signal waveform, and the frequency domain features include the signal spectrum features obtained by fast Fourier transform; inputting the to-be-recognized feature vector into the machine learning model to perform local feature perception and hierarchical abstraction on the to-be-recognized feature vector, and outputting a probability distribution corresponding to different occupant scenes; determining the occupant scene with the highest probability as the recognition result of the current in-vehicle occupant state.

[0045] In one specific embodiment of the present disclosure, time domain features and frequency domain features are extracted from the effective pressure fluctuation signal, specifically including: performing window sliding sampling on the effective pressure fluctuation signal; in the time domain, calculating the amplitude, mean and variance of the signal waveform in each time window to obtain the statistical distribution characteristics of the effective pressure fluctuation signal in the time domain; in the frequency domain, performing fast Fourier transform on the signal segment in each time window to convert the time domain signal into a frequency energy spectrum, and calculating the energy integral and dominant frequency component of the energy spectrum in the preset human activity frequency band to obtain the energy distribution characteristics of the effective pressure fluctuation signal in the frequency domain.

[0046] It can be understood that for the above recognition process, the system first extracts discriminative time domain and frequency domain features from the effective pressure fluctuation signal. Specifically, this process uses a sliding time window to segment and analyze the continuous pressure signal. For example, the system segments the signal into pieces of a certain duration, such as several seconds, and performs feature mining from both time and frequency domains.

[0047] In the time domain, the system calculates statistical features such as the amplitude, mean and variance of the signal waveform in each window, which can effectively reflect the intensity level and variation amplitude of the pressure fluctuation. For example, when the in-vehicle occupant performs relatively intense activities, such as moving in the back row, the amplitude and variance values in the time domain will usually be significantly higher than those in the single-person sitting state.

[0048] In the aspect of frequency domain feature extraction, the system performs a fast Fourier transform on the signal segment within each time window, converting the time domain signal into a frequency domain energy spectrum. It can be understood that this time-frequency conversion can reveal the periodic patterns and dominant frequency components hidden in the stress fluctuation. Illustratively, the system calculates the energy integration of the energy spectrum in the preset human activity frequency band and analyzes the dominant frequency distribution. For example, when identifying a front row multi-person conversation scene, the frequency domain analysis usually finds a significant energy concentration in the range of 1 Hz to 4 Hz, which is consistent with the syllable frequency characteristics of human speech. These time and frequency domain features together constitute a multi-dimensional feature vector, which comprehensively represents the unique "stress fingerprint" of the current in-vehicle state.

[0049] Further, after inputting the constructed feature vector into the pre-trained machine learning model, the model abstracts and understands the input features through its deep network structure. It should be noted that the machine learning model preferably adopts a one-dimensional convolutional neural network or a long short-term memory network, which is good at learning local features and long-term dependencies from time series data. Specifically, the model automatically learns the internal relationship between features through multiple layers of convolution kernels and finally outputs a probability distribution corresponding to different occupant scenes. Illustratively, the model may output a probability distribution indicating that the current state has an 85% probability of being "two people talking in the front row", a 10% probability of being "three people sitting in the front row", and the remaining probability is distributed in other scenes. The system determines the occupant scene with the highest probability as the final recognition result.

[0050] It can be understood that this recognition mechanism based on probability output greatly improves the robustness and fault tolerance of the system. Even in some complex scenarios, the model can still make the most reasonable judgment by comparing the relative probability sizes of various scenes. Illustratively, when there are both adults and children in the vehicle, the model may not explicitly distinguish specific individuals, but focus on identifying the overall activity pattern, such as outputting a scene judgment of "irregular movement exists in the back row", which is sufficient to support subsequent environmental control decisions. Through this intelligent recognition process, the system successfully converts the abstract stress fluctuation signal into specific and usable occupant state information, providing a key basis for subsequent precise environmental control.

[0051] In one specific embodiment of the present disclosure, the method further comprises pre-training the machine learning model, specifically including: collecting a plurality of sample stress signals acquired by the stress sensor under different known occupant scenes, and labeling each sample stress signal to form a training data set containing signal data and corresponding scene labels; using the training data set to iteratively train the machine learning model, and adjusting the convolution kernel weights and bias parameters in the model through a back propagation algorithm until the scene recognition accuracy of the model output reaches a preset convergence threshold.

[0052] It can be understood that a machine learning model capable of accurately identifying the occupant state must be fully data trained and parameter optimized to achieve the expected recognition accuracy in actual application. It should be noted that the training process is independent of the daily use of the vehicle and is completed in the model building stage before system deployment, but the model parameters obtained by training will be fixed as part of the system for real-time occupant state recognition.

[0053] Specifically, the training process first needs to build a high-quality training data set. Exemplarily, data collection is carried out under a plurality of known occupant scenarios, and a large number of sample pressure signals are obtained through a pressure sensor arranged in the air duct of the vehicle air conditioner. These known scenarios include but are not limited to single person sitting, front row multiple people talking, rear row people moving, rear row children moving, and different combinations of people, etc. It can be understood that each sample pressure signal needs to be accurately labeled, forming a one-to-one correspondence between signal data and corresponding scenario label. Exemplarily, when collecting data for the "front row multiple people talking" scenario, the actual number of occupants in the vehicle, the position distribution and the activity type need to be recorded synchronously, and these information are taken as the true label of this segment of pressure signal data. This strict labeling ensures the reliability of the training data and provides accurate supervision signals for model learning.

[0054] Further, after obtaining a sufficient number of labeled data, the system enters the model training stage. Specifically, the labeled training data set is input into the machine learning model to be trained in the training process, the difference between the model output and the true label is calculated through forward propagation, and the convolution kernel weight and bias parameters in the model are adjusted layer by layer using the back propagation algorithm. Exemplarily, in the early stage of training, the recognition accuracy of the model may be low, but as the number of iterations increases, the model gradually learns the internal mapping relationship between different occupant scenarios and specific pressure fluctuation patterns by continuously reducing the error between the prediction result and the true label.

[0055] It can be understood that the training process needs to continue until the scenario recognition accuracy of the model on the validation set reaches the preset convergence threshold. Exemplarily, the convergence threshold can be set to a recognition accuracy of ninety percent or higher according to actual application requirements. It should be noted that the validation set is a data subset that is separately divided from the collected data and does not participate in the training, and is used to objectively evaluate the generalization ability of the model. This training mechanism ensures that the model not only remembers the training samples, but more importantly, understands the essential characteristics of different occupant scenarios, so that it can make accurate judgments when facing new, unseen pressure signals.

[0056] Specifically, the well-trained model finally masters the recognition ability of various "stress fingerprints". For example, the model learns that "single person sitting still" usually corresponds to low-frequency and low-amplitude smooth fluctuations, while "children moving in the back row" shows high-frequency irregular and amplitude-varying fluctuation patterns. This feature recognition ability established through data-driven approach enables the system to go beyond traditional rule-based judgment and achieve more intelligent and accurate passenger state recognition.

[0057] Step 103, based on the recognition result, generating a control signal for controlling at least one execution mechanism in the vehicle to drive the execution mechanism to perform environment control based on the passenger state.

[0058] It can be understood that the passenger state recognition result obtained through the foregoing steps finally reflects its value in the precise regulation of the in-vehicle environment system, and this step completes the closed loop from "perception and understanding" to "decision and execution".

[0059] In one specific embodiment of the present disclosure, based on the recognition result, generating a control signal for controlling at least one execution mechanism in the vehicle, specifically including: analyzing the passenger number and position distribution information in the recognition result, generating independent air volume control instructions for different air outlets of the air conditioning system; wherein the control instruction contains a target air volume value corresponding to the passenger distribution area; and / or, analyzing the passenger activity type information in the recognition result, combining the pre-stored activity type and fresh air demand mapping relationship, generating a mode switching instruction for controlling the air conditioning fresh air mode and air volume to adjust the air circulation state in the vehicle.

[0060] In one specific embodiment of the present disclosure, analyzing the passenger activity type information in the recognition result to generate a mode switching instruction, specifically including: when the identified activity type is a pre-set high-metabolic activity type, generating a first control instruction to increase fresh air volume and switch to an external circulation mode to improve the air exchange efficiency in the vehicle; when the identified activity type is a pre-set low-metabolic activity type, generating a second control instruction to maintain or reduce fresh air volume and switch to an internal circulation mode to optimize the energy utilization efficiency of the air conditioning system.

[0061] It can be understood that for the above-mentioned passenger state-based environment control process, the system first analyzes the passenger number and position distribution information contained in the recognition result, and generates refined control instructions for the air conditioning system accordingly. Specifically, the system controls different air outlets of the air conditioning system independently according to the actual distribution of passengers. For example, when the system identifies that the back row is unoccupied, it will automatically generate a control instruction to reduce or close the air volume of the back row outlet; on the contrary, when it detects passenger activity in the back row, it will maintain or appropriately increase the air volume output in this area. This precise control strategy based on position distribution not only ensures passenger comfort, but also optimizes energy utilization.

[0062] Further, the system also deeply analyzes the occupant activity type information and matches it with the pre-stored activity type and fresh air demand mapping relationship. It should be noted that the "fresh air" refers to fresh air introduced from the outside of the vehicle without being circulated in the vehicle, and its introduction or not and air volume directly affect the oxygen content and carbon dioxide concentration of the cabin air. It can be understood that different types of occupant activities correspond to different metabolic levels and air quality requirements. The "high metabolic activity type" refers to activities that cause a significant increase in human oxygen consumption and carbon dioxide emissions, such as continuous conversation, body movement, or multi-person interaction; while the "low metabolic activity type" refers to a state of relative stillness and stable metabolism of the human body, such as single person sitting, resting or sleeping. For example, when identifying the high metabolic activity type of "front row multi-person conversation", the system generates control instructions to increase fresh air volume and switch to the external circulation mode, which timely discharges carbon dioxide by improving the air exchange efficiency in the vehicle, avoiding the occupants from feeling stuffy or tired. Conversely, when identifying the low metabolic activity type of "single person sitting", the system generates instructions to maintain or reduce fresh air volume and switch to the internal circulation mode, which can not only maintain the basic air quality, but also significantly optimize the energy utilization efficiency of the air conditioning system.

[0063] It should be noted that the control range of the system is not limited to the air conditioning system, but can also be extended to other vehicle body domain systems. For example, when the system identifies the "single person sitting" scenario, in addition to optimizing the air conditioning settings, it can also cooperatively control the ambient light system to dim the light tone, control the audio system to play soothing music, and create a more quiet and comfortable driving environment. When identifying "rear row child movement" with high activity intensity, the system can determine that there may be an irritable mood, and then cooperatively play soothing content with the audio system, while fine-tuning the air conditioning parameters to provide a softer air flow experience.

[0064] It can be understood that all these control decisions are intelligent responses based on a deep understanding of the occupant state. Specifically, a complete scenario-control mapping knowledge base is established inside the system, which associates different recognition results with the optimal control strategy. For example, when the system simultaneously identifies "front row one person driving" and "strong sunlight shining on the driver area", it will generate enhanced air supply instructions for the driver area based on these information, and appropriately adjust the air outlet direction to avoid direct blowing. This multi-dimensional comprehensive consideration reflects the high-level intelligence of the system. Through this precise environment control based on the actual occupant state, the system ultimately realizes the intelligent upgrade of the vehicle interior environment from "unified control" to "on-demand distribution", significantly improving the driving comfort and energy utilization efficiency.

[0065] With regard to the apparatus in the above embodiments, the specific manner in which each unit performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

[0066] Figure 2 is a structural schematic diagram of a vehicle provided by an embodiment of the present application.

[0067] As shown in the example, Figure 2 The vehicle 200 includes a memory 201 and a processor 202, where the memory 201 stores executable program code 2011, and the processor 202 is configured to invoke and execute the executable program code 2011 to perform an in-vehicle environment adjustment method.

[0068] The embodiment can divide the vehicle into functional modules according to the above-mentioned in-vehicle environment adjustment method example. For example, each functional module can be provided, or two or more functions can be integrated into one processing module. The integrated module can be implemented in the form of hardware. It should be noted that the division of the modules in the embodiment is illustrative, and is only a logical functional division. In actual implementation, another division manner can be used.

[0069] In the case where each functional module is divided according to each function, the vehicle can include: obtaining an original pressure signal collected by a pressure sensor in a vehicle air conditioning air duct, and preprocessing the original pressure signal to extract an effective pressure fluctuation signal related to the activity of the occupant; where the sampling frequency of the pressure sensor is configured to be able to capture the pressure fluctuation caused by the activity of the occupant; inputting the original pressure signal into a pre-trained machine learning model to output an identification result of the current in-vehicle occupant state; based on the identification result, generating a control signal for controlling at least one execution mechanism in the vehicle to drive the execution mechanism to perform environment control based on the occupant state.

[0070] Some embodiments of the present application provide a computer-readable storage medium corresponding to Figure 1 The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are configured to: obtaining an original pressure signal collected by a pressure sensor in a vehicle air conditioning air duct, and preprocessing the original pressure signal to extract an effective pressure fluctuation signal related to the activity of the occupant; where the sampling frequency of the pressure sensor is configured to be able to capture the pressure fluctuation caused by the activity of the occupant; inputting the original pressure signal into a pre-trained machine learning model to output an identification result of the current in-vehicle occupant state; Based on the identification result, a control signal for controlling at least one actuator in the vehicle is generated to drive the actuator to perform environment control based on the occupant state.

[0071] The various embodiments in the present application are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, the IoT device and medium embodiments are basically similar to the method embodiments, and thus are described simply. The relevant parts can be referred to the description of the method embodiments.

[0072] The system and medium provided by the embodiments of the present application are one-to-one corresponding to the method, and thus the system and medium have similar beneficial technical effects to the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system and medium will not be described here.

[0073] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer-usable program code.

[0074] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a device that implements the flow Figure 1 The function specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the function specified in one block or multiple blocks.

[0075] These computer program instructions can also be stored in a computer-readable memory that can guide the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction devices that implement the flow Figure 1 The function specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the function specified in one block or multiple blocks.

[0076] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 Figure 1 The flowchart blocks or blocks in the multiple flows and / or blocks

[0077] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0078] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) about which the computer stores information about an operating system, application software, and / or the like. Memory is an example of computer readable media.

[0079] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0080] It should also be noted that the terms "comprising", "containing" or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements does not include only those elements recited, but can also include other elements not expressly listed or inherent to such process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0081] ​The above merely provides an example of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the scope of claims of the present application.

Claims

1. A method for regulating the in-vehicle environment, characterized in that, The method includes: The system acquires raw pressure signals collected by pressure sensors inside the vehicle's air conditioning duct and preprocesses the raw pressure signals to extract effective pressure fluctuation signals related to occupant activities; wherein the sampling frequency of the pressure sensor is configured to capture pressure fluctuations caused by occupant activities. Based on the effective pressure fluctuation signal, the scene probability is calculated to determine the identification result of the current occupant status in the vehicle; Based on the identification results, a control signal is generated to control at least one actuator inside the vehicle, so as to drive the actuator to perform environmental control based on the occupant status.

2. The in-vehicle environment adjustment method according to claim 1, characterized in that, The raw pressure signal is preprocessed to extract effective pressure fluctuation signals related to occupant activity, specifically including: The original pressure signal is subjected to bandpass filtering to filter out signal components with frequencies lower than a first preset threshold and higher than a second preset threshold, thereby obtaining a pressure fluctuation signal in the target frequency band; wherein, the target frequency band includes the characteristic pressure fluctuation frequency range caused by human activity; The pressure fluctuation signal in the target frequency band is subjected to noise reduction processing to obtain a preliminary denoised signal; Based on the reference noise signal preset by the vehicle vibration sensor, the noise component related to vehicle vibration is estimated and eliminated from the preliminary denoising signal through an adaptive filtering algorithm, and the effective pressure fluctuation signal is output.

3. The in-vehicle environment adjustment method according to claim 2, characterized in that, The pressure fluctuation signal in the target frequency band is subjected to noise reduction processing to obtain a preliminary denoised signal, specifically including: Wavelet transform is used to perform multi-resolution analysis on the pressure fluctuation signal in the target frequency band, so as to decompose the pressure fluctuation signal into sub-band signals of different frequency scales; Based on a preset threshold rule, the wavelet coefficients of each sub-band signal are quantized to remove transient pulse interference, and the quantized sub-band signals are reconstructed to generate the preliminary denoised signal.

4. The in-vehicle environment adjustment method according to claim 1, characterized in that, Based on the effective pressure fluctuation signal, a scenario probability calculation is performed to determine the identification result of the current occupant status inside the vehicle, specifically including: Time-domain and frequency-domain features are extracted from the effective pressure fluctuation signal to construct a feature vector to be identified that characterizes the current in-vehicle state; wherein, the time-domain features include the amplitude, mean, and variance of the signal waveform, and the frequency-domain features include the signal spectrum features obtained by fast Fourier transform; The feature vector to be identified is input into the machine learning model to perform local feature perception and hierarchical abstraction on the feature vector to be identified, and output the probability distribution corresponding to different occupant scenarios; The occupant scenario with the highest probability is determined as the identification result of the current occupant status in the vehicle.

5. The in-vehicle environment adjustment method according to claim 4, characterized in that, Extracting time-domain and frequency-domain features from the effective pressure fluctuation signal specifically includes: The effective pressure fluctuation signal is subjected to window sliding sampling; In the time domain, the amplitude, mean, and variance of the signal waveform within each time window are calculated to obtain the statistical distribution characteristics of the effective pressure fluctuation signal in the time domain. In the frequency domain, a fast Fourier transform is performed on the signal segment within each time window to convert the time-domain signal into a frequency-domain energy spectrum. The energy integral and dominant frequency component of the energy spectrum within a preset human activity frequency band are then calculated to obtain the energy distribution characteristics of the effective pressure fluctuation signal in the frequency domain.

6. The in-vehicle environment adjustment method according to claim 1, characterized in that, Based on the recognition results, a control signal is generated for controlling at least one actuator inside the vehicle, specifically including: The system analyzes the occupant number and location distribution information from the identification results to generate independent airflow control commands for different air outlets of the air conditioning system; wherein, the control commands include target airflow values ​​corresponding to the occupant distribution areas; and / or, The occupant activity type information in the identification results is analyzed, and combined with the pre-stored mapping relationship between activity type and fresh air demand, a mode switching command for controlling the air conditioning fresh air mode and air volume is generated to adjust the air circulation state inside the vehicle.

7. The in-vehicle environment adjustment method according to claim 6, characterized in that, Parsing the occupant activity type information in the recognition result and generating a mode switching instruction specifically includes: When the identified activity type is a preset high-metabolic activity type, a first control command is generated to increase the fresh air volume and switch to the external circulation mode to improve the air exchange efficiency in the vehicle. When the identified activity type is a preset low-metabolic activity type, a second control command is generated to maintain or reduce the fresh air volume and switch to the internal circulation mode in order to optimize the energy utilization efficiency of the air conditioning system.

8. The in-vehicle environment adjustment method according to claim 1, characterized in that, The method further includes: Pre-training the machine learning model specifically includes: Multiple sample pressure signals were collected by the pressure sensor under different known occupant scenarios, and each sample pressure signal was labeled to form a training dataset containing signal data and corresponding scenario labels. The machine learning model is iteratively trained using the training dataset, and the convolution kernel weights and bias parameters in the model are adjusted using the backpropagation algorithm until the scene recognition accuracy of the model output reaches the preset convergence threshold.

9. A vehicle, characterized in that, The vehicles include: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform an in-vehicle environment adjustment method as described in any one of claims 1-8.

10. A computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, they implement a method for adjusting the in-vehicle environment as described in any one of claims 1-8.

Images