WO2022095331A1 - 压力评估方法、装置、计算机设备和存储介质 - Google Patents

压力评估方法、装置、计算机设备和存储介质 Download PDF

Info

Publication number
WO2022095331A1
WO2022095331A1 PCT/CN2021/084544 CN2021084544W WO2022095331A1 WO 2022095331 A1 WO2022095331 A1 WO 2022095331A1 CN 2021084544 W CN2021084544 W CN 2021084544W WO 2022095331 A1 WO2022095331 A1 WO 2022095331A1
Authority
WO
WIPO (PCT)
Prior art keywords
stress
hrv
test
user
sample
Prior art date
Application number
PCT/CN2021/084544
Other languages
English (en)
French (fr)
Inventor
冯澍婷
庄伯金
王少军
肖京
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2022095331A1 publication Critical patent/WO2022095331A1/zh

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/0205—Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition

Definitions

  • the present application relates to the field of artificial intelligence technology, and in particular, to a stress assessment method, device, computer equipment and storage medium.
  • Stress also known as stress, is a state in which the physiological system exhibits specific symptoms in response to stimuli.
  • physiological information fluid hormone levels, heart rate, blood pressure, respiration, pupil diameter, electrical skin, etc.
  • physical behavior information expression, voice, and body movements, etc.
  • the evaluation methods of stress index can be divided into four categories: biochemical method, subjective evaluation method, physiological response test method, and physiological parameter measurement method.
  • Biochemical methods need to extract human body fluids (blood, urine, saliva, etc.), which have high requirements on equipment and operators, and can only determine the pressure state at the time of body fluid collection, and cannot be used for long-term pressure monitoring.
  • the subjective evaluation method relies on the description of the subject's self-feeling, and is generally scored by professional scales. This method is simple and easy to implement and is widely used; however, this method is more suitable for statistical analysis of large samples and for long-term evaluation. Individuals are susceptible to environmental and short-term memory confounds when stressed.
  • the physiological response test method evaluates the stress of the test subject through various test methods such as color words and quick calculation tests, and quantifies the stress level according to the test score or physiological response; this method is affected by individual differences and has higher requirements on the test environment , During long-term pressure monitoring, the daily test content must be appropriately changed, and the implementation cost is high.
  • the physiological parameter measurement method mainly evaluates the pressure by measuring and analyzing the physiological signals (EEG, OMG, ECG, PPG, etc.)
  • the collection equipment is easy to use, and the pressure level is more objectively assessed by physiological signals, which is not affected by supervisory factors.
  • the inventor realized that the physiological parameter measurement method is most suitable for long-term stress evaluation, but when using this method to collect physiological signals, the cost of professional signal acquisition equipment is required, and The equipment is usually expensive and the operation of the equipment is complicated, which hinders the widespread promotion of this method.
  • the main purpose of the present application is to provide a stress assessment method, device, computer equipment and storage medium, aiming at overcoming the defect that professional signal acquisition equipment is required when currently collecting physiological signals for stress assessment.
  • the present application provides a stress assessment method, comprising the following steps:
  • the fingertip video of the test user is collected based on the mobile terminal; wherein, the fingertip video is obtained when the mobile terminal flash is turned on and the test user's finger blocks the camera;
  • the fingertip video is converted into RGBA video format, and the summation of all pixels in the R channel is extracted to obtain a PPG signal;
  • the application also provides a pressure assessment device, comprising:
  • a collection unit configured to collect the fingertip video of the test user based on the mobile terminal; wherein, the fingertip video is obtained when the test user's finger blocks the camera when the flashlight of the mobile terminal is turned on;
  • an extraction unit for extracting and obtaining the HRV feature of the test user based on the PPG signal
  • the evaluation unit is configured to input the HRV feature of the test user into a preset stress evaluation model, and obtain the stress evaluation result of the test user; wherein, the stress evaluation model is based on the HRV feature training set training time series processing obtained by the neural network.
  • the present application also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements a stress assessment method when executing the computer program, including the following steps:
  • the fingertip video of the test user is collected based on the mobile terminal; wherein, the fingertip video is obtained when the mobile terminal flash is turned on and the test user's finger blocks the camera;
  • the fingertip video is converted into RGBA video format, and the summation of all pixels in the R channel is extracted to obtain a PPG signal;
  • the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements a stress assessment method, comprising the following steps:
  • the fingertip video of the test user is collected based on the mobile terminal; wherein, the fingertip video is obtained when the mobile terminal flash is turned on and the test user's finger blocks the camera;
  • the fingertip video is converted into RGBA video format, and the summation of all pixels in the R channel is extracted to obtain a PPG signal;
  • the stress assessment method, device, computer equipment and storage medium provided by the present application collect the fingertip video of the test user based on the mobile terminal; wherein, the fingertip video is when the mobile terminal flash is turned on and the test user's finger blocks the camera. Capture the income; convert the fingertip video into RGBA video format, and extract the summation of all pixels in the R channel to obtain a PPG signal; based on the PPG signal, extract and obtain the HRV feature of the test user; The HRV characteristics of the test user are input into a preset stress evaluation model to obtain a stress evaluation result of the test user.
  • the present application only needs to use a mobile terminal to collect physiological signals, and does not require professional collection equipment, thereby reducing the collection cost.
  • FIG. 1 is a schematic diagram of steps of a pressure assessment method in an embodiment of the present application
  • FIG. 2 is a structural block diagram of a pressure evaluation device in an embodiment of the present application.
  • FIG. 3 is a schematic structural block diagram of a computer device according to an embodiment of the present application.
  • an embodiment of the present application provides a stress assessment method, including the following steps:
  • Step S1 based on the mobile terminal to collect the fingertip video of the test user; wherein, the fingertip video is obtained when the test user's finger blocks the camera when the flashlight of the mobile terminal is turned on;
  • Step S2 the fingertip video is converted into an RGBA video format, and the summation of all pixels in the R channel is extracted to obtain a PPG signal;
  • Step S3 based on the PPG signal, extract the HRV feature of the test user;
  • Step S4 inputting the HRV feature of the test user into a preset stress evaluation model to obtain a stress evaluation result of the test user; wherein, the stress evaluation model trains a neural network for time series processing based on the HRV feature training set income.
  • the above method is applied to assess the user's current stress, and the stress assessment is of great significance to the user's mental health.
  • the above method can also be applied to the field of smart medical technology in smart cities to promote the construction of smart cities.
  • a mobile terminal such as a mobile phone, tablet, etc.
  • the PPG signal of the test user is obtained.
  • the test user uses a finger to completely cover the camera to record the video of the blood flow of the fingertip. Since the blood flow of the fingertip will fluctuate periodically with the change of the blood vessel volume, the effect of blood on the flash light source The amount of absorption also changes, resulting in periodic changes in the pixel values of the recorded video.
  • each frame of video into RGBA format, and calculate the sum of all pixels in the R channel to obtain a discrete PPG signal, that is, each frame of video corresponds to a sampling point of the PPG signal, if the video frame rate is 30fp/s, then The PPG sampling rate is 30Hz.
  • Heart rate variability refers to the small difference between heartbeats, which can not only reflect the activity of the human autonomic nervous system, but also feedback the abnormality of the cardiovascular system. It is an effective indicator to measure the human heart activity. Specifically, short-term HRV showed a downward trend under the action of stressors, and was basically negatively correlated with stress. HRV power spectrum Low-frequency power LF, significantly decreased when sympathetic activity decreased, while high-frequency power HF, increased significantly with parasympathetic activity. Therefore, LF and HF can be used to quantitatively assess sympathetic and parasympathetic activity, respectively, The ratio LF/HF can be used to assess the balance of the autonomic nervous system.
  • HRV feature parameter acquisition difficulty and cost are relatively low.
  • the above-mentioned stress evaluation model is based on the training set of HRV feature training set to train a neural network for time series processing, and the HRV feature of the above-mentioned test user is input into a preset stress evaluation model to obtain the above-mentioned stress evaluation of the test user. result.
  • the above HRV features can be obtained by periodic short-term measurement, and the pressure is evaluated by synthesizing HRV features of multiple short-term measurements, which can avoid random errors caused by single data fluctuations, and the pressure evaluation results are more robust.
  • HRV features can also be extracted from ECG and BCG signals, so as to evaluate the user stress result.
  • the step of extracting and obtaining the HRV feature of the test user based on the PPG signal includes:
  • the HRV feature of the test user is extracted.
  • the sampling rate of the device is different, and the sampling rate of the obtained PPG signal is different. Therefore, in this solution, the sampling rate of the PPG signal needs to be converted into a specified sampling rate by up-sampling or down-sampling. (such as 250Hz), in the process of up-sampling or down-sampling, in order to ensure that the waveform shape is not deformed, spline interpolation or least squares interpolation method is usually used.
  • the RR interval of the noise signal segment is invalid, so in this embodiment, the RR interval of the noise segment is processed (eliminated or interpolated) to avoid affecting subsequent HRV feature extraction. Finally, only the HRV features of sinus beats are valid. In order to avoid the abnormal cardiac cycle affecting the calculation of HRV features, in this example, the RR intervals with large differences (for example: the difference exceeds 20% or 30%) are also calculated. Same culling or interpolation processing. After obtaining the RR interval of the sinus beat, the HRV feature of the test user can be extracted.
  • the HRV features include HRV time domain parameters and HRV frequency domain parameters
  • the HRV time domain parameter includes at least the standard deviation SDNN of the RR interval of the sinus beat
  • the HRV frequency domain parameters at least include low frequency power LF, high frequency power HF, and the ratio of LF to HF.
  • the low frequency power LF is the low frequency energy of the RR interval power spectrum
  • the high frequency power HF is the high frequency energy of the RR interval power spectrum.
  • the HRV time domain parameters further include rMSSD and PNN50
  • the HRV frequency domain parameters further include parameters such as LF_nu (low frequency) and HF_nu (high frequency).
  • the method before step S4 of inputting the HRV characteristics of the test user into the preset stress evaluation model, the method includes:
  • Step S10 collect the physiological signals of the sample user based on the preset collection period and the preset collection duration, and obtain the pressure measurement result of the sample user each time the physiological signal is collected; wherein, the pressure measurement result is based on the sample user's Measured by the preset pressure gauge;
  • Step S20 obtaining HRV characteristics corresponding to the sample user based on the physiological signals collected each time;
  • Step S30 taking the HRV features corresponding to the physiological signals collected by the sample user each time and the corresponding pressure measurement results as a set of HRV feature training sets;
  • Step S40 according to the time sequence of collecting physiological signals, input each group of the HRV feature training set into the neural network training of time series processing to obtain the stress evaluation model.
  • the above preset collection period and preset collection duration are determined according to the current application scenario.
  • Pressure is a variable that characterizes the state of the body. Unlike complex and changeable physiological signals, the pressure state of the human body is short-term stable and does not require real-time monitoring. Therefore, this embodiment only considers the user's periodic pressure detection, and the monitoring period is related to the application scenario. For example, when planning a day's work/study schedule reasonably, a short-term stress test needs to be performed, and the user's pressure should be monitored multiple times in a day. ; During long-term stress testing, monitoring is only required once every 1 to 2 days; during chronic stress testing, the monitoring period may be longer.
  • the collection time is 30 ⁇ 60s; when measuring once a day, the collection time is 1 ⁇ 5min; when once a week, the collection time >10min. In this embodiment, it is mainly used in the application scenario of measuring once a day.
  • the time for a single data collection is set to 2 minutes.
  • a mobile terminal which can collect PPG and BCG signals
  • the pressure gauge Then choose "PERCEIVED STRESS QUESTIONNAIRE" that can reflect the user's short-term pressure.
  • the physiological signal collection of the above-mentioned sample users is usually collected in the early morning when the sample users are not disturbed by external things, and the initial state of daily stress of the sample users is collected. In order to reflect the real stress state of the sample users in their daily life, no stress-inducing operation is performed.
  • the sample users are required to collect PPG signals with their mobile phones in a calm state at a fixed time period (eg 7 ⁇ 8am) every morning, and then fill in the pressure scale to obtain the pressure measurement results, which are the training data set labels.
  • the HRV feature corresponding to the sample user is obtained.
  • the extraction process of the HRV feature is the same as that in the above-mentioned embodiment, and details are not repeated here.
  • the HRV features corresponding to the physiological signals collected by the sample user each time and the corresponding stress measurement results are used as a set of HRV feature training sets, and the HRV feature training sets are used to train the above-mentioned time series processing neural network.
  • the neural network for time series processing includes any one of RNN, LSTM, GUR, and IndRNN.
  • the traditional BP neural network can only process the input features in sequence, and there is no relationship between the front and rear input features. Therefore, the BP neural network cannot reflect the relationship between the historical HRV features and the current pressure.
  • Recurrent Neural Network RNN can be divided into three parts: input layer, hidden layer and output layer.
  • RNN can be used, and the value St of the hidden layer of the RNN and the value Ot of the output layer are not only related to the current input xt, but also depend on the value St-1 of the previous hidden layer.
  • the above stress assessment model reflects the interaction between the front and back of the sequence data, and is more suitable for processing sequences with time series.
  • the above RNN unit can be replaced with a variety of time series processing units with better performance such as LSTM/GUR/IndRNN.
  • Traditional RNNs are prone to gradient disappearance/explosion problems due to the sharing of parameters in time.
  • Traditional RNNs are difficult to reasonably explain the behavior of neurons due to the interconnected neurons in the layers.
  • LSTM/GRU can solve the problem of vanishing/exploding gradients in traditional RNN layers, while IndRNN improves the interpretability of neurons and can be used to process longer sequence information.
  • this model structure can be used not only for regression calculation of pressure index, but also for pressure classification judgment (for example: high pressure, low pressure, no pressure classification, etc.), just modify the output layer and loss function (for example: output layer
  • the activation function is set to softmax and the loss function is set to multi class cross entropy).
  • the step S10 of collecting the physiological signals of the sample user includes:
  • the fingertip video of the sample user is collected based on the mobile terminal; wherein, the fingertip video is captured when the sample user's finger blocks the camera when the flashlight of the mobile terminal is turned on;
  • the steps include:
  • the application scenario type of the stress test includes a short-term stress test, a long-term stress test, and a chronic stress test;
  • the types of the stress test include perceived stress and psychological stress pressure;
  • a preset pressure scale for measuring the pressure measurement result of the sample user is acquired.
  • the above-mentioned preset collection period and preset collection time are both determined according to the current application scenario, and pressure is a variable representing the physical state. It is not necessary to perform real-time monitoring, so in this embodiment, only the user's periodic pressure detection is considered, and the length of the monitoring period is related to the application scenario. Monitor user pressure multiple times within the system; during long-term stress testing, it only needs to be monitored once every 1 to 2 days; during chronic stress testing, the monitoring period may be longer.
  • the stress on the human body can be divided into many types according to different classification methods, such as: perceived stress and psychological stress; the above-mentioned perceived stress and psychological stress include work stress, life stress, study stress, training stress, chronic stress, sudden stress, physical stress induced and emotionally induced stress, etc.
  • perceived stress and psychological stress include work stress, life stress, study stress, training stress, chronic stress, sudden stress, physical stress induced and emotionally induced stress, etc.
  • Different application situations need to evaluate different types of stress, and scholars who study stress have designed corresponding evaluation scales or test methods for different types of stress.
  • the scale for perceived stress "PERCEIVED STRESS QUESTIONNAIRE", “Perceived Psychological Stress Scale (CPSS)” for psychological stress, etc.
  • CPSS Perceived Psychological Stress Scale
  • the method further includes:
  • the preset stress evaluation model, stress evaluation results, fingertip videos and HRV characteristics are stored in the blockchain.
  • blockchain is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • an embodiment of the present application further provides a pressure assessment device, including:
  • the collection unit 10 is configured to collect the fingertip video of the test user based on the mobile terminal; wherein, the fingertip video is obtained when the test user's finger blocks the camera when the flashlight of the mobile terminal is turned on;
  • Conversion unit 20 for converting the fingertip video into RGBA video format, and extracting the summation of all pixels in the R channel to obtain a PPG signal
  • Extraction unit 30 for extracting and obtaining the HRV feature of the test user based on the PPG signal
  • the evaluation unit 40 is configured to input the HRV feature of the test user into a preset stress evaluation model to obtain the stress evaluation result of the test user; wherein, the stress evaluation model is based on the HRV feature training set training time series processing obtained by the neural network.
  • the extraction unit 30 is specifically configured to:
  • the HRV feature of the test user is extracted.
  • the HRV features include HRV time domain parameters and HRV frequency domain parameters
  • the HRV time domain parameter includes at least the standard deviation SDNN of the RR interval of the sinus beat
  • the HRV frequency domain parameters at least include low frequency power LF, high frequency power HF, and the ratio of LF to HF.
  • the above-mentioned pressure evaluation device further includes:
  • a sample collection unit configured to collect the physiological signals of the sample users based on the preset collection period and the preset collection duration, and obtain the pressure measurement results of the sample users each time the physiological signals are collected; wherein the pressure measurement results are the Measured by sample users based on preset pressure gauges;
  • a sample feature extraction unit configured to obtain the HRV feature corresponding to the sample user based on the physiological signal collected each time
  • a training set construction unit used for taking the HRV feature corresponding to the physiological signal collected by the sample user each time and the corresponding stress measurement result as a set of HRV feature training set;
  • the training unit is configured to input each group of the HRV feature training set into the neural network training of time series processing according to the time sequence of collecting the physiological signals, so as to obtain the stress evaluation model.
  • the sample collection unit collects the physiological signals of the sample user, including:
  • the fingertip video of the sample user is collected based on the mobile terminal; wherein, the fingertip video is captured when the sample user's finger blocks the camera when the flashlight of the mobile terminal is turned on;
  • the above-mentioned pressure evaluation device further includes:
  • a type obtaining unit configured to obtain the application scenario type of the stress test and the type of the stress test; wherein, the application scenario type of the stress test includes a short-term stress test, a long-term stress test and a chronic stress test; Types include perceived stress as well as psychological stress;
  • a first obtaining unit configured to obtain the preset collection period and preset collection time for collecting the physiological signal of the sample user according to the corresponding relationship between the application scenario type of the stress test, the collection period and the collection time;
  • a second obtaining unit configured to obtain a preset pressure scale for measuring the pressure measurement result of the sample user according to the corresponding relationship between the type of the pressure test and the pressure scale.
  • the neural network for time series processing includes any one of RNN, LSTM, GUR, and IndRNN.
  • an embodiment of the present application further provides a computer device.
  • the computer device may be a server, and its internal structure may be as shown in FIG. 3 .
  • the computer device includes a processor, memory, a network interface and a database connected by a system bus.
  • the processor of the computer design is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium, an internal memory.
  • the nonvolatile storage medium stores an operating system, a computer program, and a database.
  • the internal memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium.
  • the database of the computer equipment is used to store stress assessment results and the like.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer program when executed by a processor, implements a stress assessment method comprising the following steps:
  • the fingertip video of the test user is collected based on the mobile terminal; wherein, the fingertip video is obtained when the mobile terminal flash is turned on and the test user's finger blocks the camera;
  • the fingertip video is converted into RGBA video format, and the summation of all pixels in the R channel is extracted to obtain a PPG signal;
  • FIG. 3 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied.
  • An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, a method for evaluating stress is implemented, including the following steps:
  • the fingertip video of the test user is collected based on the mobile terminal; wherein, the fingertip video is obtained when the mobile terminal flash is turned on and the test user's finger blocks the camera;
  • the fingertip video is converted into RGBA video format, and the summation of all pixels in the R channel is extracted to obtain a PPG signal;
  • the computer-readable storage medium in this embodiment may be a volatile readable storage medium, and may also be a non-volatile readable storage medium.
  • the fingertip video of the test user is collected based on the mobile terminal; wherein, the fingertip video is when the flashlight of the mobile terminal is turned on.
  • the test results are obtained when the user's finger blocks the camera; the video of the fingertip is converted into RGBA video format, and the sum of all pixels in the R channel is extracted to obtain the PPG signal; based on the PPG signal, the extraction is obtained.
  • HRV characteristics of the test user inputting the HRV characteristics of the test user into a preset stress evaluation model to obtain a stress evaluation result of the test user.
  • the present application only needs to use a mobile terminal to collect physiological signals, and does not require professional collection equipment, thereby reducing the collection cost.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Animal Behavior & Ethology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Physiology (AREA)
  • Cardiology (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Psychiatry (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Developmental Disabilities (AREA)
  • Child & Adolescent Psychology (AREA)
  • Social Psychology (AREA)
  • Psychology (AREA)
  • Hospice & Palliative Care (AREA)
  • Educational Technology (AREA)
  • Pulmonology (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

一种压力评估方法、装置、计算机设备和存储介质，涉及人工智能技术领域，该方法包括：基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得（S1）；将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号（S2）；基于所述PPG信号，提取得到所述测试用户的HRV特征（S3）；将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得（S4）。该方法在测试用户的压力时，只需要用到移动终端进行生理信号的采集，不需要专业的采集设备，降低采集成本。

Description

压力评估方法、装置、计算机设备和存储介质
本申请要求于2020年11月09日提交中国专利局、申请号为2020112404444，发明名称为“压力评估方法、装置、计算机设备和存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域，特别涉及一种压力评估方法、装置、计算机设备和存储介质。
 背景技术
随着社会经济的快速发展，人们的工作、生活节奏越来越快，来自各方面的压力也越来越大，很多人在不同时期，都会呈现出不同程度、不同诱因的压力问题。压力长期得不到释放、缓解时，会严重影响身体与精神健康，轻则引起头痛、失眠焦虑等一系列不适，重则会增加心脑血管疾病、糖尿病和癌症等一系列慢病风险。由此可见，了解并掌握自身压力状态，及时采取减压措施，对于个体身心健康至关重要。
压力也称为应激，是生理系统应对刺激时表现出特殊症状的一种状态。当一个人处于压力状态时，不仅其生理信息(体液激素水平、心率、血压、呼吸、瞳孔直径、皮肤电等)会有变化，身体行为信息(表情、声音和肢体动作等)也会发生变化。现有技术通过检测这两类状态的变化即可评估人体压力水平。
压力指数评估方法可分为：生化法、主观评定法、生理反应测试法、生理参数测定法四类。
1、生化法需要提取人体体液(血液，尿液，唾液等)，对设备与操作人员要求都很高，且只能判定体液采集时刻的压力状态，无法用于长期压力监测。
2、主观评定法依赖于被测者自我感觉的描述，一般通过专业量表进行评分，这种方法简单易行，应用较为普遍；但此方法更适合用于大样本统计分析，用于长期评估单个个体的压力时，易受环境和短时记忆混淆的影响。
3、生理反应测试法通过色词、速算测试等多种测试方法评估被测者压力，根据测试分值或生理反应情况量化压力水平；这种方法受个体差异影响，而且对测试环境要求较高，长期压力监测时，每日测试内容要有适当的变化，实现成本较高。
4、生理参数测定法主要通过测量并分析被测者的生理信号(脑电，眼电，心电，PPG等)评估压力，随着传感技术的发展，上述生理信号均可由专用的生理信号采集设备采集，使用方便，且由生理信号评估压力水平更加客观，不受主管因素影响。
综上所述，在目前的压力指数评估方法中，发明人意识到生理参数测定法最适合用于长期的压力评估，但是使用该方法在采集生理信号时，需要专业的信号采集设备成本，且设备通常较昂贵，设备操作复杂，影响了该方法的广泛推广。
 技术问题
本申请的主要目的为提供一种压力评估方法、装置、计算机设备和存储介质，旨在克服目前采集生理信号以进行压力评估时，需要专业信号采集设备的缺陷。
 技术解决方案
为实现上述目的，本申请提供了一种压力评估方法，包括以下步骤：
基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
基于所述PPG信号，提取得到所述测试用户的HRV特征；
将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
本申请还提供了一种压力评估装置，包括：
采集单元，用于基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
转换单元，用于将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
提取单元，用于基于所述PPG信号，提取得到所述测试用户的HRV特征；
评估单元，用于将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
本申请还提供一种计算机设备，包括存储器和处理器，所述存储器中存储有计算机程序，所述处理器执行所述计算机程序时实现一种压力评估方法，包括以下步骤：
基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
基于所述PPG信号，提取得到所述测试用户的HRV特征；
将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
本申请还提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现一种压力评估方法，包括以下步骤：
基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
基于所述PPG信号，提取得到所述测试用户的HRV特征；
将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
 有益效果
本申请提供的压力评估方法、装置、计算机设备和存储介质，基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；基于所述PPG信号，提取得到所述测试用户的HRV特征；将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果。本申请在测试用户的压力时，只需要用到移动终端进行生理信号的采集，不需要专业的采集设备，降低采集成本。
 附图说明
图1 是本申请一实施例中压力评估方法步骤示意图；
图2是本申请一实施例中压力评估装置结构框图；
图3 为本申请一实施例的计算机设备的结构示意框图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
 本发明的最佳实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
参照图1，本申请一实施例中提供了一种压力评估方法，包括以下步骤：
步骤S1，基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
步骤S2，将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
步骤S3，基于所述PPG信号，提取得到所述测试用户的HRV特征；
步骤S4，将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
在本实施例中，上述方法应用于评估用户的当前压力，而压力的评估对用户的心理健康具有重大意义。上述方法也可应用于智慧城市的智慧医疗技术领域中，以推动智慧城市的建设。
如上述步骤S1-S2所述的，本实施例中不需要使用专业的采集设备，只需要使用移动终端（如手机、平板等）进行指尖视频的采集，进而获取到测试用户的PPG信号。具体地，打开移动终端闪光灯，并打开拍摄功能，测试用户用一根手指完全遮挡摄像头录制指尖血流视频，由于指尖血流量会随血管容积变化发生周期性的波动，血液对闪光灯光源的吸收量也随之变化，导致录制视频像素值周期性的变化。将每帧视频转化为RGBA格式，并求出R通道所有像素点求和，可得到离散的PPG信号，即每一帧视频对应PPG信号的一个采样点，如果视频帧率为30fp/s，则PPG采样率为30Hz。
如上述步骤S3所述的，压力与心脏的活动关系密切，超过了其他一些生理信号，如肌电和呼吸信号与压力之间的相关性。考虑不同个体之间的心血管系统差异，需要提取普遍适用的特征来表现心脏活动。心率变异性(HRV特征)是指逐次心跳间期的微小差异，这一指标不仅可以反应人体自主神经系统活动，还能反馈心血管系统异常，是衡量人体心脏活动的有效指标。具体地，短时HRV在压力源作用下呈下降趋势，与压力基本呈负相关关系。HRV功率谱低频功率LF，在交感神经活性降低时显著降低，而高频功率HF，随副交感神经活性升高而显著升高，因此，LF和HF可分别用于定量评估交感与副交感神经活性，二者比值LF/ HF可用于评估自主神经系统均衡性。
在本实施例中，在生理参数评估压力的方法中，与脑电、眼电、肌电等生理信号相比，HRV特征参数采集难度与成本都比较低。
如上述步骤S4所述的，上述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得，将上述测试用户的HRV特征输入至预设的压力评估模型中，得到上述测试用户的压力评估结果。
在本实施例中，无需使用专业的生理信号采集设备，只需要使用移动终端，便可以实现信号的采集，无需用户长时间佩戴传感器。在一实施例中，上述HRV特征可采用定期短时测量方式获得，综合多次短时测量的HRV特征评估压力，可以避免单次数据波动引起的随机误差，压力评估结果鲁棒性更强。
可以理解的是，在其它实施例中，也可以从ECG、BCG信号中提取HRV特征，进而评估用户压力结果。
在一实施例中，所述基于所述PPG信号，提取得到所述测试用户的HRV特征的步骤，包括：
采用样条插值或最小二乘插值法，将所述PPG信号转换为指定采样率的转换信号；
提取所述转换信号中窦性心博的RR间期；
基于所述有效的RR间期，提取得到所述测试用户的HRV特征。
在本实施例中，受移动终端的影响，其设备采样率不同，得到的PPG信号的采样率不同，因此，本方案需通过升采样或降采样方法将PPG信号的采样率转换为指定采样率（如250Hz），在升采样或降采样过程中，为保证波形形态不发生形变，通常使用样条插值或最小二乘插值方法。
由于信号可能受随机动作或硬件噪声干扰，噪声信号段RR间期无效，所以本实施例中对噪声段RR间期进行了处理(剔除或插值)，避免影响后续HRV特征提取。最后，仅有窦性心博的HRV特征有效，为避免异常心动周期影响HRV特征的计算，本实施例中对相邻差异较大的RR间期(例：差异超过20%或30%)也进行了相同的剔除或插值处理。在得到上述窦性心博的RR间期之后，便可以提取得到测试用户的HRV特征。
具体地，所述HRV特征包括HRV时域参数、HRV频域参数；
所述HRV时域参数至少包括所述窦性心博的RR间期的标准差SDNN；
所述HRV频域参数至少包括低频功率LF、高频功率HF，LF与HF的比值。其中，低频功率LF为RR间期功率谱低频能量，高频功率HF为RR间期功率谱高频能量。
在其他实施例中，HRV时域参数还包括rMSSD、PNN50，HRV频域参数还包括LF_nu(低频)、HF_nu(高频)等参数。
在一个实施例中，所述将所述测试用户的HRV特征输入至预设的压力评估模型中的步骤S4之前，包括：
步骤S10，基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果；其中，所述压力测量结果为所述样本用户基于预设压力量表测量所得；
步骤S20，基于每一次采集的生理信号，得到所述样本用户对应的HRV特征；
步骤S30，将所述样本用户每一次采集的生理信号所对应的HRV特征以及对应的压力测量结果作为一组HRV特征训练集；
步骤S40，依照采集生理信号的时间顺序，将每一组所述HRV特征训练集输入至时间序列处理的神经网络训练，得到所述压力评估模型。
在本实施例中，提出训练得到上述压力评估模型的方法。
上述预设采集周期以及预设采集时长均是根据当前的应用场景所确定，压力是表征身体状态的变量，与复杂多变的生理信号不同，人体压力状态是短时稳定的，不需要进行实时监测，所以本实施例中仅考虑为用户周期性的检测压力，监测周期长短与应用场景有关，例如：合理规划一天工作/学习安排时，需要进行短期压力测试，在一天内多次监测用户压力；长期压力测试时，仅需要1~2天监测一次；慢性压力测试时，监测周期可能更长。
采集时间间隔越长，则对采集数据质量的稳定性要求越高，因此一天测量多次时，采集时间为30~60s；一天测量一次时，采集时间1~5min；1周一次时，采集时间>10min。本实施例中，主要用于一天测量一次的应用场景中，单次数据采集时间定为2min，数据采集时，采用无需增加额外硬件成本的移动终端(可采集PPG与BCG信号)，压力量表则选择能体现用户短期压力的“PERCEIVED STRESS QUESTIONNAIRE”。
上述样本用户的生理信号采集，通常在清晨样本用户未受外界事物干扰时采集，收集样本用户每日压力初始状态。为体现样本用户日常生活中真实的压力状态，不进行任何压力诱导操作。要求样本用户每日清晨固定时间段(如：7~8am)在心情平静状态下用手机采集PPG信号，随后通过填写压力量表获得压力测量结果，即为训练数据集标签。
进而基于每一次采集的生理信号，得到所述样本用户对应的HRV特征，上述HRV特征的提取过程与上述实施例中相同，在此不再进行赘述。
将所述样本用户每一次采集的生理信号所对应的HRV特征以及对应的压力测量结果作为一组HRV特征训练集，该HRV特征训练集则用于训练上述时间序列处理的神经网络。
在本实施例中，所述时间序列处理的神经网络包括RNN、LSTM、GUR、IndRNN中的任一种。
传统BP神经网络只能单独按顺序处理输入特征，前后输入特征间是完全没有关系的，因此，BP神经网络无法体现历史HRV特征与当前压力之间的联系。循环神经网络(RNN)，可分为输入层，隐藏层，输出层三部分。
本实施例中，可采用RNN，RNN的隐藏层的值St与输出层的值Ot不仅与当前输入xt有关，还取决于上一次隐藏层的值St-1。上述压力评估模型体现出了序列数据前后之间的相互影响，更适合处理带有时序的序列。
上述RNN单元可替换为LSTM/GUR/IndRNN等多种性能更优的时间序列处理的单元。传统RNN因为在时间上参数共享，容易出现梯度消失/爆炸问题，传统RNN由于层内神经元相互联系，难以对神经元的行为进行合理的解释。LSTM/GRU可以解决传统RNN层内梯度消失/爆炸问题，而IndRNN提高了神经元的可解释性，可用于处理更长的序列信息。
最后，此模型结构不仅可用于回归计算压力指数，也可用于进行压力分级判断(例如：压力大，压力小，无压力分类等)，只需修改输出层与损失函数即可(例如：输出层激活函数设置为softmax，损失函数设置为multi class cross entropy)。
在一实施例中，所述采集样本用户的生理信号的步骤S10，包括：
基于移动终端采集所述样本用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，样本用户手指遮挡摄像头时拍摄所得；
将所述样本用户的指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到所述样本用户的PPG信号。
在本实施例中，上述采集生理信号的具体实现过程与上述实施例中一致，在此不再进行赘述。
在一实施例中，所述基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果的步骤S10之前，包括：
获取压力测试的应用场景类型，以及所述压力测试的类型；其中，所述压力测试的应用场景类型包括短期压力测试、长期压力测试以及慢性压力测试；所述压力测试的类型包括感知压力以及心理压力；
根据所述压力测试的应用场景类型与采集周期、采集时长的对应关系，获取采集所述样本用户的生理信号的所述预设采集周期以及预设采集时长；
根据所述压力测试的类型与压力量表的对应关系，获取测量所述样本用户压力测量结果的预设压力量表。
在本实施例中，上述预设采集周期以及预设采集时长均是根据当前的应用场景所确定，压力是表征身体状态的变量，与复杂多变的生理信号不同，人体压力状态是短时稳定的，不需要进行实时监测，所以本实施例中仅考虑为用户周期性的检测压力，监测周期长短与应用场景有关，例如：合理规划一天工作/学习安排时，需要进行短期压力测试，在一天内多次监测用户压力；长期压力测试时，仅需要1~2天监测一次；慢性压力测试时，监测周期可能更长。
人体承受的压力根据不同分类方法可以分为很多种，例如：感知压力以及心理压力；上述感知压力以及心理压力包括工作压力、生活压力、学习压力、训练压力、慢性压力、突发性压力、身体诱发与情绪诱发的压力等。不同的应用场合需要评估不同类型的压力，而研究压力的学者已针对不同类型压力的压力设计了对应的评估量表或测试方法。例如：针对感知压力的量表“PERCEIVED STRESS QUESTIONNAIRE”，针对心理压力的“知觉心理压力量表(CPSS)”等。
在一实施例中，所述方法还包括：
将所述预设的压力评估模型、压力评估结果、指尖视频以及HRV特征存储至区块链中。其中，区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain)，本质上是一个去中心化的数据库，是一串使用密码学方法相关联产生的数据块，每一个数据块中包含了一批次网络交易的信息，用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层。
参照图2，本申请一实施例中还提供了一种压力评估装置，包括：
采集单元10，用于基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
转换单元20，用于将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
提取单元30，用于基于所述PPG信号，提取得到所述测试用户的HRV特征；
评估单元40，用于将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
在一实施例中，所述提取单元30，具体用于：
采用样条插值或最小二乘插值法，将所述PPG信号转换为指定采样率的转换信号；
提取所述转换信号中窦性心博的RR间期；
基于所述有效的RR间期，提取得到所述测试用户的HRV特征。
在一实施例中，所述HRV特征包括HRV时域参数、HRV频域参数；
所述HRV时域参数至少包括所述窦性心博的RR间期的标准差SDNN；
所述HRV频域参数至少包括低频功率LF、高频功率HF，LF与HF的比值。
在一实施例中，上述压力评估装置，还包括：
样本采集单元，用于基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果；其中，所述压力测量结果为所述样本用户基于预设压力量表测量所得；
样本特征提取单元，用于基于每一次采集的生理信号，得到所述样本用户对应的HRV特征，
训练集构建单元，用于将所述样本用户每一次采集的生理信号所对应的HRV特征以及对应的压力测量结果作为一组HRV特征训练集；
训练单元，用于依照采集生理信号的时间顺序，将每一组所述HRV特征训练集输入至时间序列处理的神经网络训练，得到所述压力评估模型。
在一实施例中，所述样本采集单元采集样本用户的生理信号，包括：
基于移动终端采集所述样本用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，样本用户手指遮挡摄像头时拍摄所得；
将所述样本用户的指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到所述样本用户的PPG信号。
在一实施例中，上述压力评估装置，还包括：
类型获取单元，用于获取压力测试的应用场景类型，以及所述压力测试的类型；其中，所述压力测试的应用场景类型包括短期压力测试、长期压力测试以及慢性压力测试；所述压力测试的类型包括感知压力以及心理压力；
第一获取单元，用于根据所述压力测试的应用场景类型与采集周期、采集时长的对应关系，获取采集所述样本用户的生理信号的所述预设采集周期以及预设采集时长；
第二获取单元，用于根据所述压力测试的类型与压力量表的对应关系，获取测量所述样本用户压力测量结果的预设压力量表。
在一实施例中，所述时间序列处理的神经网络包括RNN、LSTM、GUR、IndRNN中的任一种。
在本实施例中，上述装置实施例中各个单元的具体实现，请参照上述方法实施例中所述，在此不再进行赘述。
参照图3，本申请实施例中还提供一种计算机设备，该计算机设备可以是服务器，其内部结构可以如图3所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中，该计算机设计的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储压力评估结果等。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种压力评估方法，包括以下步骤：
基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
基于所述PPG信号，提取得到所述测试用户的HRV特征；
将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
本领域技术人员可以理解，图3中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定。
本申请一实施例还提供一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现一种压力评估方法，包括以下步骤：
基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
基于所述PPG信号，提取得到所述测试用户的HRV特征；
将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
可以理解的是，本实施例中的计算机可读存储介质可以是易失性可读存储介质，也可以为非易失性可读存储介质。
综上所述，为本申请实施例中提供的压力评估方法、装置、计算机设备和存储介质，基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；基于所述PPG信号，提取得到所述测试用户的HRV特征；将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果。本申请在测试用户的压力时，只需要用到移动终端进行生理信号的采集，不需要专业的采集设备，降低采集成本。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其它变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、装置、物品或者方法不仅包括那些要素，而且还包括没有明确列出的其它要素，或者是还包括为这种过程、装置、物品或者方法所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、装置、物品或者方法中还存在另外的相同要素。
以上所述仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其它相关的技术领域，均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种压力评估方法，其中，包括以下步骤：
    基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
    将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
    基于所述PPG信号，提取得到所述测试用户的HRV特征；
    将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
  2. 根据权利要求1所述的压力评估方法，其中，所述基于所述PPG信号，提取得到所述测试用户的HRV特征的步骤，包括：
    采用样条插值或最小二乘插值法，将所述PPG信号转换为指定采样率的转换信号；
    提取所述转换信号中窦性心博的RR间期；
    基于所述有效的RR间期，提取得到所述测试用户的HRV特征。
  3. 根据权利要求2所述的压力评估方法，其中，所述HRV特征包括HRV时域参数、HRV频域参数；
    所述HRV时域参数至少包括所述窦性心博的RR间期的标准差SDNN；
    所述HRV频域参数至少包括低频功率LF、高频功率HF，LF与HF的比值。
  4. 根据权利要求1所述的压力评估方法，其中，所述将所述测试用户的HRV特征输入至预设的压力评估模型中的步骤之前，包括：
    基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果；其中，所述压力测量结果为所述样本用户基于预设压力量表测量所得；
    基于每一次采集的生理信号，得到所述样本用户对应的HRV特征，
    将所述样本用户每一次采集的生理信号所对应的HRV特征以及对应的压力测量结果作为一组HRV特征训练集；
    依照采集生理信号的时间顺序，将每一组所述HRV特征训练集输入至时间序列处理的神经网络训练，得到所述压力评估模型。
  5. 根据权利要求4所述的压力评估方法，其中，所述采集样本用户的生理信号的步骤，包括：
    基于移动终端采集所述样本用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，样本用户手指遮挡摄像头时拍摄所得；
    将所述样本用户的指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到所述样本用户的PPG信号。
  6. 根据权利要求4所述的压力评估方法，其中，所述基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果的步骤之前，包括：
    获取压力测试的应用场景类型，以及所述压力测试的类型；其中，所述压力测试的应用场景类型包括短期压力测试、长期压力测试以及慢性压力测试；所述压力测试的类型包括感知压力以及心理压力；
    根据所述压力测试的应用场景类型与采集周期、采集时长的对应关系，获取采集所述样本用户的生理信号的所述预设采集周期以及预设采集时长；
    根据所述压力测试的类型与压力量表的对应关系，获取测量所述样本用户压力测量结果的预设压力量表。
  7. 根据权利要求1所述的压力评估方法，其中，所述时间序列处理的神经网络包括RNN、LSTM、GUR、IndRNN中的任一种。
  8. 一种压力评估装置，其中，包括：
    采集单元，用于基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
    转换单元，用于将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
    提取单元，用于基于所述PPG信号，提取得到所述测试用户的HRV特征；
    评估单元，用于将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
  9. 一种计算机设备，包括存储器和处理器，所述存储器中存储有计算机程序，其中，所述处理器执行所述计算机程序时实现一种压力评估方法的步骤：包括：
    基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
    将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
    基于所述PPG信号，提取得到所述测试用户的HRV特征；
    将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
  10. 根据权利要求9所述的计算机设备，其中，所述基于所述PPG信号，提取得到所述测试用户的HRV特征的步骤，包括：
    采用样条插值或最小二乘插值法，将所述PPG信号转换为指定采样率的转换信号；
    提取所述转换信号中窦性心博的RR间期；
    基于所述有效的RR间期，提取得到所述测试用户的HRV特征。
  11. 根据权利要求10所述的计算机设备，其中，所述HRV特征包括HRV时域参数、HRV频域参数；
    所述HRV时域参数至少包括所述窦性心博的RR间期的标准差SDNN；
    所述HRV频域参数至少包括低频功率LF、高频功率HF，LF与HF的比值。
  12. 根据权利要求9所述的计算机设备，其中，所述将所述测试用户的HRV特征输入至预设的压力评估模型中的步骤之前，包括：
    基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果；其中，所述压力测量结果为所述样本用户基于预设压力量表测量所得；
    基于每一次采集的生理信号，得到所述样本用户对应的HRV特征，
    将所述样本用户每一次采集的生理信号所对应的HRV特征以及对应的压力测量结果作为一组HRV特征训练集；
    依照采集生理信号的时间顺序，将每一组所述HRV特征训练集输入至时间序列处理的神经网络训练，得到所述压力评估模型。
  13. 根据权利要求12所述的计算机设备，其中，所述采集样本用户的生理信号的步骤，包括：
    基于移动终端采集所述样本用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，样本用户手指遮挡摄像头时拍摄所得；
    将所述样本用户的指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到所述样本用户的PPG信号。
  14. 根据权利要求12所述的计算机设备，其中，所述基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果的步骤之前，包括：
    获取压力测试的应用场景类型，以及所述压力测试的类型；其中，所述压力测试的应用场景类型包括短期压力测试、长期压力测试以及慢性压力测试；所述压力测试的类型包括感知压力以及心理压力；
    根据所述压力测试的应用场景类型与采集周期、采集时长的对应关系，获取采集所述样本用户的生理信号的所述预设采集周期以及预设采集时长；
    根据所述压力测试的类型与压力量表的对应关系，获取测量所述样本用户压力测量结果的预设压力量表。
  15. 根据权利要求9所述的计算机设备，其中，所述时间序列处理的神经网络包括RNN、LSTM、GUR、IndRNN中的任一种。
  16. 一种计算机可读存储介质，其上存储有计算机程序，其中，所述计算机程序被处理器执行时实现一种压力评估方法的步骤，包括：
    基于移动终端采集测试用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，测试用户手指遮挡摄像头时拍摄所得；
    将所述指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到PPG信号；
    基于所述PPG信号，提取得到所述测试用户的HRV特征；
    将所述测试用户的HRV特征输入至预设的压力评估模型中，得到所述测试用户的压力评估结果；其中，所述压力评估模型基于HRV特征训练集训练时间序列处理的神经网络所得。
  17. 根据权利要求16所述的计算机可读存储介质，其中，所述基于所述PPG信号，提取得到所述测试用户的HRV特征的步骤，包括：
    采用样条插值或最小二乘插值法，将所述PPG信号转换为指定采样率的转换信号；
    提取所述转换信号中窦性心博的RR间期；
    基于所述有效的RR间期，提取得到所述测试用户的HRV特征。
  18. 根据权利要求17所述的计算机可读存储介质，其中，所述HRV特征包括HRV时域参数、HRV频域参数；
    所述HRV时域参数至少包括所述窦性心博的RR间期的标准差SDNN；
    所述HRV频域参数至少包括低频功率LF、高频功率HF，LF与HF的比值。
  19. 根据权利要求16所述的计算机可读存储介质，其中，所述将所述测试用户的HRV特征输入至预设的压力评估模型中的步骤之前，包括：
    基于预设采集周期以及预设采集时长采集样本用户的生理信号，并获取所述样本用户每一次采集生理信号时的压力测量结果；其中，所述压力测量结果为所述样本用户基于预设压力量表测量所得；
    基于每一次采集的生理信号，得到所述样本用户对应的HRV特征，
    将所述样本用户每一次采集的生理信号所对应的HRV特征以及对应的压力测量结果作为一组HRV特征训练集；
    依照采集生理信号的时间顺序，将每一组所述HRV特征训练集输入至时间序列处理的神经网络训练，得到所述压力评估模型。
  20. 根据权利要求19所述的计算机可读存储介质，其中，所述采集样本用户的生理信号的步骤，包括：
    基于移动终端采集所述样本用户的指尖视频；其中，所述指尖视频为所述移动终端闪光灯开启状态下，样本用户手指遮挡摄像头时拍摄所得；
    将所述样本用户的指尖视频转换为RGBA视频格式，并提取出R通道中的所有像素点的求和，得到所述样本用户的PPG信号。
PCT/CN2021/084544 2020-11-09 2021-03-31 压力评估方法、装置、计算机设备和存储介质 WO2022095331A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011240444.4A CN112370057A (zh) 2020-11-09 2020-11-09 压力评估方法、装置、计算机设备和存储介质
CN202011240444.4 2020-11-09

Publications (1)

Publication Number Publication Date
WO2022095331A1 true WO2022095331A1 (zh) 2022-05-12

Family

ID=74578427

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/084544 WO2022095331A1 (zh) 2020-11-09 2021-03-31 压力评估方法、装置、计算机设备和存储介质

Country Status (2)

Country Link
CN (1) CN112370057A (zh)
WO (1) WO2022095331A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115281676A (zh) * 2022-10-08 2022-11-04 齐鲁工业大学 基于gru神经网络和ecg信号的疲劳检测方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112370057A (zh) * 2020-11-09 2021-02-19 平安科技（深圳）有限公司 压力评估方法、装置、计算机设备和存储介质
CN113257418A (zh) * 2021-03-29 2021-08-13 广州科克里特生命科技有限公司 一种用于腰背痛的风险检测系统与方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017016086A1 (zh) * 2015-07-30 2017-02-02 华南理工大学 基于生理信息的抑郁症评估系统及其评估方法
CN109106350A (zh) * 2018-09-20 2019-01-01 中山大学 低生理负荷的睡眠呼吸暂停低通气综合征初步筛查方法
CN110477895A (zh) * 2019-07-24 2019-11-22 苏州国科医疗科技发展有限公司 基于血液容积波的多光源探测器连续心率测量方法
CN111797817A (zh) * 2020-07-30 2020-10-20 平安科技（深圳）有限公司 情绪识别方法、装置、计算机设备及计算机可读存储介质
CN112370057A (zh) * 2020-11-09 2021-02-19 平安科技（深圳）有限公司 压力评估方法、装置、计算机设备和存储介质

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR3079127B1 (fr) * 2018-03-21 2021-07-30 Magid Hallab Systeme pour le suivi du stress chez un individu
CN110477932B (zh) * 2019-07-31 2022-04-15 商丘师范学院 一种基于互联网与云计算的学生心理压力评估方法及系统
CN111513730B (zh) * 2020-03-20 2023-06-09 合肥工业大学 基于多通道生理数据的心理压力预测方法和系统
CN111714144A (zh) * 2020-07-24 2020-09-29 长春理工大学 基于视频非接触测量的精神压力分析方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017016086A1 (zh) * 2015-07-30 2017-02-02 华南理工大学 基于生理信息的抑郁症评估系统及其评估方法
CN109106350A (zh) * 2018-09-20 2019-01-01 中山大学 低生理负荷的睡眠呼吸暂停低通气综合征初步筛查方法
CN110477895A (zh) * 2019-07-24 2019-11-22 苏州国科医疗科技发展有限公司 基于血液容积波的多光源探测器连续心率测量方法
CN111797817A (zh) * 2020-07-30 2020-10-20 平安科技（深圳）有限公司 情绪识别方法、装置、计算机设备及计算机可读存储介质
CN112370057A (zh) * 2020-11-09 2021-02-19 平安科技（深圳）有限公司 压力评估方法、装置、计算机设备和存储介质

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115281676A (zh) * 2022-10-08 2022-11-04 齐鲁工业大学 基于gru神经网络和ecg信号的疲劳检测方法

Also Published As

Publication number Publication date
CN112370057A (zh) 2021-02-19

Similar Documents

Publication Publication Date Title
JP7367099B2 (ja) せん妄患者の脳症の存在をスクリーニングするためのシステム
EP3698707B1 (en) Electrocardiogram information dynamic monitoring system, computer program and computer readable storage medium
WO2022095331A1 (zh) 压力评估方法、装置、计算机设备和存储介质
JP6160029B2 (ja) ストレスモニタリングシステム
JP6044111B2 (ja) 人のストレスモデルを生成するシステム
US20160051203A1 (en) Method and system for pulse diagnosis
WO2020006845A1 (zh) 一种健康储备评估方法、装置及其应用
JP2012239888A (ja) 腎臓のドップラ超音波法により作成されたストレスプロファイルを使用したストレスの連続したモニタリング
CN113133752B (zh) 基于心率变异分析的心理评估方法、系统、设备和介质
CN114246589B (zh) 记忆认知能力测评方法及系统
CN104367306A (zh) 生理心理职业测评系统及实现方法
Xu et al. Automatic detection of artifacts in heart period data
Tiwari et al. Stress and anxiety measurement" in-the-wild" using quality-aware multi-scale hrv features
CN117158938A (zh) 一种应用于智能手表的健康监测方法、装置及电子设备
TWI661384B (zh) 生理監控回饋系統及其運作方法
TW201247169A (en) Heart rate variability device and cloud health management system
TWI516247B (zh) 居家式憂鬱傾向生理訊號之情緒分析方法
Luo et al. Quantitative analysis of heart rate variability parameter and mental stress index
Koryahin et al. Integration of Information Systems in the Control of Heart Rate in the Process of Physical Education
Ide et al. Workplace stress estimation method based on multivariate analysis of physiological indices
Farhan et al. Impact of Heart Rate Variability on Physiological Stress: Systematic Review
WO2019058293A1 (en) EQUIPMENT FOR DETERMINING A HUMAN HEALTH STATUS USING SYNCHRONIZED DATA FROM MULTIPLE SENSORS AND METHOD OF OPERATING THE SAME
CN114334090B (zh) 一种数据分析方法、装置及电子设备
RU129680U1 (ru) Система определения функционального состояния группы людей
CN114530250B (zh) 基于数据增强的可穿戴血糖检测方法、系统及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21888035

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 09.08.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21888035

Country of ref document: EP

Kind code of ref document: A1