CN116130086A - Motion sickness prediction system - Google Patents
Motion sickness prediction system Download PDFInfo
- Publication number
- CN116130086A CN116130086A CN202211734857.7A CN202211734857A CN116130086A CN 116130086 A CN116130086 A CN 116130086A CN 202211734857 A CN202211734857 A CN 202211734857A CN 116130086 A CN116130086 A CN 116130086A
- Authority
- CN
- China
- Prior art keywords
- data
- vector
- motion sickness
- prediction
- peak
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
- 201000003152 motion sickness Diseases 0.000 title claims abstract description 152
- 230000008447 perception Effects 0.000 claims abstract description 14
- 239000013598 vector Substances 0.000 claims description 170
- 238000000034 method Methods 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 17
- 230000001133 acceleration Effects 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 15
- 238000009423 ventilation Methods 0.000 claims description 14
- 230000002708 enhancing effect Effects 0.000 claims description 13
- 238000013528 artificial neural network Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 10
- 238000004140 cleaning Methods 0.000 claims description 7
- 238000012217 deletion Methods 0.000 claims description 6
- 230000037430 deletion Effects 0.000 claims description 6
- 230000036760 body temperature Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 208000024891 symptom Diseases 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000007493 shaping process Methods 0.000 claims description 2
- 238000012549 training Methods 0.000 claims description 2
- 230000007787 long-term memory Effects 0.000 claims 1
- 238000001514 detection method Methods 0.000 abstract 1
- 238000009434 installation Methods 0.000 abstract 1
- 238000013178 mathematical model Methods 0.000 abstract 1
- 238000012544 monitoring process Methods 0.000 abstract 1
- 230000006872 improvement Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 206010028813 Nausea Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008693 nausea Effects 0.000 description 1
- 230000036544 posture Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003238 somatosensory effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H80/00—ICT specially adapted for facilitating communication between medical practitioners or patients, e.g. for collaborative diagnosis, therapy or health monitoring
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a prediction system for motion sickness, which is characterized in that a set of flexible and effective detection and prediction models for motion sickness of buses are established, individual qualitative and quantitative data of users are collected, specific motion sickness sensing experiences of the users taking buses are acquired more accurately, a mathematical model is established, the current motion sickness sensing experiences of the buses can be evaluated and predicted in real time, and the users are guided to relieve motion sickness sensing through output dynamic strategies. The motion sickness prediction system provided by the invention has the advantages of low cost, more accuracy, easiness in deployment, easiness in use, convenience in installation, simplicity in operation and simplicity in flow, is more convenient for monitoring the operation condition of an own vehicle, adjusting the motion dynamics of the vehicle, improving the riding comfort and reducing the motion sickness perception for a public transport operation system.
Description
Technical Field
The invention belongs to the technical field of data acquisition and processing, and particularly relates to a motion sickness prediction system.
Background
According to researches, the motion sickness of a user in the process of taking a bus is accumulated, the feeling intensity of the motion sickness physiological symptoms can be enhanced by accumulated composite superposition according to various external elements experienced by passengers in the process of taking the bus, the elements can be divided into external elements and internal elements, the external elements are mainly cabin environment and vehicle dynamic elements of the bus, the internal elements are psychological elements of the user, and the user perceives the external environment and evaluates the physical condition of the user. The external elements may be obtained by sensors, while the internal elements are obtained by qualitative means.
At present, the main carsickness research is directed at the driver of a car, for example, patent document CN108639061A discloses a method for actively preventing and assisting driving of a car, which mainly aims at the problem that the car sickness of passengers is easy to occur in the car, and establishes an optimal control problem with the carsickness prevention as an evaluation index by improving the riding comfort of the car so as to prevent the passengers from nausea and carsickness, dynamically calculates an acceleration command meeting the active carsickness prevention, and realizes the active and anti-carsickness assisting driving control of the car. Patent document CN113795416a discloses a method for predicting and alleviating disturbance caused by motion sickness, mainly for an occupant, a feature number that describes the occurrence probability of disturbance caused by motion sickness is determined by detecting the stimulus of the occupant, the personal susceptibility of the occupant to motion sickness, the type of activity done during driving operation, and at least one individual measure in a measure list for preventing disturbance caused by motion sickness is suggested or automatically started to the occupant depending on the determined feature number.
However, according to the green travel advocated by the country, a quite large group selects a public transportation mode for traveling, in view of the complex condition of public transportation, accurate and comprehensive external environment information of a user is difficult to obtain accurately by a quantitative method, and meanwhile, according to research results, acceleration, crowding degree and temperature sensed by different positions and sitting postures of the user on a bus are different; in addition, in view of actual domestic traffic conditions, the public transportation means, especially passengers on buses, can be subjected to motion sickness prediction in a very obvious trip peak period, and the problem to be solved in the current social environment is urgent.
Disclosure of Invention
In order to solve the technical problems, the invention discloses a motion sickness prediction system, which aims at the specific environment of a bus, fuses peak period data of traffic, collects qualitative data and quantitative data of a user, performs pretreatment, processes the pretreated input data by using a motion sickness prediction model to obtain a prediction result, and improves the accuracy of motion sickness prediction.
A motion sickness prediction system comprises a mobile terminal, a data preprocessing module, a motion sickness prediction module and a terminal result display module;
the mobile terminal is internally provided with a front-end data acquisition module which is used for acquiring quantitative data and qualitative data of a user;
the data preprocessing module is used for preprocessing the quantitative data and the qualitative data of the user of the acquisition tool to obtain preprocessed data;
the motion sickness prediction module is used for processing the preprocessed data to obtain motion sickness prediction results; the motion sickness prediction module comprises a peak period judgment and data enhancement branch and motion sickness perception score prediction module; the peak period judging and data enhancing branch is used for judging the peak period and enhancing the corresponding data according to the judging result; the motion sickness awareness score prediction module is used for outputting motion sickness prediction data of a user;
the terminal display module is used for receiving the prediction data obtained by the data processing module, visualizing the prediction data, and outputting a dynamic strategy through the graphical human-computer interface to guide a user to relieve motion sickness perception.
Further improvements include local time, current location information, X-axis acceleration magnitude at 30HZ frequency, Y-axis acceleration magnitude at 30HZ frequency, Z-axis acceleration magnitude at 30HZ frequency, X, Y, Z axis angular velocity magnitude, noise value, vibration frequency, ride location data, including somatosensory temperature, perceived motion sickness score, perceived in-vehicle congestion level, ventilation level, and peak traffic level; the user performs an input of qualitative data on the mobile terminal.
Further improved, the data preprocessing module preprocesses quantitative data and qualitative data as follows:
s2-1: the quantitative and qualitative data ranges of the user are adjusted to [0,1], and the normalization formula is as follows:
wherein x= [ x ] 1 ,x 2 ,…,x i …,x n ]N is a natural number, x, which is an n-dimensional column vector of qualitative data i Is the ith shaping data;
for qualitative data, the mobile terminal converts the qualitative data into One-Hot code by using 1 minute as a frequency band, the One-Hot code is carried out by adopting a mode of filling in a subjective evaluation scale for qualitative data input, the motion is carried out by adopting a motion 10-point scale, the body temperature is detected by adopting a body temperature 3-point scale, the crowding degree is detected by adopting a crowding degree 3-point scale, the ventilation degree is detected by adopting a ventilation degree 3-point scale, and the driving peak degree is detected by adopting a driving peak 3-point scale.
S2-2: data cleaning;
if the finished quantitative data have missing values, the values of the two time slice data before and after the missing values are selected by the missing position data values, and the values are averaged and filled; if the encoded qualitative data has a missing value, the most value of all the qualitative data is used for complementing the missing; and obtaining the qualitative data and the quantitative data after cleaning.
S2-3: data arrangement
The cleaned qualitative data and quantitative data are arranged and tidied into an input vector X which can be input by a model, and a data tag Y vector:
the input vector X comprises: x, Y, Z axis acceleration and angular velocity 6-dimensional degree vector, local time vector 1-dimensional degree vector, noise value vector 1-dimensional degree vector, perceived ventilation degree 3-dimensional degree vector, sitting position 5-dimensional degree vector, perceived crowding degree 3-dimensional degree vector, ventilation degree 3-dimensional degree vector and current time (T time) motion perception scoring 10-dimensional degree vector;
the Y vector of the data tag contains: vector y of running peak 2 point scale peek Vector y of T+1 motion 10 point scale at next time T+1 The method comprises the steps of carrying out a first treatment on the surface of the And obtaining the preprocessed data.
The motion sickness prediction module is a long-short-term memory LSTM network, and the long-short-term memory LSTM network comprises a peak period judgment and data enhancement branch and a perceived motion sickness score prediction module; the peak period judging and data enhancing branch comprises a peak period judging module and a data enhancing module; the perceived motion sickness score prediction module comprises a perceived motion sickness score prediction LSTM layer, a self-attention layer and a motion sickness prediction layer;
training the preprocessed data as input of the motion sickness prediction module, so as to obtain a trained motion sickness prediction module;
the trained motion sickness prediction module comprises the following steps:
taking the collected input vector X as the input of a peak period judging module, and outputting a peak type by the peak period judging module, wherein the peak type is a peak period or a non-peak period; when the peak type is the peak period, the peak judgment data is input into the data enhancement module to be used as the input of the perceived motion sickness score prediction LSTM layer after the data enhancement and the peak judgment data, otherwise, the peak judgment data are directly input into the perceived motion sickness score prediction LSTM layer, the output of the perceived motion sickness score prediction LSTM layer is used as the input of the self-attention layer, the position of the self-attention layer is used as the input of the motion sickness prediction layer, and the motion sickness prediction layer outputs the motion sickness prediction result.
Further improvement, the data processing steps of the rush hour judging module are as follows:
the input vector X is taken as the 29-dimensional input vector X of the LSTM layer in Will be 29 dimensionsInput vector x in The size of the sequence is segmented into seq_len_batch_size by adopting a sliding window according to time sequence, and then the seq_batch_size is input into an LSTM network; wherein batch length batch_size=20, feature dimension size=29, and sequence length seq_len is time sequence length L of all data data The value after subtracting the batch_size, seq_len, is calculated as follows:
seq len ＝L data -batch_size
29-dimensional input vector x in Obtaining intermediate level features x after entering LSTM layer mid Intermediate level feature x mid The vector size is seq_len_batch_size 16: LSTM () represents LSTM layer processing;
x mid ＝LSTM(x in )
the next full link layer contains a full link convolution Linear () with output size set to 2 and a Softmax () active operation, outputting the LSTM layer result x mid Sending the motion vector into a full connection layer to obtain an output vector y for predicting the peak degree of the traveling crane pred ；
y pred ＝Softmax(Linear(x mid ))
The Softmax () function is as follows:
wherein x is i The value of the ith class vector, C is the number of classified classes, 3, x c A predictive weight vector value representing class c, e being a natural logarithm; the output result of the full connection layer is peak period and off-peak period.
Further improvement, the data enhancement module comprises a parallel random exchange unit, a random deletion unit and a Seq2Seq unit;
the random exchange unit randomly selects two values from the column vector, exchanges the positions of the two values, and repeats the whole vector n times to obtain an output vector after random exchange enhancement, which is marked as x aug1 The method comprises the steps of carrying out a first treatment on the surface of the The random deleting unit randomly deletes the numbers in the vector with a preset fixed probabilityThe value, the output vector after random deletion enhancement is obtained and is marked as x aug2 The method comprises the steps of carrying out a first treatment on the surface of the The Seq2Seq unit intercepts the 3-dimensional vector y for predicting the peak degree of the driving in prediction in the prediction stage pred Input data x as time series of peak periods peek Take out x peek Intermediate level feature x in a network after entering LSTM layer mid ，x mid Namely the enhancement feature of the Seq2 Seq;
the output result of the peak period judgment and data enhancement branch is x out When the peak period is determined, the peak period is determined and the output result x of the data enhancement branch is obtained out To enhance vector features [ x ] aug1 ,x aug2 ,x mid ]Plus the LSTM layer 29-dimensional input vector x in The method comprises the steps of carrying out a first treatment on the surface of the When the judgment is off-peak, the judgment is carried out in peak and the output result x of the data enhancement branch is obtained out Input vector x for the 29-dimensional dimension of the LSTM layer in 。
Further improved, the processing method of the perceived motion sickness score prediction module is as follows: inputting the input x of the perceived motion sickness score prediction module into the perceived motion sickness score prediction LSTM layer to obtain a 10-dimensional output vector y of the perceived motion sickness score prediction LSTM layer out1 Perceived motion sickness score predicts 10-dimensional output vector y of LSTM layer out1 Is fed as input into the self-attention layer;
the self-attention layer comprises the following matrix operations:
the self-attention layer first predicts the perceived motion sickness score to the 10-dimensional output vector y of the LSTM layer out1 Divided into 10 dimension vectors a by dimension 1 ,a 2 ,…,a 10 Each dimension vector is multiplied by different neural network matrixes to obtain 3 different 10-dimension vectors q= (q) 1 ,q 2 ,…,q 10 ) They are calculated in each dimension i as follows:
q i ＝W q *a i
k i ＝W k *a i
v i ＝W v *a i
wherein W is k And W is v Is a learnable parameter in a neural network;
then take the vector q of the first dimension of the q vector 1 K for each dimension of the k vector i Doing an attention multiplication to get q 1 And each k i Is a degree of proximity alpha of (a) 1,i ：
Wherein d is q 1 And k i Because of q 1 *k i The value of (2) increases with the dimension, so it is divided by
Corresponding to the normalized effect;
and then to the obtained alpha 1,i Vector softmax () operation results
Then the same calculation is carried out to obtain the complete product
Vector: />
Wherein i=1, 2 …,10;
the self-attention layer will finally
V corresponding to the dimension of v vector i Multiplication of values, in particular, +.>
Multiplying by v 1 、/>
Multiplying by
Multiplying by v 10 . Then the product results are added up to obtain b 1 At this time b 1 With q 1 And k is equal to i And v i All associated information:
similarly, use and obtain b 1 The same operation vector k 2 To k 10 Attention is paid to get b 2 To b 10 A 10-dimensional vector y output from the attention layer is obtained out2 ＝(b 1 ,b 2 ,…,b 10 )；
Finally, outputting the 10-dimensional vector y from the attention layer out2 Sending the motion sickness prediction layer;
the motion sickness prediction layer comprises a fully connected layer of a neural network, a Softmax () function and a max () function; 10-dimensional vector y output from the attention layer out2 A mapping operation is carried out through the full connection layer Linear () to output a 2-dimensional predictive vector y out2 ' 2-point fixed range degree probability representing motion symptom, 2-dimensional predictive vector y out2 ' finally, class probability is obtained through a Softmax () function, and class probability is processed through a max () function to obtain a level predictive value y of the motion sickness degree of the next minute out3 ：
y out3 ＝max(Softmax(Linear(y out2 )))
Level prediction value y of motion sickness degree of next minute out3 If 0 is not carsickness, y out3 A value of 1 indicates motion sickness 1.
Further improvements, the prediction system directs a user predicted to be motion sickness to a location predicted to be motion sickness based on the collected quantitative and qualitative data at a public transportation means.
The invention has the beneficial effects that:
1. the data input of qualitative data and quantitative data serving as the motion sickness sensing prediction model is combined, so that the model can learn subjective sensing experiences of more passengers, better learn mapping relations before individual difference characteristics of users, and improve motion sickness experience prediction generalization capability of the model for different user differences.
2. And a time sequence model is used for better adapting to the physiological characteristics of dynamic changes of motion sickness. Research shows that the triggering intensity of the motion sickness is cumulative, the motion sickness sensing experience of a user is dynamic along with the increase of the riding time and the change of the driving conditions such as temperature, acceleration frequency, intensity and the like, the change of the state in the bus riding can be captured based on a time sequence deep learning model, the future state can be deduced and predicted according to the state input in real time and the past state memory capacity (lstm), and the motion sickness sensing method has good performance in application.
3. Compared with the passenger car, the bus has difference in riding experience, and the method collects corresponding data influencing the motion sickness perception of the user based on the riding characteristics of the bus, so that the motion sickness perception of the bus can be predicted better.
Drawings
Fig. 1 is a block diagram of an motion sickness prediction system according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a peak determination and data enhancement branch in accordance with an embodiment of the present invention.
FIG. 3 is a schematic diagram of slide sampling according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention are further described below with reference to the accompanying drawings and examples. It should be noted that the examples do not limit the scope of the invention as claimed.
Example 1
As shown in fig. 1 to 2, a method for using a motion sickness prediction system includes the following steps:
s1: receiving qualitative data and quantitative data of a user acquired by a mobile terminal;
s2: preprocessing the collected qualitative data and quantitative data to obtain preprocessed input data;
s3: processing the preprocessed input data by using an motion sickness prediction model to obtain a prediction result; the motion sickness prediction model comprises a peak period judgment and data enhancement branch and a motion sickness sensing score prediction module.
Acquiring quantitative data of an environment where a user is located by using a mobile terminal, wherein the quantitative data comprises local time, current position information, X-axis acceleration of 30HZ frequency, Y-axis acceleration of 30HZ frequency, Z-axis acceleration of 30HZ frequency, angular velocity of each axis, noise value, vibration frequency and riding position data; and the user inputs qualitative data on the mobile terminal, wherein the qualitative data comprises body temperature sensing, motion sickness sensing scoring, in-vehicle congestion sensing degree, ventilation degree, driving peak degree and the like.
Preprocessing the collected qualitative data and quantitative data comprises the following steps:
s2-1: data arrangement and coding;
for quantitative data, the data are segmented at a period of 1 minute (30 Hz per second) and are arranged in a time sequence, and then in order to avoid the problem that the model cannot be converged due to overlarge difference of the numerical ranges of the vector data, normalization processing is carried out on the quantitative data, and the data range is adjusted to be between [0,1] so as to facilitate model convergence. The normalization formula is as follows:
wherein x= [ x ] 1 ,x 2 ,…,x i …,x n ]N is a natural number, which is an n-dimensional column vector of qualitative data.
For qualitative data, the segmentation was performed in 1 minute (60 seconds) frequency band. The mobile terminal converts qualitative data into One-Hot coding, the qualitative data is input by filling an FMS evaluation scale to carry out One-Hot coding, motion adopts a motion 20-point scale, motion temperature adopts a motion temperature 3-point scale, congestion degree adopts a congestion degree 3-point scale, ventilation degree 3-point scale, and driving peak degree adopts a driving peak 3-point scale.
S2-2: data cleaning;
the data cleaning is mainly the missing value processing, namely if the missing value exists in the sorted quantitative data, the missing position data value selects the value of each two time slice data before and after the missing position data value to average for filling; if the encoded qualitative data has missing values, the most value of all the qualitative data is used for complementing the missing. Thus, qualitative and quantitative data after cleaning are obtained.
S2-3: data arrangement
Due to the requirement of subsequent model input, the cleaned qualitative data and quantitative data need to be arranged and tidied into an input vector X which can be input by the model and a data label Y vector.
The input vector X comprises: the XYZ axis acceleration and angular velocity 6-dimensional degree vector, the local time vector 1-dimensional degree vector, the noise value vector 1-dimensional degree vector, the perceived ventilation degree 3-dimensional degree vector, the riding position 5-dimensional degree vector, the perceived crowding degree 3-dimensional degree vector and the current time (T time) motion perception scoring 10-dimensional degree vector;
the Y vector of the data tag contains: vector y of running peak 2 point scale peek Vector y of T+1 motion 10 point scale at next time T+1 The method comprises the steps of carrying out a first treatment on the surface of the And obtaining the preprocessed data.
Because the acquired data input and output have time continuity, the data correlation between the result at the next moment and the data at the previous moment is very high, and multiple motion data are used for multi-step time sequence prediction, the modeling is performed by using a long-short-period memory LSTM network with the dominant processing time sequence data, a attention mechanism capable of effectively extracting the connection between multiple types of data and a neural network. The motion sickness prediction model comprises a peak period judgment and data enhancement branch and a motion sickness perception score prediction module.
The peak period judging and data enhancing branch comprises a peak period judging module and a data enhancing module, wherein the peak period judging module comprises an LSTM layer and a full connection layer.
The local time vector, the noise value vector, the vibration frequency vector, the 3 degree vector of the perceived crowding degree, the 3 degree vector of the ventilation degree and the 10 perceived motion sickness score vectors after pretreatment are taken together as the 29-dimensional input vector x of the LSTM layer in Input vector x of 29 dimensions in Is segmented into seq_len_size_size in time series and then is input into an LSTM network. Wherein, batch_size=20, and seq_len is the time sequence length L of all data data The value after subtracting the batch_size, seq_len, is calculated as follows:
seq len ＝L data -batch_size
(As shown in FIG. 3, which samples in a sliding window according to the batch length batch_size, there is overlap in the samples due to the sliding window
29-dimensional input vector x in Obtaining intermediate level features x after entering LSTM layer mid Intermediate level feature x mid The vector size is seq_len_batch_size 16: LSTM () represents LSTM layer processing;
x mid ＝LSTM(x in )
the next full link layer contains a full link convolution Linear () with output size set to 2 and a Softmax () active operation, outputting the LSTM layer result x mid Sending the motion vector into a full connection layer to obtain an output vector y for predicting the peak degree of the traveling crane pred ；
y pred ＝Softmax(Linear(x mid ))
The Softmax () function is as follows:
wherein x is i The value of the ith class vector is C, the number of classified classes is 2, and e is natural logarithm; the output result of the full connection layer is that the business trip is performedRush hour and off-peak rush hour.
The input data output by the full connection layer and output as dense peak period and general peak period is then intercepted and is characterized by x ori Features x of input data for data-intensive peak periods and for peak periods in general ori And inputting the data enhancement module to enhance the data in the peak period. The data enhancement module comprises a parallel random exchange unit, a random deleting unit and a Seq2Seq unit.
Random exchange and random deletion are equivalent to introducing a certain amount of noise under the condition that the overall accuracy of the model is not affected, so that the robustness of the model can be improved. The random exchange is to randomly select two values from the column vector, exchange their positions, repeat n times in the whole vector to obtain the output vector after the random exchange is enhanced, and record as x aug1 The method comprises the steps of carrying out a first treatment on the surface of the The numerical value in the vector is deleted randomly with a certain fixed probability to obtain an output vector after the random deletion and enhancement, which is marked as x aug2 . The Seq2Seq module intercepts a 3-dimensional vector y for predicting the driving peak degree in a prediction stage pred Inputting data x for time series of peak periods peek Take out x peek Intermediate level feature x in a network after entering LSTM layer mid ，x mid I.e. the enhancement feature of Seq2 Seq.
The output result of the peak period judgment and data enhancement branch is x out When the peak period is determined, the peak period is determined and the output result x of the data enhancement branch is obtained out To enhance vector features [ x ] aug1 ,x aug2 ,x mid ]Plus the LSTM layer 29-dimensional input vector x in The method comprises the steps of carrying out a first treatment on the surface of the When the judgment is off-peak, the judgment is carried out in peak and the output result x of the data enhancement branch is obtained out 29-dimensional input vector x for LSTM layer in 。
Will x out Together with the input vector X and 3 One-Hot encoding vectors of the driving peak degree, the vector X is used as the input X of the perceived motion sickness score prediction module.
The perceived motion sickness score prediction module comprises a perceived motion sickness score prediction LSTM layer, a self-attention layer and a motion sickness prediction layer.
Prediction module for motion sickness scoreIs input into the perceived motion sickness score prediction LSTM layer to obtain a 10-dimensional output vector y of the perceived motion sickness score prediction LSTM layer out1 Perceived motion sickness score predicts 10-dimensional output vector y of LSTM layer out1 Is fed as input into the self-attention layer. Adding self-attention can make the model better focus perception motion sickness score predict 10-dimensional output vector y of LSTM layer out1 Is more useful for final prediction of perceived motion sickness, and improves prediction accuracy.
The self-attention layer comprises the following matrix operations:
the self-attention layer first predicts the perceived motion sickness score to the 10-dimensional output vector y of the LSTM layer out1 Divided into 10 dimension vectors a by dimension 1 ,a 2 ,…,a 10 Each dimension vector is multiplied by different neural network matrixes to obtain 3 different 10-dimension vectors q= (q) 1 ,q 2 ,…,q 10 ) They are calculated in each dimension i as follows:
q i ＝W q *a i
k i ＝W k *a i
v i ＝W v *a i
wherein W is k And W is v Is a learnable parameter in a neural network.
Then take the vector q of the first dimension of the q vector 1 K for each dimension of the k vector i Doing an attention multiplication to get q 1 And each k i Is a degree of proximity alpha of (a) 1,i ：
Wherein d is q 1 And k i Because of q 1 *k i The value of (2) increases with the dimension, so it is divided by
Is equivalent to the normalization effect. />
And then to the obtained alpha 1,i Vector softmax () operation results
Then the same calculation is carried out to obtain the complete product
Vector: />
Where i=1, 2 …,10.
The self-attention layer will finally
V corresponding to the dimension of v vector i Multiplication of values, in particular +.>
Multiplying by v 1 、/>
Multiplication with->
Multiplying by v 10 . Then the product results are added up to obtain b 1 At this time b 1 With q 1 And k is equal to i And v i All the associated information is:
similarly, use and obtain b 1 The same operation vector k 2 To k 10 Attention is paid to get b 2 To b 10 Self-injection is obtained10-dimensional vector y of the force layer output out2 ＝(b 1 ,b 2 ,…,b 10 )。
Finally, outputting the 10-dimensional vector y from the attention layer out2 And sending the motion sickness prediction layer.
The motion sickness prediction layer includes a fully connected layer of a neural network, a Softmax () function, and a max () function. 10-dimensional vector y output from the attention layer out2 A mapping operation is carried out through the full connection layer Linear () to output a 2-dimensional predictive vector y out2 ' 2-point fixed range degree probability representing motion symptom, 2-dimensional predictive vector y out2 ' finally, class probability is obtained through a Softmax () function, and class probability is processed through a max () function to obtain a level predictive value y of the motion sickness degree of the next minute out3 ：
y out3 ＝max(Softmax(Linear(y out2 )))
Level prediction value y of motion sickness degree of next minute out3 If 0 is not carsickness, y out3 A value of 1 indicates motion sickness 1.
A motion sickness prediction system comprises a front-end data acquisition module, a data processing module and a terminal result display module.
The front-end data acquisition module is used for acquiring quantitative data and qualitative data of a user; the front-end data acquisition module is mainly realized through a mobile terminal.
The data processing module comprises an motion sickness prediction model, wherein the motion sickness prediction model comprises a peak period judging and data enhancing branch and a motion sickness sensing score prediction module; the peak period judging and data enhancing branch comprises a peak period judging module and a data enhancing module. The rush hour judging module comprises an LSTM layer and a full connection layer; the perceived motion sickness score prediction module comprises a perceived motion sickness score prediction LSTM layer, a self-attention layer and a motion sickness prediction layer.
The terminal display module is used for receiving the prediction data obtained by the data processing module, visualizing the prediction data, and outputting a dynamic strategy through the graphical human-computer interface to guide a user to relieve motion sickness perception.
The prediction method and the prediction system for the motion sickness provided by the invention have low cost, and a user can acquire data by holding the mobile intelligent terminal equipment, so that the operation condition of the own vehicle is more conveniently monitored for a public transport operation system, the motion dynamics of the vehicle is adjusted, the riding comfort is improved, and the motion sickness perception is reduced.
Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered by the scope of the claims of the present invention.
Claims (8)
1. The motion sickness prediction system is characterized by comprising a mobile terminal, a data preprocessing module, a motion sickness prediction module and a terminal result display module;
the mobile terminal is internally provided with a front-end data acquisition module which is used for acquiring quantitative data and qualitative data of a user;
the data preprocessing module is used for preprocessing the quantitative data and the qualitative data of the user of the acquisition tool to obtain preprocessed data;
the motion sickness prediction module is used for processing the preprocessed data to obtain motion sickness prediction results; the motion sickness prediction module comprises a peak period judgment and data enhancement branch and motion sickness perception score prediction module; the peak period judging and data enhancing branch is used for judging the peak period and enhancing the corresponding data according to the judging result; the motion sickness awareness score prediction module is used for outputting motion sickness prediction data of a user;
the terminal display module is used for receiving the prediction data obtained by the data processing module, visualizing the prediction data, and outputting a dynamic strategy through the graphical human-computer interface to guide a user to relieve motion sickness perception.
2. The motion sickness prediction system of claim 1 wherein the quantitative data includes local time, current location information, X-axis acceleration magnitude at 30HZ frequency, Y-axis acceleration magnitude at 30HZ frequency, Z-axis acceleration magnitude at 30HZ frequency, X, Y, Z axis angular velocity magnitude, noise value, vibration frequency, seating location data, the qualitative data including body temperature, perceived motion sickness score, perceived congestion level in the vehicle, ventilation level, and peak driving level; the user performs an input of qualitative data on the mobile terminal.
3. The motion sickness prediction system according to claim 1, wherein the data preprocessing module preprocesses quantitative data and qualitative data as follows:
s2-1: the quantitative and qualitative data ranges of the user are adjusted to [0,1], and the normalization formula is as follows:
wherein x= [ x ] 1 ,x 2 ,…,x i …,x n ]N is a natural number, x, which is an n-dimensional column vector of qualitative data i Is the ith shaping data;
for qualitative data, segmenting the frequency band for 1 minute, converting the qualitative data into One-Hot coding by the mobile terminal, carrying out One-Hot coding by inputting the qualitative data in a manner of filling in a subjective evaluation scale, wherein motion is carried out by using a motion 10-point scale, motion is carried out by using a body temperature 3-point scale, congestion is carried out by using a congestion 3-point scale, ventilation is carried out by using a ventilation 3-point scale, and driving peak is carried out by using a driving peak 3-point scale;
s2-2: data cleaning;
if the finished quantitative data have missing values, the values of the two time slice data before and after the missing values are selected by the missing position data values, and the values are averaged and filled; if the encoded qualitative data has a missing value, the most value of all the qualitative data is used for complementing the missing; obtaining qualitative data and quantitative data after cleaning;
s2-3: data arrangement
The cleaned qualitative data and quantitative data are arranged and tidied into an input vector X which can be input by a model, and a data tag Y vector:
the input vector X comprises: x, Y, Z axis acceleration and angular velocity 6-dimensional degree vector, local time vector 1-dimensional degree vector, noise value vector 1-dimensional degree vector, perceived ventilation degree 3-dimensional degree vector, sitting position 5-dimensional degree vector, perceived crowding degree 3-dimensional degree vector, ventilation degree 3-dimensional degree vector and current time (T time) motion perception scoring 10-dimensional degree vector;
the Y vector of the data tag contains: vector y of running peak 2 point scale peek Vector y of T+1 motion 10 point scale at next time T+1 The method comprises the steps of carrying out a first treatment on the surface of the And obtaining the preprocessed data.
4. A motion sickness prediction system as claimed in claim 3 wherein said motion sickness prediction module is a long-term memory LSTM network including peak-time decision and data enhancement branch and perceived motion sickness score prediction modules; the peak period judging and data enhancing branch comprises a peak period judging module and a data enhancing module; the perceived motion sickness score prediction module comprises a perceived motion sickness score prediction LSTM layer, a self-attention layer and a motion sickness prediction layer;
training the preprocessed data as input of the motion sickness prediction module, so as to obtain a trained motion sickness prediction module;
the trained motion sickness prediction module comprises the following steps:
taking the collected input vector X as the input of a peak period judging module, and outputting a peak type by the peak period judging module, wherein the peak type is a peak period or a non-peak period; when the peak type is the peak period, the peak judgment data is input into the data enhancement module to be used as the input of the perceived motion sickness score prediction LSTM layer after the data enhancement and the peak judgment data, otherwise, the peak judgment data are directly input into the perceived motion sickness score prediction LSTM layer, the output of the perceived motion sickness score prediction LSTM layer is used as the input of the self-attention layer, the position of the self-attention layer is used as the input of the motion sickness prediction layer, and the motion sickness prediction layer outputs the motion sickness prediction result.
5. The motion sickness prediction system according to claim 4 wherein the peak period determination module data processing steps are as follows:
the input vector X is taken as the 29-dimensional input vector X of the LSTM layer in Input vector x of 29 dimensions in The size of the sequence is segmented into seq_len_batch_size by adopting a sliding window according to time sequence, and then the seq_batch_size is input into an LSTM network; wherein batch length batch_size=20, feature dimension size=29, and sequence length seq_len is time sequence length L of all data data The value after subtracting the batch_size, seq_len, is calculated as follows:
seq len ＝L data -batch_size
29-dimensional input vector x in Obtaining intermediate level features x after entering LSTM layer mid Intermediate level feature x mid The vector size is seq_len_batch_size 16: LSTM () represents LSTM layer processing;
x mid ＝LSTM(x in )
the next full link layer contains a full link convolution Linear () with output size set to 2 and a Softmax () active operation, outputting the LSTM layer result x mid Sending the motion vector into a full connection layer to obtain an output vector y for predicting the peak degree of the traveling crane pred ；
y pred ＝Softmax(Linear(x mid ))
The Softmax () function is as follows:
wherein x is i The value of the ith class vector, C is the number of classified classes, 3, x c A predictive weight vector value representing class c, e being a natural logarithm; of fully-connected layersThe output results are peak and off-peak.
6. The motion sickness prediction system of claim 5 wherein the data enhancement module comprises a parallel random exchange unit, a random deletion unit, and a Seq2Seq unit;
the random exchange unit randomly selects two values from the column vector, exchanges the positions of the two values, and repeats the whole vector n times to obtain an output vector after random exchange enhancement, which is marked as x aug1 The method comprises the steps of carrying out a first treatment on the surface of the The random deleting unit randomly deletes the numerical value in the vector with a preset fixed probability to obtain an output vector after random deletion and enhancement, and the output vector is marked as x aug2 The method comprises the steps of carrying out a first treatment on the surface of the The Seq2Seq unit intercepts the 3-dimensional vector y for predicting the peak degree of the driving in prediction in the prediction stage pred Input data x as time series of peak periods peek Take out x peek Intermediate level feature x in a network after entering LSTM layer mid ，x mid Namely the enhancement feature of the Seq2 Seq;
the output result of the peak period judgment and data enhancement branch is x out When the peak period is determined, the peak period is determined and the output result x of the data enhancement branch is obtained out To enhance vector features [ x ] aug1 ,x aug2 ,x mid ]Plus the LSTM layer 29-dimensional input vector x in The method comprises the steps of carrying out a first treatment on the surface of the When the judgment is off-peak, the judgment is carried out in peak and the output result x of the data enhancement branch is obtained out Input vector x for the 29-dimensional dimension of the LSTM layer in 。
7. The motion sickness prediction system according to claim 6 wherein the perceived motion sickness score prediction module is processed as follows: inputting the input x of the perceived motion sickness score prediction module into the perceived motion sickness score prediction LSTM layer to obtain a 10-dimensional output vector y of the perceived motion sickness score prediction LSTM layer out1 Perceived motion sickness score predicts 10-dimensional output vector y of LSTM layer out1 Is fed as input into the self-attention layer;
the self-attention layer comprises the following matrix operations:
the self-attention layer will first perceive motion sicknessScoring the 10-dimensional output vector y of the predictive LSTM layer out1 Divided into 10 dimension vectors a by dimension 1 ,a 2 ,…,a 10 Each dimension vector is multiplied by different neural network matrixes to obtain 3 different 10-dimension vectors q= (q) 1 ,q 2 ,…,q 10 ) They are calculated in each dimension i as follows:
q i ＝W q *a i
k i ＝W k *a i
v i ＝W v *a i
wherein W is k And W is v Is a learnable parameter in a neural network;
then take the vector q of the first dimension of the q vector 1 K for each dimension of the k vector i Doing an attention multiplication to get q 1 And each k i Is a degree of proximity alpha of (a) 1,i ：
Wherein d is q 1 And k i Because of q 1 *k i The value of (2) increases with the dimension, so it is divided by
Corresponding to the normalized effect;
and then to the obtained alpha 1,i Vector softmax () operation results
Then the same calculation is carried out to obtain the complete product
Vector: />
Wherein i=1, 2 …,10;
the self-attention layer will finally
V corresponding to the dimension of v vector i Multiplication of values, in particular, +.>
Multiplying by v 1 、/>
Multiplying by
Multiplying by v 10 The method comprises the steps of carrying out a first treatment on the surface of the Then the product results are added up to obtain b 1 At this time b 1 With q 1 And k is equal to i And v i All associated information:
similarly, use and obtain b 1 The same operation vector k 2 To k 10 Attention is paid to get b 2 To b 10 A 10-dimensional vector y output from the attention layer is obtained out2 ＝(b 1 ,b 2 ,…,b 10 )；
Finally, outputting the 10-dimensional vector y from the attention layer out2 Sending the motion sickness prediction layer;
the motion sickness prediction layer comprises a fully connected layer of a neural network, a Softmax () function and a max () function; 10-dimensional vector y output from the attention layer out2 A mapping operation is carried out through the full connection layer Linear () to output 2-dimensionalPredictive vector y out2 ' 2-point fixed range degree probability representing motion symptom, 2-dimensional predictive vector y out2 ' finally, class probability is obtained through a Softmax () function, and class probability is processed through a max () function to obtain a level predictive value y of the motion sickness degree of the next minute out3 ：
y out3 ＝max(Softmax(Linear(y out2 )))
Level prediction value y of motion sickness degree of next minute out3 If 0 is not carsickness, y out3 A value of 1 indicates motion sickness 1.
8. The motion sickness prediction system of claim 1 wherein the prediction system directs a user predicted to be motion sickness to a location predicted to be motion sickness based on collected quantitative and qualitative data at a public transportation means.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211734857.7A CN116130086B (en) | 2023-04-20 | 2023-04-20 | Motion sickness prediction system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211734857.7A CN116130086B (en) | 2023-04-20 | 2023-04-20 | Motion sickness prediction system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116130086A true CN116130086A (en) | 2023-05-16 |
| CN116130086B CN116130086B (en) | 2024-05-17 |
Family
ID=86298673
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211734857.7A Active CN116130086B (en) | 2023-04-20 | 2023-04-20 | Motion sickness prediction system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116130086B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190022347A1 (en) * | 2017-07-19 | 2019-01-24 | GM Global Technology Operations LLC | Motion sickness mitigation system and method |
| CN110507288A (en) * | 2019-08-29 | 2019-11-29 | 重庆大学 | Vision based on one-dimensional convolutional neural networks induces motion sickness detection method |
| KR20200073109A (en) * | 2018-12-13 | 2020-06-23 | 한국과학기술원 | Method for vr sickness assessment considering neural mismatch model and the apparatus thereof |
| CN112867528A (en) * | 2018-07-10 | 2021-05-28 | 摩拜利克斯有限责任公司 | System and method for predicting and preventing motion sickness |
| CN113104048A (en) * | 2021-03-25 | 2021-07-13 | 浙江大学 | Vehicle control system and method considering motion sickness of passengers in vehicle |
| CN113212437A (en) * | 2021-05-21 | 2021-08-06 | 吉林大学 | Motion sickness estimation method during vehicle curve running |
| CN115376041A (en) * | 2022-07-15 | 2022-11-22 | 宁波大学 | Three-dimensional panoramic video motion sickness degree prediction method based on content perception |
-
2023
- 2023-04-20 CN CN202211734857.7A patent/CN116130086B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190022347A1 (en) * | 2017-07-19 | 2019-01-24 | GM Global Technology Operations LLC | Motion sickness mitigation system and method |
| CN112867528A (en) * | 2018-07-10 | 2021-05-28 | 摩拜利克斯有限责任公司 | System and method for predicting and preventing motion sickness |
| KR20200073109A (en) * | 2018-12-13 | 2020-06-23 | 한국과학기술원 | Method for vr sickness assessment considering neural mismatch model and the apparatus thereof |
| CN110507288A (en) * | 2019-08-29 | 2019-11-29 | 重庆大学 | Vision based on one-dimensional convolutional neural networks induces motion sickness detection method |
| CN113104048A (en) * | 2021-03-25 | 2021-07-13 | 浙江大学 | Vehicle control system and method considering motion sickness of passengers in vehicle |
| CN113212437A (en) * | 2021-05-21 | 2021-08-06 | 吉林大学 | Motion sickness estimation method during vehicle curve running |
| CN115376041A (en) * | 2022-07-15 | 2022-11-22 | 宁波大学 | Three-dimensional panoramic video motion sickness degree prediction method based on content perception |
Non-Patent Citations (3)
| Title |
|---|
| CHUNG-YEN LIAO: "Using EEG and Deep Learning to Predict Motion Sickness Under Wearing a Virtual Reality Device", 《IEEE TRANSACTIONS AND JOURNALS》, 31 December 2016 (2016-12-31) * |
| 张鑫;: "运用晕动症敏感性问卷评估晕动症患者的敏感性", 中医临床研究, no. 27, 25 September 2014 (2014-09-25) * |
| 赵蕾蕾;李翀;季林红;杨铁牛;: "基于虚拟现实的自动驾驶模式中晕动受试者的脑电特征", 清华大学学报(自然科学版), no. 12, 31 December 2020 (2020-12-31) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116130086B (en) | 2024-05-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109147279B (en) | Driver fatigue driving monitoring and early warning method and system based on Internet of vehicles | |
| CN109902558B (en) | CNN-LSTM-based human health deep learning prediction method | |
| CN107506692A (en) | A kind of dense population based on deep learning counts and personnel's distribution estimation method | |
| CN113990494B (en) | Tic disorder auxiliary screening system based on video data | |
| CN110781873A (en) | Driver fatigue grade identification method based on bimodal feature fusion | |
| CN111204348A (en) | Method and device for adjusting vehicle running parameters, vehicle and storage medium | |
| CN110781872A (en) | Driver fatigue grade recognition system with bimodal feature fusion | |
| CN108537198A (en) | A kind of analysis method of the driving habit based on artificial intelligence | |
| CN111259737B (en) | Method and device for predicting failure of steering wheel of vehicle, electronic equipment and storage medium | |
| CN112529284A (en) | Private car residence time prediction method, device and medium based on neural network | |
| CN109903555A (en) | A kind of bus passenger based on big data is got off data predication method and system | |
| CN111144706A (en) | Method for grading and classifying network taxi appointment drivers | |
| CN112598331A (en) | Dynamic scheduling method, system, computer equipment and storage medium for rail transit | |
| CN117592788B (en) | Bus running risk identification method and device | |
| Aloqaily et al. | C-healthier: A cooperative health intelligent emergency response system for c-its | |
| CN111310650A (en) | Vehicle riding object classification method and device, computer equipment and storage medium | |
| CN114973214A (en) | Unsafe driving behavior identification method based on face characteristic points | |
| CN116130086B (en) | Motion sickness prediction system | |
| CN107867295A (en) | Be in danger accidents early warning method, storage device and the car-mounted terminal of probability based on vehicle | |
| Wei et al. | A driver distraction detection method based on convolutional neural network | |
| CN107590332A (en) | A kind of modularization redesign method of public transport chassis | |
| CN116443027A (en) | Vehicle comfort evaluation method, device, server and storage medium | |
| Zhu et al. | Passenger Comfort Quantification for Automated Vehicle Based on Stacking of Psychophysics Mechanism and Encoder-Transformer Model | |
| CN112036352B (en) | Training method of fatigue detection model, and fatigue driving detection method and device | |
| CN118411673B (en) | Optimization method, device, equipment and storage medium for congestion degree of subway carriage |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |