Individual difference-oriented driving fatigue music intervention method

Abstract

The invention discloses an individual difference-oriented driving fatigue music intervention method, and belongs to the technical field of driving fatigue intervention. According to the invention, identity recognition is carried out on a driver through an identity recognition module, and identity information corresponding to the driver in the system is obtained; the fatigue state of the driver is judged through a fatigue identification module; if the driver is in a fatigue state, searching an optimal intervention strategy corresponding to the current fatigue state in an optimal intervention strategy library of the driver in the system through a music intervention module, and performing music intervention on the driver according to the optimal intervention strategy; the fatigue state of the driver after intervention is judged, and if the driver is still in the fatigue state, the driver is intervened again through the music intervention module; and the change of the fatigue intervention strategy model in the intervention process is recorded, and the fatigue intervention strategy model and the optimal intervention strategy library corresponding to the driver in the system are replaced through a model adjustment module. Interference is carried out for individual differences, and the driving fatigue music intervention effect is fully guaranteed.

Description

Technical Field

[0001] The invention belongs to the technical field of driving fatigue intervention, and in particular relates to a driving fatigue music intervention method oriented to individual differences. Background Art

[0002] Driving fatigue refers to the phenomenon that after a long period of continuous driving, the driver's physiological and psychological functions become unbalanced, and his driving skills objectively decline. Data show that 21% of traffic accidents are caused by fatigue driving, accounting for more than 40% of major traffic accidents and 83% of traffic accident mortality. Therefore, it is of great practical significance to take effective measures to reduce driving fatigue.

[0003] Driving fatigue intervention refers to the use of various technical means to remind, warn, and help drivers stay awake and alert when they are fatigued, thereby reducing the risks of fatigue driving to traffic safety. These technical means include but are not limited to sound prompts, vibration prompts, image prompts, and intelligent driving assistance.

[0004] Listening to music is usually a habitual behavior that accompanies driving and is believed to help drivers pass the time easily. A study surveyed 1,780 British adults, of which 68% admitted to listening to music or radio while driving; nearly half reported humming along with music while driving; 62% felt that music could relax the mood and make the driving task more comfortable; and 25% believed that music helped them stay alert. At the same time, drivers generally believe that listening to music has a lower risk of accidents than other distracting activities such as mobile phone calls. Compared with other intervention methods, auditory intervention is easy to operate, has strong practicality, and has a controllable intervention time and can be intervened for a long time.

[0005] Existing music intervention methods for driving fatigue often use the same music type and rhythm, but different drivers may have different preferences and reactions to music, and lack of personalized music intervention may not achieve the best results. Different individuals have different physiological responses to music, and choosing the right music type is crucial to the intervention effect.

[0006] Therefore, a new technical solution is urgently needed in the prior art to solve this problem. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a driving fatigue music intervention method oriented to individual differences to solve the technical problem that the existing driving fatigue music intervention method has the defect of ignoring personalized needs.

[0008] A driving fatigue music intervention method for individual differences is provided, which discriminates the driver's fatigue state through an identity recognition module, a fatigue identification module I, a fatigue identification module II, a music intervention module, a model adjustment module and a cloud database, and selects the best music intervention strategy to perform fatigue intervention on the driver to meet individual differences; the method comprises the following steps, and the following steps are performed in sequence:

[0009] Step 1: Identify the driver's ID x through the identity recognition module, and obtain the corresponding stored preferred song list, listening habits, fatigue intervention strategy model library and optimal intervention strategy library in the cloud database;

[0010] Step 2: The fatigue identification module I identifies the fatigue state of the driver x at a set time interval;

[0011] Step 3: Driver x is identified as being in a fatigue state α, and the music intervention module calls the optimal intervention strategy library P_strategy corresponding to driver x (x) , find the best intervention strategy corresponding to fatigue state α The music intervention module performs music intervention on driver x according to the optimal intervention strategy;

[0012] Step 4: The fatigue state of driver x after each intervention is identified by the fatigue identification module II, and it is determined whether driver x is awake after the intervention. If driver x is awake, step 5 is performed; if driver x is still in a fatigue state, the fatigue state of driver x is identified, and the current fatigue state of driver x is output, and the process returns to step 3, and music intervention is performed on driver x again by the music intervention module;

[0013] Step 5: The model adjustment module records the fatigue intervention strategy model of driver x after intervention changes, thereby adjusting the fatigue intervention strategy model And the optimal intervention strategy library P_strategy (x) And upload it to the cloud database to update the fatigue intervention strategy model library and the optimal intervention strategy library.

[0014] In the step 1, the identity recognition module identifies the identity ID of driver x. If the cloud database does not store the identity information of driver x, driver x is reminded to upload a preferred song list and set a listening habit. The cloud database assigns driver x an identity ID of x (x∈N + ), and use the initial fatigue intervention strategy model stored in the cloud database as the fatigue intervention strategy model of the driver x, and the initial optimal intervention strategy library as the optimal intervention strategy library of the driver x.

[0015] In step 2, the fatigue identification module I identifies the fatigue state of the driver x:

[0016] The driving load of driver x is obtained through the cloud, and the fatigue state of driver x is divided into active fatigue and passive fatigue. The facial information of driver x is obtained through the cloud, and these two fatigue states are further divided into three levels according to the set thresholds: mild fatigue, moderate fatigue, and severe fatigue; if the driver x is judged to be in a sober state, it is judged again after T1 minutes whether the driver x is in a fatigue state; if the driver x is judged to be in a fatigue state, the fatigue identification module I determines which fatigue state α the driver x is in and outputs it.

[0017] The specific steps of performing music intervention on driver x in step 3 are as follows:

[0018] Step 1: Driver x is identified as a fatigue state α, then the optimal intervention strategy library P_strategy corresponding to driver x is called (x) , find the best intervention strategy corresponding to fatigue state α Perform music intervention on driver x according to the optimal intervention strategy;

[0019] Among them, the best intervention strategy library is divided into two types, one is the initial best intervention strategy library, and the other is the best intervention strategy library adjusted in step five;

[0020] The fatigue intervention strategy model library is composed of fatigue intervention strategy models, and β represents each fatigue intervention strategy corresponding to the fatigue intervention strategy model. The fatigue intervention strategy is a combination of music rhythm, loudness and duration, that is, rhythm×loudness×duration;

[0021] The initial fatigue intervention strategy model with fatigue state as α and fatigue intervention strategy as β As follows:

[0022]

[0023] in, is the change of the initial pre-intervention normalized fatigue composite value over time t; is the change of the normalized fatigue comprehensive value after the initial setting intervention over time t, where α is the driving fatigue state, β represents the fatigue intervention strategy, and (0) represents the initial setting, that is, the driver ID is 0; are the initial model parameters; The moment for the musical intervention to begin; It is the moment when the driver enters the awake state from the fatigue state after the music intervention;

[0024] The fatigue intervention strategy model formula is used to calculate the optimal intervention strategy β under each fatigue state α;

[0025] The initial fatigue intervention strategy library consists of the initial optimal initial fatigue intervention strategy under each fatigue state, expressed as:

[0026]

[0027] Among them, P_strategy (0) The initial fatigue intervention strategy library; strategy1 (0) With β i1 The fatigue intervention strategy for the first fatigue state, namely mild active fatigue; strategy2 (0) With β i2 The fatigue intervention strategy for the first fatigue state, namely moderate active fatigue; strategy3 (0) With β i3 The fatigue intervention strategy for the first fatigue state, i.e. severe active fatigue; strategy 4 (0) With β i4 The fatigue intervention strategy for the first fatigue state, namely mild passive fatigue; strategy5 (0) With β i5 The fatigue intervention strategy for the first fatigue state, namely moderate passive fatigue; strategy6 (0) With β i6 The fatigue intervention strategy for the first fatigue state, i.e., severe passive fatigue; u (1) ,u (2) ,u (3) ,u (4) ,u (5) ,u (6) ∈{fast tempo, medium tempo, slow tempo}, v (1) ,v (2) ,v (3) ,v (4) ,v (5) ,v (6) ∈{high volume, medium volume, low volume}, q (1) ,q (2) ,q (3) ,q (4) ,q (5) ,q (6) ∈{10min,20min,30min,40min,50min,60min};

[0028] Driver x (x = 1, 2, 3...) uses this method for the first time, then the fatigue intervention strategy model F corresponding to driver x (x) (t) and the optimal intervention strategy library P_strategy (x)All used the initial fatigue intervention strategy model and the initial optimal intervention strategy library;

[0029] Step 2: According to the optimal intervention strategy described in the first step, find songs with a rhythm that matches this strategy in the driver x's preferred song list, and perform fatigue intervention on the driver in the music intervention module according to the loudness and duration settings, specifically:

[0030] Fatigue identification module I determines that the fatigue state of driver x is α, and the optimal intervention strategy library P_strategy corresponding to driver x (x) Find the best intervention strategy corresponding to fatigue state α in strategy α =β→rhythm u, loudness v, duration q;

[0031] In the preferred song list of driver x, find the song corresponding to rhythm u, arrange the order of songs according to driver x's listening habits, and use loudness v and duration q to perform music intervention on driver x.

[0032] Adjust the fatigue intervention strategy model in step 5 And the optimal intervention strategy library P_strategy (x) The specific method is:

[0033] Adjusting the fatigue intervention strategy model means adjusting the model parameters θ(a, b, c, k, l, m, n) using stochastic gradient descent (SGD). The steps are as follows:

[0034] Set the parameters of the initial fatigue intervention strategy model as θ(a, b, c, k, l, m, n) = θ (0) ;

[0035] The first step is to define the loss function:

[0036]

[0037] In the formula, f i is the normalized fatigue comprehensive value of the driver recorded after the intervention, where N is the total number of samples, i∈[1,N]; F i (t i ; θ) is the normalized comprehensive fatigue value of the driver stored in the current cloud database;

[0038] The second step is to get the complete gradient:

[0039]

[0040] Simplifying the expression, we get the gradient vector ▽R(θ):

[0041]

[0042] Update parameters:

[0043]

[0044] θ new is the updated model parameter, η is the learning rate, which is used to control the step size of each parameter update;

[0045] Get new model parameter values ​​through stochastic gradient descent SGD Replace the original model parameter values ​​to obtain a new fatigue intervention strategy model Recalculate and adjust the optimal intervention strategy library P_strategy for driver x (x,new) ;

[0046] The third step is to integrate the new fatigue intervention strategy model and the optimal intervention strategy library P_strategy (x,new) The data is stored in the cloud database to replace the original model and data, and corresponds to the ID of driver x for easy retrieval next time.

[0047] Through the above design scheme, the present invention can bring the following beneficial effects:

[0048] The present invention identifies the driver through an identity recognition module to obtain the identity information corresponding to the driver in the system; the fatigue state of the driver is judged through a fatigue recognition module; if the driver is in a certain fatigue state, the best intervention strategy corresponding to the current fatigue state is searched in the best intervention strategy library of the driver in the system through a music intervention module, and music intervention is performed on the driver according to the strategy; the fatigue state of the driver after the intervention is judged, and if the driver is still in the fatigue state, the driver is intervened again through the music intervention module; the changes of the fatigue intervention strategy model during the intervention process are recorded, and the fatigue intervention strategy model and the best intervention strategy library corresponding to the driver in the system are replaced through a model adjustment module, so as to realize the best driving fatigue music intervention for individual differences.

[0049] The present invention performs music intervention according to individual differences of drivers, fully guarantees the effectiveness and pertinence of music intervention for driving fatigue, and has important practical significance for alleviating driving fatigue. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The present invention is further described below with reference to the accompanying drawings and specific embodiments:

[0051] Figure 1 This is a schematic flow chart of a driving fatigue music intervention method for individual differences according to the present invention;

[0052] Figure 2This is a flow chart of an identity recognition module in a driving fatigue music intervention method for individual differences of the present invention;

[0053] Figure 3 This is a flow chart of fatigue identification module I in a driving fatigue music intervention method for individual differences of the present invention;

[0054] Figure 4 A flowchart of a music intervention module in a driving fatigue music intervention method for individual differences according to the present invention;

[0055] Figure 5 This is a flow chart of fatigue identification module II in a driving fatigue music intervention method for individual differences of the present invention;

[0056] Figure 6 A flow chart of a model adjustment module in a driving fatigue music intervention method for individual differences according to the present invention;

[0057] Figure 7 This is a flow chart of the gradient descent method in the model adjustment module of a driving fatigue music intervention method oriented to individual differences of the present invention. DETAILED DESCRIPTION

[0058] A method for music intervention for driving fatigue targeting at individual differences comprises the following steps: a driver facial recognition module is provided on the identity recognition module for obtaining facial information and uploading it to a cloud database, the identity recognition module identifies the driver and obtains the identity information corresponding to the driver in the cloud database; the facial information of the driver x is obtained through the cloud by the fatigue recognition module, and the fatigue state of the driver is judged. If the driver is in a certain fatigue state, the best intervention strategy corresponding to the current fatigue state is searched in the best intervention strategy library of the driver through the music intervention module, and the driver is intervened with music according to the strategy, and the fatigue state of the driver after the intervention is judged again. If the driver is still in the fatigue state, the driver is intervened again through the music intervention module; the changes of the fatigue intervention strategy model during the intervention process are recorded, and the fatigue intervention strategy model and the best intervention strategy library corresponding to the driver in the cloud database are replaced through the model adjustment module.

[0059] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical scheme in the present invention is clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the present invention is not limited to the following embodiments, and the specific implementation method can be determined according to the technical scheme of the present invention and the actual situation. In order to avoid confusing the essence of the present invention, the known methods, processes and procedures are not described in detail.

[0060] like Figure 1As shown, a driving fatigue music intervention method for individual differences includes the following steps:

[0061] Step 1: Identify the driver's ID x through the identity recognition module, and obtain the corresponding preferred song list, listening habits, fatigue intervention strategy model and optimal intervention strategy library in the cloud database.

[0062] If the cloud database does not store the identity information of the driver x, the driver x is reminded to upload the preferred song list and set the listening habits, and the identity ID of the driver x is given as x (x∈N + ) and use the initial fatigue intervention strategy model as the fatigue intervention strategy model of the driver x, and the initial optimal intervention strategy library as the optimal intervention strategy library of the driver x.

[0063] Step 2: The fatigue identification module I identifies the fatigue status of the driver x every T1 minutes.

[0064] The normalized fatigue comprehensive value of driver x is calculated every T1 minutes as the normalized fatigue comprehensive value Fb of driver x before intervention (x) , record Fb (x) Curve Fb changes with time t (x) (t).

[0065]

[0066] Among them, Fb (x) (t) is the normalized fatigue comprehensive value change curve before the driver x intervenes, Fb (x) The larger the value of (t), the more fatigued the driver x is; t is the driving time of driver x; a, b, c are model parameters determined by step five.

[0067] By Facebook (x)(t) curve determines whether driver x is in a fatigue state. If driver x is awake, then after T1 minutes, it is determined again whether driver x is in a fatigue state; if driver x is in a fatigue state, the fatigue identification module I determines which fatigue state α driver x is in and outputs it. The driving load of driver x is obtained through the cloud. If the driving load of driver x is greater than or equal to the threshold z1, driver x is determined to be actively fatigued; if the driving load of driver x is less than the threshold z, the driver is determined to be passively fatigued. The driving load is determined by information such as traffic flow and vehicle movement. The two fatigue states are divided into three levels according to the set thresholds based on the facial information of driver x: mild fatigue, moderate fatigue, and severe fatigue. This is the prior art, not the focus of the protection of the present invention, and will not be elaborated in detail. Set the driving fatigue state to α, then α∈(1,2,3,4,5,6). Among them, 1=mild active fatigue, 2=moderate active fatigue, 3=severe active fatigue, 4=mild passive fatigue, 5=moderate passive fatigue, and 6=severe passive fatigue.

[0068] Step 3: When driver x is identified as a fatigue state α, the optimal intervention strategy library P_strategy corresponding to driver x is called (x) , find the best intervention strategy corresponding to fatigue state α Perform music intervention on driver x according to the optimal intervention strategy.

[0069] There are two types of optimal intervention strategy libraries: one is the initial optimal intervention strategy library. The other is the optimal intervention strategy library adjusted in step 5. When the cloud database does not store the information of driver x, the ID of driver x is set to x, and the optimal intervention strategy library corresponding to driver x is set to the initial optimal intervention strategy library P_strategy (x) =P_strategy (0) For the initial setting, the default ID of driver x is 0; and the optimal driver intervention strategy library P_strategy after adjustment in step 5 (x) It will vary from driver to driver.

[0070] The fatigue intervention strategy model library is composed of fatigue intervention strategies. Fatigue intervention strategies are combinations of music rhythm, loudness, and duration, that is, rhythm × loudness × duration. The rhythm of music depends on the size of the song's beats per minute (bpm). The bpm of fast-paced music is greater than 120, the bpm of medium-paced music is 70-120, and the bpm of slow-paced music is less than 70; the loudness of music is 70-80dB for high loudness, 60-70dB for medium loudness, and 50-60dB for low loudness; the music intervention duration is the time of music playback, which is 10min, 20min, 30min, 40min, 50min, and 60min. When β represents the fatigue intervention strategy, β∈{1,2,3,...,54}, that is, 3(rhythm)×3(loudness)×6(duration). Each β has its corresponding fatigue intervention strategy model.

[0071] The initial fatigue intervention strategy model with fatigue state as α and fatigue intervention strategy as β Depend on and composition:

[0072]

[0073] in, It is obtained from formula (1), that is, the change of the initial normalized fatigue comprehensive value before intervention with time t; is the change of the normalized fatigue comprehensive value after the initial setting intervention over time t. Where α is the driving fatigue state, α∈(1,2,3,4,5,6). β represents the fatigue intervention strategy, β∈{1,2,3,...,54}. (0) represents the initial setting, that is, the driver ID is 0. are the initial model parameters, where is related to the fatigue level of driver 0 before intervention in fatigue state α; Indicates the rate at which the comprehensive fatigue value of driver 0 changes over time; represents the fatigue growth rate constant of driver 0 in the process of reaching fatigue state α; It represents the rate of fatigue reduction after the intervention strategy β is implemented for driver 0 in fatigue state α; It represents the intensity of the intervention effect of intervention strategy β when driver 0 is in fatigue state α; It represents the rate at which the comprehensive fatigue value recovers over time after the intervention strategy β is implemented when the driver 0 is in fatigue state α; When driver 0 is in fatigue state α, the rate constant of fatigue recovery after intervention strategy β is implemented; The moment for the musical intervention to begin; The time when the driver enters the awake state from the fatigue state after the music intervention. Parameters corresponding to different fatigue states Different fatigue intervention strategies correspond to as well as Are also different.

[0074] Suppose the initial intervention effect value corresponding to the βth fatigue intervention strategy under the αth fatigue state for:

[0075]

[0076] in, is the cumulative fatigue value before intervention when fatigue state is α, is the cumulative fatigue value after intervention strategy β is used when fatigue state is α, For curve The horizontal coordinate of the inflection point.

[0077] The initial optimal intervention effect under the αth fatigue state is:

[0078]

[0079] That is, the initial optimal intervention effect under the αth fatigue state is the maximum value of the initial intervention effect values ​​corresponding to all fatigue intervention strategies under the αth fatigue state. In the formula, α is the driving fatigue state, α∈(1,2,3,4,5,6). β represents the fatigue intervention strategy, β∈{1,2,3,...,54}.

[0080] In formula (5) For driver 0, the best intervention effect is achieved by using the βth fatigue intervention strategy to deal with the αth fatigue state among all fatigue intervention strategies. This effect is expressed as To measure.

[0081] The initial optimal intervention effect library P_effect (0) for:

[0082]

[0083] The initial optimal intervention effect in the αth fatigue state was obtained Then the fatigue intervention strategy β adopted for the initial optimal intervention effect is the initial optimal fatigue intervention strategy under the αth fatigue state:

[0084]

[0085] is the initial optimal fatigue intervention strategy under the αth fatigue state;

[0086] u∈{fast pace, medium pace, slow pace}, v∈{high volume, medium volume, low volume}, q∈{10min,20min,30min,40min,50min,60min}.

[0087] The initial fatigue intervention strategy library consists of the initial optimal fatigue intervention strategy under each fatigue state. That is,

[0088]

[0089] Among them, P_strategy (0) The initial fatigue intervention strategy library; strategy1 (0) With β i1 The fatigue intervention strategy for the first fatigue state, namely mild active fatigue; strategy2 (0) With β i2 The fatigue intervention strategy for the first fatigue state, namely moderate active fatigue; strategy3 (0) With β i3 The fatigue intervention strategy for the first fatigue state, i.e. severe active fatigue; strategy 4 (0) With β i4 The fatigue intervention strategy for the first fatigue state, namely mild passive fatigue; strategy5 (0) With β i5 The fatigue intervention strategy for the first fatigue state, namely moderate passive fatigue; strategy6 (0) With β i6 The fatigue intervention strategy for the first fatigue state, i.e., severe passive fatigue; u (1) ,u (2) ,u (3) ,u (4) ,u (5) ,u (6) ∈{fast tempo, medium tempo, slow tempo}, v (1) ,v (2) ,v (3) ,v (4) ,v (5) ,v (6) ∈{high volume, medium volume, low volume}, q (1) ,q (2) ,q (3) ,q (4) ,q (5) ,q (6) ∈{10min,20min,30min,40min,50min,60min};

[0090] After completing the above settings, start using it. The first step is that when driver x (x = 1, 2, 3...) uses the method or system for the first time, the fatigue intervention strategy model F corresponding to driver x (x) (t), best intervention effect library P_effect (x) And the optimal intervention strategy library P_strategy (x) All use the initial model or library, that is,

[0091]

[0092] In the second step, according to the optimal intervention strategy described in the first step, songs with a rhythm that meets this strategy are found in the driver x's preferred playlist, and fatigue intervention is performed on driver x in the music intervention module according to the loudness and duration settings.

[0093] When the fatigue identification module I determines that the fatigue state of driver x is α, the music intervention module selects the optimal intervention strategy library P_strategy corresponding to driver x. (x) Find the best intervention strategy corresponding to fatigue state α in strategy α =β→rhythm u, loudness v, duration q.

[0094] In the preferred song list of driver x, find the song corresponding to rhythm u, arrange the order of the songs according to driver x's listening habits, and use loudness v and duration q to intervene in the driver's music.

[0095] During the entire music intervention process, the model adjustment module records the fatigue intervention strategy model of driver x. changes.

[0096] Step 4: Identify the fatigue state of driver x after intervention in fatigue identification module II.

[0097] In the fatigue identification module II, the fatigue state of driver x after the intervention is identified to determine whether driver x is awake after the intervention. If driver x is awake, proceed to step 5; if driver x is still in a fatigued state, the fatigue state of driver x is identified and the current fatigue state of driver x is output, and the music intervention module performs music intervention on driver x again.

[0098] Step 5: The model adjustment module records the fatigue intervention strategy model of driver x after each intervention changes, thereby adjusting And the optimal intervention strategy library P_strategy (x) .

[0099] Adjusting the fatigue intervention strategy model F(t) means adjusting the model parameters θ(a, b, c, k, l, m, n). The Stochastic Gradient Descent (SGD) method is used.

[0100] Gradient descent method is an optimization algorithm based on gradient descent. It iteratively updates the parameters along the negative gradient direction of the error function to minimize the sum of the square errors between the fitting function and the actual data. This algorithm is applicable to various nonlinear fitting problems.

[0101] The basic idea of ​​stochastic gradient descent is to use only the gradient of one sample to update the model parameters at each iteration. Since only the gradient of one sample needs to be calculated at a time, stochastic gradient descent is efficient on large-scale datasets. When using SGD, the goal is to minimize the loss function. The loss function is a function that measures the difference between the model's predicted value and the actual value.

[0102] When using stochastic gradient descent, the steps to adjust F(t;θ) are as follows:

[0103] The initial parameters are the parameters of the initial fatigue intervention strategy model θ(a, b, c, k, l, m, n) = θ (0) .

[0104] The first step is to define the loss function:

[0105]

[0106] Among them, f i is the normalized fatigue comprehensive value of the driver recorded after intervention; where N is the total number of samples, i∈[1,N]; F i (t i ; θ) is the normalized comprehensive fatigue value of the driver stored in the current cloud database;

[0107] In the second step, for each sample i, perform the following operations until in, The gradient is the vector of partial derivatives of the loss function R with respect to each parameter θ, and threshold is the threshold:

[0108] Compute the gradient:

[0109]

[0110] Among them, R i is the loss of a single sample, For R i The gradient with respect to the parameters θ.

[0111] First, calculate the single sample error term:

[0112] e i =(f i -F(t i ;θ)) 2 ;

[0113] Then, calculate the single sample error term e i The gradient of the loss function (f i -F(t i ;θ)) 2 The derivative with respect to θ is:

[0114]

[0115] where θ j ∈θ(a,b,c,k,l,m,n),j∈{1,2,3...7} are the parameters of the model.

[0116] Sum the gradients of all samples to get the loss function for each parameter θ j The total gradient of:

[0117]

[0118] Full gradient:

[0119]

[0120] Simplify the expression and get the gradient vector

[0121]

[0122] Finally, update the parameters.

[0123]

[0124] θ new is the updated model parameter. η is the learning rate, which is used to control the step size of each parameter update.

[0125] Get new model parameter values ​​through SGD Replace the original model parameter values ​​to obtain a new fatigue intervention strategy model Recalculate the optimal intervention effect under the αth fatigue state Thus adjusting the optimal intervention effect library P_effect of driver x (x,new) And the optimal intervention strategy library P_strategy (x,new) .

[0126] The third step is to adjust the model and P_effect (x) 、P_strategy(x) The data and models are stored in the cloud database to replace the original data and models, and correspond to the driver's ID for easy retrieval next time.

[0127]

[0128] The adjusted model is in line with the individual differences of drivers and can achieve better intervention effects.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions and ideas of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A driving fatigue music intervention method for individual differences, which discriminates the driver's fatigue state through an identity recognition module, a fatigue identification module I, a fatigue identification module II, a music intervention module, a model adjustment module and a cloud database, and selects the best music intervention strategy to intervene in the driver's fatigue to meet individual differences; its characteristics are: The method comprises the following steps, and the following steps are performed in sequence: Step 1: Identify the driver's ID x through the identity recognition module, and obtain the corresponding stored preferred song list, listening habits, fatigue intervention strategy model library and optimal intervention strategy library in the cloud database; Step 2: The fatigue identification module I identifies the fatigue state of the driver x at a set time interval; Step 3: Driver x is identified as being in a fatigue state α, and the music intervention module calls the optimal intervention strategy library P_strategy corresponding to driver x (x) , find the best intervention strategy corresponding to fatigue state α The music intervention module performs music intervention on driver x according to the optimal intervention strategy; Step 4: The fatigue state of driver x after each intervention is identified by the fatigue identification module II, and it is determined whether driver x is awake after the intervention. If driver x is awake, step 5 is performed; if driver x is still in a fatigue state, the fatigue state of driver x is identified, and the current fatigue state of driver x is output, and the process returns to step 3, and music intervention is performed on driver x again by the music intervention module; Step 5: The model adjustment module records the fatigue intervention strategy model of driver x after intervention changes, thereby adjusting the fatigue intervention strategy model and the optimal intervention strategy library P_strategy (x) And upload it to the cloud database to update the fatigue intervention strategy model library and the optimal intervention strategy library.

2. The driving fatigue music intervention method for individual differences according to claim 1 is characterized by: In the step 1, the identity recognition module identifies the identity ID of driver x. If the cloud database does not store the identity information of driver x, driver x is reminded to upload a preferred song list and set a listening habit. The cloud database assigns driver x an identity ID of x (x∈N + ), and use the initial fatigue intervention strategy model stored in the cloud database as the fatigue intervention strategy model of the driver x, and the initial optimal intervention strategy library as the optimal intervention strategy library of the driver x.

3. The driving fatigue music intervention method for individual differences according to claim 1 is characterized by: In step 2, the fatigue identification module I identifies the fatigue state of the driver x: The driving load of driver x is obtained through the cloud, and the fatigue state of driver x is divided into active fatigue and passive fatigue. The facial information of driver x is obtained through the cloud, and these two fatigue states are further divided into three levels according to the set thresholds: mild fatigue, moderate fatigue, and severe fatigue; if the driver x is judged to be in a sober state, it is judged again after T1 minutes whether the driver x is in a fatigue state; if the driver x is judged to be in a fatigue state, the fatigue identification module I determines which fatigue state α the driver x is in and outputs it.

4. The driving fatigue music intervention method for individual differences according to claim 1 is characterized by: The specific steps of performing music intervention on driver x in step 3 are as follows: Step 1: Driver x is identified as a fatigue state α, then the optimal intervention strategy library P_strategy corresponding to driver x is called (x) , find the best intervention strategy corresponding to fatigue state α Perform music intervention on driver x according to the optimal intervention strategy; Among them, the best intervention strategy library is divided into two types, one is the initial best intervention strategy library, and the other is the best intervention strategy library adjusted in step five; The fatigue intervention strategy model library is composed of fatigue intervention strategy models, and β represents each fatigue intervention strategy corresponding to the fatigue intervention strategy model. The fatigue intervention strategy is a combination of music rhythm, loudness and duration, that is, rhythm×loudness×duration; The initial fatigue intervention strategy model with fatigue state as α and fatigue intervention strategy as β As follows: in, is the change of the initial pre-intervention normalized fatigue composite value over time t; is the change of the normalized fatigue comprehensive value after the initial setting intervention over time t, where α is the driving fatigue state, β represents the fatigue intervention strategy, and (0) represents the initial setting, that is, the driver ID is 0; are the initial model parameters; The moment for the musical intervention to begin; It is the moment when the driver enters the awake state from the fatigue state after the music intervention; The fatigue intervention strategy model formula is used to calculate the optimal intervention strategy β under each fatigue state α; The initial fatigue intervention strategy library consists of the initial optimal initial fatigue intervention strategy under each fatigue state, expressed as: Among them, P_strategy (0) The initial fatigue intervention strategy library; strategy1 (0) With β i1 The fatigue intervention strategy for the first fatigue state, namely mild active fatigue; strategy2 (0) With β i2 The fatigue intervention strategy for the first fatigue state, namely moderate active fatigue; strategy3 (0) With β i3 The fatigue intervention strategy for the first fatigue state, i.e. severe active fatigue; strategy 4 (0) With β i4 The fatigue intervention strategy for the first fatigue state, namely mild passive fatigue; strategy5 (0) With β i5 The fatigue intervention strategy for the first fatigue state, namely moderate passive fatigue; strategy6 (0) With β i6 The fatigue intervention strategy for the first fatigue state, i.e., severe passive fatigue; u (1) ,u (2) ,u (3) ,u (4) ,u (5) ,u (6) ∈{fast tempo, medium tempo, slow tempo}, v (1) ,v (2) ,v (3) ,v (4) ,v (5) ,v (6) ∈{high volume, medium volume, low volume}, q (1) ,q (2) ,q (3) ,q (4) ,q (5) ,q (6) ∈{10min,20min,30min,40min,50min,60min}; Driver x (x = 1, 2, 3...) uses this method for the first time, then the fatigue intervention strategy model F corresponding to driver x (x) (t) and the optimal intervention strategy library P_strategy (x) All used the initial fatigue intervention strategy model and the initial optimal intervention strategy library; Step 2: According to the optimal intervention strategy described in the first step, find songs with a rhythm that matches this strategy in the driver x's preferred song list, and perform fatigue intervention on the driver in the music intervention module according to the loudness and duration settings, specifically: Fatigue identification module I determines that the fatigue state of driver x is α, and the optimal intervention strategy library P_strategy corresponding to driver x (x) Find the best intervention strategy corresponding to fatigue state α in strategy α =β→rhythm u, loudness v, duration q; In the preferred song list of driver x, find the song corresponding to rhythm u, arrange the order of songs according to driver x's listening habits, and use loudness v and duration q to perform music intervention on driver x.

5. The driving fatigue music intervention method for individual differences according to claim 1 is characterized by: Adjust the fatigue intervention strategy model in step 5 and the optimal intervention strategy library P_strategy (x) The specific method is: Adjusting the fatigue intervention strategy model means adjusting the model parameters θ(a, b, c, k, l, m, n) using stochastic gradient descent (SGD). The steps are as follows: Set the parameters of the initial fatigue intervention strategy model as θ(a, b, c, k, l, m, n) = θ (0) ; The first step is to define the loss function: In the formula, f i is the normalized fatigue comprehensive value of the driver recorded after the intervention, where N is the total number of samples, i∈[1,N]; F i (t i ; θ) is the normalized comprehensive fatigue value of the driver stored in the current cloud database; The second step is to get the complete gradient: Simplifying the expression, we get the gradient vector ▽R(θ): Update parameters: i new =θ-η▽R(θ); θ new is the updated model parameter, η is the learning rate, which is used to control the step size of each parameter update; Get new model parameter values ​​through stochastic gradient descent SGD Replace the original model parameter values ​​to obtain a new fatigue intervention strategy model Recalculate and adjust the optimal intervention strategy library P_strategy for driver x (x ,new) ; The third step is to integrate the new fatigue intervention strategy model and the optimal intervention strategy library P_strategy (x,new) The data is stored in the cloud database to replace the original model and data, and corresponds to the driver's ID, so that it can be easily called next time.

Images