Vehicle-mounted man-machine interaction method and system based on big data

Abstract

The invention provides a vehicle-mounted man-machine interaction method and system based on big data, and the method achieves the automatic setting of a current user for a vehicle in a current scene through different schemes provided by three embodiments, and builds a candidate set through different methods in the three embodiments. The relationships between users, between settings and between users and settings are mined, the user feature vectors are utilized to represent the user features, and the setting feature vectors are utilized to represent the setting features. And meanwhile, the preference degree of the user to the setting is calculated by utilizing the correlation among the characteristics, and the vehicle is automatically set according to the set preference degree sequence, so that compared with the existing independent set learning and human-computer interaction for a single user, the method is more intelligent, and the user experience is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of human - machine interaction, and particularly relates to a vehicle - mounted human - machine interaction method and system based on big data. Background Art

[0002] In vehicle - mounted human - machine interaction, the same user may have different preferences for vehicle human - machine interaction settings in different scenarios. Taking the vehicle air - conditioner as an example, a male user likes to set the air - conditioner temperature to 18°C on a hot day with direct sunlight, and likes to set the air - conditioner temperature to 26°C on a cold day with autumn wind. Different users may also have different preferences for air - conditioner settings in the same scenario. For example, in the same hot day with direct sunlight, a female user likes to set the air - conditioner temperature to 22°C. In order to provide personalized human - machine interaction services to users, in the prior art, the system uses machine - learning models to learn users' preferences and habits based on the historical data of users' human - machine interaction settings. The learned model can predict the human - machine interaction settings that users may like according to the current scenario.

[0003] However, the above - mentioned machine - learning models in the prior art learn independently for each user, and each user has an independent model. When the user's interaction habit changes or the user makes new settings in a new scenario, the update of the machine - learning model can only be for a single user and has no effect on other users. However, when this user makes new settings in a new scenario, other users with similar user characteristics are also very likely to make the same settings in the same new scenario. However, the existing machine - learning models do not fully explore the relationships between users, between users and settings, and between settings, resulting in insufficient intelligence in human - machine interaction and room for further improvement in user experience. Summary of the Invention

[0004] The present invention aims to improve the intelligence level of vehicle - mounted human - machine interaction and further enhance the human - machine interaction experience, and provides a vehicle - mounted human - machine interaction method and system based on big data.

[0005] To achieve this purpose, the present invention adopts the following technical solutions:

[0006] Provide a vehicle - mounted human - machine interaction method based on big data, including the steps of:

[0007] L1, obtaining the vehicle - mounted human - machine interaction habit data of the current user, including user information, scenario information, setting information, and setting interaction preferences, to construct a user feature vector associated with the current user and obtain a setting feature vector;

[0008] L2, constructing a candidate set for predicting vehicle settings for the current user;

[0009] L3, extract the highest preference degree among the preference degrees of each setting in the candidate set for the current user, and use the setting associated with the highest preference degree as the prediction result for vehicle setting for the current user.

[0010] Preferably, the user information includes any one or more of: user ID, user age, user gender, clothing index, physiological information, and emotional information;

[0011] The scenario information includes any one or more of: outdoor temperature, indoor temperature, vehicle speed, remaining battery power, remaining fuel quantity, weather condition, light intensity, current time, geographical information where the vehicle is located, road navigation information, current road type, and driving slope;

[0012] The setting information includes any one or more of: air conditioner setting information, seat setting information, fragrance setting information, window setting information, atmosphere light setting information, and audio setting information.

[0013] Preferably, the user feature vector is a vector composed of the preference degrees of n settings that the user u has interacted with in history; the obtained setting feature vector is a vector composed of the similarities of the setting s with each of the n settings. i The preference degrees of n settings that the user u has interacted with in history; the obtained setting feature vector is a vector composed of the similarities of the setting s with each of the n settings. 1 With each setting s among the n settings 2 of the similarities.

[0014] Preferably, the method for constructing the candidate set in step L2 is:

[0015] Obtain the n settings that the current user has interacted with in history, and then find k settings that have setting feature similarities with each of the n settings. A total of n×k settings constitute the candidate set used as the basis for vehicle automatic setting for the current user.

[0016] Preferably, the similarity between setting s 1 and setting s 2 is:

[0017]

[0018] S C = S 1 ∩S 2 S c represents the user intersection of the user set S 1 associated with the setting s 1 and the user set S 2 associated with the setting s 2 .

[0019] Preferably, the user feature vector is: user u iA vector composed of the similarities with the top p users having user feature similarities with it and arranged in descending order of similarity; the obtained setting feature vector is: the preferences of p users for the setting s interacted with for the current scenario 1 constitute a vector.

[0020] Preferably, in step L2, the method for constructing the candidate set is:

[0021] Find p users having user feature similarities with the current user, and then find n settings where the scenario information of each user's historical interaction among the p users is consistent with the current scenario information of the vehicle to be set by the current user. A total of n×p settings are used as the candidate set.

[0022] Preferably, the similarity between user u 1 and u 2 is:

[0023]

[0024] K C = K 1 ∩K 2 where K c represents the set intersection of the set K of each setting of the historical interaction of the user u 1 and the set K of each setting of the historical interaction of the user u 1 with the user u 2 and the set K of each setting of the historical interaction of the user u 2 with the user u

[0025] Preferably, the setting is an extended setting, and the extended setting is: the fusion information including the setting information and the scenario information for setting the vehicle in the corresponding scenario.

[0026] Preferably, step L3 specifically includes the steps:

[0027] L31, remove the duplicate and inconsistent settings with the current scenario information in the candidate set;

[0028] L32, calculate the preference degree of the current user for each setting remaining in the candidate set after being removed in step L31 by dot product;

[0029] L33, extract the highest preference degree among the preference degrees of the current user for each setting remaining in the candidate set after being removed in step L31, and use the setting associated with the highest preference degree as the prediction result for setting the vehicle for the current user.

[0030] Preferably, in step L1, after separating discrete information and continuous information from the in-vehicle human-computer interaction habit data area and performing preprocessing, the preprocessed vector is input into a neural network to output the user feature vector and the setting feature vector;

[0031] The user discrete information of the user information includes user ID and / or user gender, and the user continuous information includes user age and / or clothing index; the scene discrete information of the scene information includes weather conditions and / or geographical information where the vehicle is located, and the scene continuous information includes outside temperature and / or inside temperature; the setting discrete information of the setting information includes air outlet mode and / or circulation mode, and the setting continuous information includes set temperature and / or set wind speed;

[0032] The discrete and / or continuous user information and / or the scene information are input into a user neural network after preprocessing to obtain the user feature vector; the discrete and / or continuous setting information are input into a setting neural network after preprocessing to obtain the setting feature vector;

[0033] The preprocessing method for user discrete information, scene discrete information, and setting discrete information is: converting the one-hot vector corresponding to the discrete information into an Embedding matrix to obtain an Embedding vector as the discrete information preprocessing vector;

[0034] The preprocessing method for user continuous information, scene continuous information, and setting continuous information is: after binning the continuous information, converting it into discrete information, and then multiplying the one-hot vector corresponding to the discrete information by the Embedding matrix to obtain an Embedding vector as the discrete information preprocessing vector; or performing normalization processing on the continuous information.

[0035] Preferably, the binning process is to group the continuous information, and the normalization process is to map the continuous values of the continuous information to a specified numerical interval using linear transformation.

[0036] Preferably, the method for the neural network to predict the settings of the vehicle by the current user in the current scene is:

[0037] Calculate the cosine similarity between the user feature vector and the setting feature vector as the preference degree of the current user for the setting in the current scene, and use the setting with the highest preference degree for vehicle control.

[0038] Preferably, the calculation method for the user's preference degree for the setting is:

[0039] After the user makes a setting for the vehicle in a specified scene, add "1" to the value of the setting preference degree corresponding to the setting of the vehicle by the user in the specified scene;

[0040] After the system makes the above settings for the vehicle by the user in the specified scenario, if the system monitors that it enters the specified scenario again and the user does not change the settings made by the system after entering the specified scenario this time, the preference degree of the user for the settings of the vehicle in the specified scenario is incremented by "1"; if the user changes the settings made by the system, the preference degree of the user for the settings in the specified scenario is decremented by "1".

[0041] Or map the preference as positive to "1" and the preference as negative to "-1".

[0042] Preferably, step L3 specifically includes the steps of:

[0043] L301, calculating the preference degree of the current user for each setting in the candidate set by dot product or cosine similarity;

[0044] L302, extracting the highest preference degree among the preference degrees of the current user for each setting in the candidate set, and using the setting associated with the highest preference degree as the prediction result for the vehicle setting of the current user.

[0045] The present invention also provides a vehicle-mounted human-computer interaction system based on big data, which can implement the vehicle-mounted human-computer interaction method based on big data described above.

[0046] The present invention realizes the automatic setting of the vehicle for the current user in the current scenario through different solutions provided by 3 embodiments. In the 3 embodiments, candidate sets are constructed by different methods, the relationships between users and users, between settings and settings, and between users and settings are mined, and user feature vectors are used to represent user features, and setting feature vectors are used to represent setting features. At the same time, the correlation between the above features is used to calculate the preference degree of the user for the setting, and the vehicle is automatically set according to the sorted preference degree of the setting. Compared with the existing independent setting learning for a single user, the human-computer interaction is more intelligent and the user experience is greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0048] Figure 1 is the implementation step diagram of the vehicle-mounted human-computer interaction method provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The technical solution of the present invention will be further described below in conjunction with the accompanying drawings and through specific embodiments.

[0050] Among them, the accompanying drawings are only for illustrative purposes, showing only schematic diagrams, not physical diagrams, and should not be construed as a limitation on this patent; in order to better illustrate the embodiments of the present invention, some components in the accompanying drawings will be omitted, enlarged or reduced, which do not represent the dimensions of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the accompanying drawings may be omitted.

[0051] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if terms such as "upper", "lower", "left", "right", "inner", "outer", etc. are used to indicate the orientation or positional relationship, it is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the accompanying drawings are only for illustrative purposes and should not be construed as a limitation on this patent. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0052] In the description of the present invention, unless otherwise clearly specified and limited, if terms such as "connection" are used to indicate the connection relationship between components, this term should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components or the interaction relationship between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0053] First, the technical principle of the in-vehicle human-computer interaction method based on big data provided by this application will be described, and then the technical implementation process of this application will be specifically described through 3 embodiments.

[0054] The in-vehicle human-computer interaction method based on big data provided by this application includes the following technical parts:

[0055] 1. Collect the user habits of in-vehicle human-computer interaction to obtain big data on user habits of in-vehicle human-computer interaction, including: user information, scenario information, setting information, and setting interaction preferences;

[0056] User information includes any one or more of: user ID, user age, user gender, clothing index, physiological information, and emotional information; the clothing index is the clothing weather index, which is determined by the thickness of the clothing and is specifically expressed in Table 1 below;

[0057] Table 1 Dressing Meteorological Index and Corresponding Service Terms

[0058] Clothing Weather Index Thickness of Clothing H Service Terms 1 H＞25.0 Suitable for wearing thick down jackets, gloves and other winter clothing for severe cold 2 15.0＜H≤25.0 Suitable for wearing cotton-padded clothes, leather coats, and thick sweaters inside and other winter clothing 3 8.0＜H≤15.0 Suitable for wearing jackets and other early winter clothing 4 4.0＜H≤8.0 Suitable for wearing suits, jackets and other early spring and late autumn clothing 5 1.5＜H≤4.0 Suitable for wearing cotton shirts, long-sleeved T-shirts, denim and other spring and autumn clothing 6 -1.2＜H≤1.5 Suitable for wearing light shirts and long skirts and other summer clothing 7 H≤-1.2 Suitable for wearing short-sleeved T-shirts, short skirts, shorts and other midsummer clothing

[0059] Physiological information includes respiratory rate, heart rate, blood oxygen saturation, etc.

[0060] The described scenario information includes any one or more of the following: outside vehicle temperature, inside vehicle temperature, vehicle speed, remaining battery power, remaining fuel volume, weather conditions (sunny, cloudy, rainy, etc.), light intensity, current time, geographical information of the vehicle (latitude and longitude information where the vehicle is currently located, city information where it is located, etc.), road navigation information (road congestion conditions, etc.), current road type (elevated road, tunnel, community, etc.), driving slope.

[0061] The described setting information includes any one or more of the following: air conditioner setting information (air conditioner temperature, air outlet mode of the air conditioner, air conditioner wind speed, etc.), seat setting information (seat massage, seat heating, seat ventilation, etc.), fragrance setting information (fragrance mixing ratio, fragrance concentration, etc.), window setting information (side window opening degree, sunroof opening degree, etc.), ambient light setting information (dynamic effect of the ambient light, brightness, color, etc.), audio setting information (volume size, reverberation effect, etc.).

[0062] There is a corresponding relationship among user information, scenario information, setting information, and setting interaction preferences in one piece of data. For example, for User A, aged 35 and female, in a scenario where the outside vehicle temperature is 36°C and the light intensity is 500 lx, the preference for a set temperature of 22.5°C, a set air volume of 3, and a set air conditioner interaction mode of blowing on the face and in the internal circulation is "Preference 3", which constitutes one piece of data. In this piece of data, User A, aged 35, and female are user information; the outside vehicle temperature of 36°C and the light intensity of 500 lx are scenario information; the set temperature of 22.5°C, the set air volume of 3, and the set blowing on the face and in the internal circulation are setting information; the set air conditioner interaction preference of "Preference 3" and "3" is the setting interaction preference. Another example is that for User B, when passing by near a chemical plant, the preference for setting the opening degree of all windows to 0% is "Preference 2", which also constitutes one piece of data. Big data is a collection of a large number of such pieces of data.

[0063] The determination methods of setting interaction preferences include the following three methods:

[0064] The first method is: when a certain user actively makes a certain setting for the vehicle in a certain scenario, the value of the setting interaction preference (setting preference degree) corresponding to this setting of the vehicle by this user in this scenario is incremented by "1".

[0065] The second one is: after the system makes a certain setting for a vehicle by a certain user in a certain scenario, if the user does not change the system setting, add "1" to the setting interaction preference value corresponding to this setting of the vehicle by this user in this scenario; if the user changes this system setting, subtract "1" from the setting interaction preference value corresponding to this setting of the vehicle by this user in this scenario.

[0066] The third one is: map the setting interaction preference with a positive value to "1", and map the setting interaction preference with a negative value to "-1". For example, when the user closes the window once when driving onto the highway, and then drives onto the highway continuously twice later, and the system actively sets the window to close twice, the value of its setting interaction preference is 3, which is a positive value. Therefore, map the setting interaction preference with a positive value to "1"; when user D changes the recommended foot-blowing air outlet mode twice at -10°C, the value of its setting interaction preference is -2, which is a negative value. Therefore, map the setting interaction preference with a negative value to "-1".

[0067] 2. Construct a user feature vector and a setting feature vector according to the user information, scenario information, setting information, and setting interaction preference;

[0068] In different embodiments, the expression forms of the user feature vector are different, and the expression forms of the setting feature vector are also different, which will be described in detail in the following specific embodiments.

[0069] 3. Predict the preference degree of the user for the setting through correlation calculation according to the user feature vector and the setting feature vector;

[0070] The method of correlation calculation is: calculate the dot product of the user feature vector and the setting feature vector, or calculate the cosine similarity between the user feature vector and the setting feature vector. The specific calculation process will be described in the following embodiments.

[0071] 4. Set the vehicle according to the predicted preference degree.

[0072] The following specifically illustrates the specific implementation process of the vehicle-mounted human-computer interaction method based on big data provided by this application through 3 embodiments.

[0073] Embodiment 1

[0074] The technical implementation of the vehicle-mounted human-computer interaction method based on big data provided by Embodiment 1 includes the following steps:

[0075] 1. Construct a user feature vector and a setting feature vector according to the user information, scenario information, setting information, and setting interaction preference. The specific method is: the user feature vector is user u i(where \(i\) represents the \(i\)-th user) a vector composed of the preference degrees of \(n\) settings that have been interacted with historically (or the most recently interacted with historically); the obtained setting feature vector is setting \(s\). 1 and each setting \(s\) among the aforementioned \(n\) settings 2 forms a vector of similarities.

[0076] 2. The similarity between settings is determined by the audiences of the two settings. The higher the audience overlap, the more similar the two settings are. The specific calculation process is as follows: Denote the user set of setting \(s\) 1 as \(S\) 1 , and the user set of setting \(s\) 2 as \(S\) 2 , and their user intersection is \(S\) C = \(S\) 1 ∩\(S\) 2 , then the similarity between setting \(s\) 1 and setting \(s\) 2 is:

[0077]

[0078] 3. Obtain the \(n\) settings that the current user has interacted with historically to form a setting feature vector, and then find the \(k\) settings arranged in descending order of similarity for each setting in the setting feature vector associated with the current user, which constitutes a candidate set for vehicle automatic setting for this user;

[0079] For example, if the setting feature vector of user \(u\) i includes \(n\) settings, then after finding the \(k\) settings arranged in descending order of similarity for each setting in the setting feature vector associated with this user, a total of \(n×k\) settings constitute the candidate set for vehicle setting for this user \(u\) i .

[0080] 4. Set the vehicle according to the predicted preference degree, including sorting the preference degrees corresponding to multiple settings and selecting the setting with the highest preference degree for vehicle control. The specific method for obtaining the preference degrees of multiple settings is as follows:

[0081] Obtain the \(n\) settings that a certain user \(u\) i has interacted with most recently;

[0082] For each of the \(n\) settings, find the corresponding \(k\) settings respectively. The similarity calculation method for finding the settings is:

[0083]

[0084] Remove duplicate settings from the candidate set and remove the extended settings in the candidate set that are inconsistent with the current scenario information. Then, calculate the preference degree of the user for each remaining setting in the candidate set through dot product, and use the setting with the highest associated preference degree as the prediction result for how to set the vehicle for the current user in the current scenario.

[0085] The method for predicting the preference degree of the user for the setting through dot product is as follows: Obtain the preference degree of this user for this setting through the known preference degrees of the user for n settings (i.e., the user feature vector) and the similarity between a certain setting and these n settings (i.e., the setting feature vector). For example, the user feature vector Setting feature vector The dot product of the two is

[0086]

[0087] In addition, it should be noted that in Embodiment 1, the setting is preferably an extended setting, and the extended setting is: the fusion information including the setting information and the scenario information for setting the vehicle in the corresponding scenario. That is, the setting does not include the scenario information, while the extended setting includes the scenario information.

[0088] Embodiment 2

[0089] One of the differences between Embodiment 2 and Embodiment 1 lies in the difference between the user feature vector and the setting feature vector. In Embodiment 2, the user feature vector is: the user u i And the vector composed of the similarities with the top p users having user feature similarities with it and arranged in descending order of similarity. For the user u i The obtained setting feature vector is: the vector composed of the preference degrees of p users for each setting s 1 historically interacted with for the same scenario.

[0090] Another difference between Embodiment 2 and Embodiment 1 is that in Embodiment 2, the similarity between two users is determined by the setting preferences of the two users. The higher the preference overlap, the more similar the two users are. The specific calculation process is as follows:

[0091] Denote the set of settings historically interacted with by the user u 1 as K 1 , denote the set of settings historically interacted with by the user u 2 as K 2 , the intersection of their settings is K C =K 1 ∩K 2 , the similarity between the user u 1 and the user u 2 is:

[0092]

[0093] Then, find p users with user feature similarities to the current user, and find n settings where the historical interaction scenario information of each of the p users is consistent with the current scenario information of the vehicle to be set by the current user. A total of n×p settings are used as the candidate set.

[0094] For example, if there are p users with user feature similarities to user u i Then, find n settings where the historical interaction scenario information of each of the p users is consistent with the current scenario information of the vehicle to be set by the current user. A total of n×p settings form the candidate set for automatically setting the vehicle for the current user u i for vehicle automatic setting.

[0095] Then, similar to Embodiment 1, set the vehicle according to the predicted preference degree, including sorting the preference degrees corresponding to the n×p settings, and selecting the setting with the highest preference degree as the current setting for predicting the vehicle control for the current user. The method for obtaining the preference degrees of multiple settings is the same as that in Embodiment 1 and will not be elaborated here.

[0096] In addition, it should be noted that in Embodiment 2, the setting is preferably an extended setting, and the extended setting is: the fusion information including setting information and scenario information for setting the vehicle in the corresponding scenario. That is, the setting does not include scenario information, while the extended setting includes scenario information.

[0097] Embodiment 3

[0098] The vehicle-mounted human-computer interaction method based on big data provided by Embodiment 3 specifically includes the following steps:

[0099] 1. Construct the user feature vector and setting feature vector of the current user according to the user information, scenario information, setting information, and setting interaction preference. The specific method is: separate the discrete information and continuous information of the user information, scenario information, and setting information, and perform preprocessing respectively to obtain the preprocessed vectors. Then send the preprocessed vectors into the neural network to obtain the feature vectors.

[0100] The user discrete information of the user information includes the user ID and / or user gender, and the user continuous information includes the user age and / or clothing index; the scenario discrete information of the scenario information includes the weather condition and / or the geographical information where the vehicle is located, and the scenario continuous information includes the outside temperature and / or the inside temperature of the vehicle; the setting discrete information of the setting information includes the air outlet mode and / or the circulation mode, and the setting continuous information includes the set temperature and / or the set wind speed.

[0101] The discrete and / or continuous user information and / or scenario information are preprocessed and then input into the user neural network to obtain a user feature vector; the discrete and / or continuous setting information and / or scenario information are preprocessed and then input into the setting neural network to obtain a setting feature vector;

[0102] The preprocessing method for user discrete information, scenario discrete information, and setting discrete information is as follows: Multiply the one-hot vector corresponding to the discrete information by the Embedding matrix to obtain the Embedding vector as the preprocessed vector for the discrete information;

[0103] Examples of the one-hot vectors corresponding to the discrete information are as follows:

[0104] Suppose there are 10 users in a simplified large data set. At this time, the one-hot vector is a 10-dimensional vector. For example:

[0105] The one-hot vector corresponding to the user with user ID A is (1, 0, 0, 0, 0, 0, 0, 0, 0, 0) T ,

[0106] The one-hot vector corresponding to the user with user ID B is (0, 1, 0, 0, 0, 0, 0, 0, 0, 0) T ,

[0107] The one-hot vector corresponding to the user with user ID C is (0, 0, 1, 0, 0, 0, 0, 0, 0, 0) T ,

[0108] …

[0109] The one-hot vector corresponding to the user with user ID J is (0, 0, 0, 0, 0, 0, 0, 0, 0, 1) T ,

[0110] An example of the process of multiplying the one-hot vector by the Embedding matrix is as follows:

[0111] The Embedding matrix is an m×n matrix, where n is the same as the dimension of the one-hot vector. m is a quantity less than n. Similarly, in the large data set with 10 users, let m be set to 3. Then the Embedding matrix E is a 3×10 matrix. Calculating the Embedding vector of user C is as follows:

[0112]

[0113] The preprocessing method for user continuous information, scenario continuous information, and setting continuous information is as follows: After binning the continuous information, it is converted into discrete information, and then the one-hot vector corresponding to the discrete information is multiplied by the Embedding matrix to obtain the Embeddin g vector as the discrete information preprocessing vector; or perform normalization processing on the continuous information.

[0114] Taking the user's age as an example, binning means dividing users into groups such as children, youth, middle-aged, and elderly according to their age. After binning, the continuous information is converted into discrete information and processed according to the above discrete information processing method. Normalization processing uses linear transformation to map continuous values to a specified interval. For example, for the set temperature t ∈ [15, 30], through f(t) = (t - 15) / 15, it is mapped to the interval [0, 1].

[0115] The method for the neural network to process the preprocessed information to obtain the user feature vector and the setting feature vector is as follows:

[0116] The parameters in the neural network and the parameters in the Embedding matrix are obtained by an optimization algorithm. The neural network is composed of a combination of multiple basic operation modules. The multiple basic operation modules include linear layers, non-linear activation functions, normalization layers, and attention layers. The parameters of the neural network are the parameters of the multiple operation modules. The parameters of the Embedding matrix are the elements of this matrix. For example, obtaining the parameter W in the neural network and the parameter E in the Embedding matrix is transformed into the following optimization problem:

[0117]

[0118] Among them, p is the set interaction preference information, u is the user feature vector output by the user neural network, s is the setting feature vector output by the setting neural network, and cosine(u, s) is used to calculate the cosine similarity of the two vectors. The specific expression is:

[0119]

[0120] L is the loss function, which can be the mean squared loss, cross-entropy loss, or triplet loss. The above optimization problem can be solved for W and E by various optimization algorithms. Optimization algorithms include stochastic gradient descent, momentum method, AdamW method, and RMSprop method. Briefly, the optimization goal is to make the preference degree cosine(u, s) predicted by the algorithm as close as possible to the set interaction preference p in the big data. For example, there is such a piece of data in the big data: User A, 25 years old, male, in the scenario where the outdoor temperature is 36°C, has a set interaction preference of "1" for the set temperature of 22.5°C, set air volume of 3, set to blow the face and in the internal circulation mode. Then, according to the above user information, scenario information, and setting information, the closer the prediction result obtained through data preprocessing and the neural network is to "1", the better. Through the optimization algorithm, the parameters in the neural network and the parameters in the Embedding matrix are updated to make the prediction result closer and closer to the true value of the set interaction preference.

[0121] The method for the user neural network and the setting neural network to predict vehicle settings for the current user in the current scenario is as follows:

[0122] First, after data processing based on the input preprocessed vector, a user feature vector and a setting feature vector are output. The specific method is as follows:

[0123] The user preprocessed vector and the scenario preprocessed vector are combined as the input vector u of the user neural network i . The output of the user neural network is the user feature vector u o . An example of a user neural network is as follows:

[0124] l 1 =W 1 u i +b 1

[0125] l n =a(W n l n-1 +b n ),n=2,...,N - 1

[0126] u o =a(W N l N-1 +b N )

[0127] Among them, the basic operation module in the form of Wl + b is the linear layer, and W and b are the parameters of the linear layer. a(·) is a non-linear activation function, which is tanh here, and can also be sigmoid, ReLU, or PReLU.

[0128] The set preprocessed vector is used as the input vector s of the set neural network iSetting the output of the neural network means setting the feature vector s o An example of setting a neural network is as follows:

[0129] l 1 = W 1 s i + b 1

[0130] l n = a(W n l n-1 + b n ), n = 2, …, N - 1

[0131] s o = a(W N l N-1 + b N )

[0132] Among them, the basic operation module in the form of Wl + b is the linear layer, and W and b are the parameters of the linear layer. a(·) is a non - linear activation function, which is tanh here, and can also be sigmoid, ReLU, PReLU.

[0133] Combine the setting information elements to obtain all possible setting information as the candidate set. Calculate the preference degree of the current user for each setting in the candidate set through dot - product or cosine similarity. Finally, extract the highest preference degree among the preference degrees of the current user for each setting in the candidate set, and use the setting associated with the highest preference degree as the prediction result for vehicle setting of the current user.

[0134] In summary, the above - mentioned three embodiments provided by the present application for the vehicle - mounted human - machine interaction method based on big data, as Figure 1 shown, include the steps:

[0135] L1. Obtain the vehicle - mounted human - machine interaction habit data of the current user, including user information, scenario information, setting information, and setting interaction preferences, to construct the user feature vector associated with the current user and obtain the setting feature vector;

[0136] L2. Construct a candidate set for predicting vehicle settings for the current user;

[0137] L3. Extract the highest preference degree among the preference degrees of the current user for each setting in the candidate set, and use the setting associated with the highest preference degree as the prediction result for vehicle setting of the current user.

[0138] In summary, the present invention realizes the automatic setting of the vehicle for the current user in the current scenario through different solutions provided by three embodiments. In the three embodiments, candidate sets are constructed by different methods, and the relationships between users, between settings, and between users and settings are mined. The user features are represented by user feature vectors, and the setting features are represented by setting feature vectors. At the same time, the correlation between the above features is used to calculate the preference degree of the user for the setting, and the vehicle is automatically set according to the sorted preference degree of the setting. Compared with the existing independent setting learning for a single user, the human-computer interaction is more intelligent, and the user experience is greatly improved.

[0139] It should be noted that the above specific embodiments are only the preferred embodiments of the present invention and the applied technical principles. Those skilled in the art should understand that various modifications, equivalent replacements, changes, etc. can be made to the present invention. However, as long as these transformations do not deviate from the spirit of the present invention, they should be within the protection scope of the present invention. In addition, some terms used in the specification and claims of this application are not restrictive, but are only for the convenience of description.

Claims

1. A vehicle-mounted human-computer interaction method based on big data, characterized in that: Includes steps: L1, obtaining the in-vehicle human-computer interaction habit data of the current user, including user information, scene information, setting information, and setting interaction preferences, so as to construct a user feature vector associated with the current user and obtain a setting feature vector; L2, constructing a candidate set for predicting vehicle settings for the current user; L3, extracting the highest preference degree of the current user for each setting in the candidate set, and associating the setting with the highest preference degree as a prediction result of vehicle setting for the current user.

2. The vehicle-mounted human-computer interaction method based on big data according to claim 1 is characterized in that: The user information includes: any one or more of user ID, user age, user gender, clothing index, physiological information, and emotional information; The scene information includes: any one or more of the following: vehicle exterior temperature, vehicle interior temperature, vehicle speed, remaining battery power, remaining fuel, weather conditions, light intensity, current time, vehicle location information, road navigation information, current road type, and driving slope; The setting information includes: any one or more of air conditioning setting information, seat setting information, fragrance setting information, window setting information, ambient light setting information, and audio setting information.

3. The vehicle-mounted human-computer interaction method based on big data according to claim 1, characterized in that: The user feature vector is user u i A vector composed of the preference degrees of n settings that have been historically interacted; the obtained setting feature vector is a vector composed of the similarity between setting s1 and each setting s2 in the n settings.

4. The vehicle-mounted human-computer interaction method based on big data according to claim 1 or 3, characterized in that: The method for constructing the candidate set in step L2 is: Obtain n settings that the current user has interacted with historically, and then find k settings that have setting feature similarity with each of the n settings, and a total of n×k settings constitute the candidate set as a basis for automatic vehicle settings for the current user.

5. The vehicle-mounted human-computer interaction method based on big data according to claim 4 is characterized in that: The similarity between setting s1 and setting s2 is: S C =S1∩S2, S c Represents the user intersection of the user set S1 of setting s1 and the user set S2 of setting s2.

6. The vehicle-mounted human-computer interaction method based on big data according to claim 1, characterized in that: The user feature vector is: i A vector consisting of similarities with the first p users who have user feature similarity and are arranged in descending order of similarity; the obtained setting feature vector is: a vector consisting of the preferences of the p users for the setting s1 that they have interacted with for the current scene.

7. The vehicle-mounted human-computer interaction method based on big data according to claim 1 or 6, characterized in that: In step L2, the method for constructing the candidate set is: Find p users with user feature similarity to the current user, and then find n settings whose scene information of historical interactions of each user among the p users is consistent with the current scene information of the vehicle to be set by the current user, and a total of n×p settings are used as the candidate set.

8. The vehicle-mounted human-computer interaction method based on big data according to claim 1 or 7, characterized in that: The similarity between users u1 and u2 is: k C = k1∩k2, k c The setting intersection of the setting set K1 of each setting that the user u1 has historically interacted with and the setting set K2 of each setting that the user u2 has historically interacted with.

9. The vehicle-mounted human-computer interaction method based on big data according to claim 3 or 6, characterized in that: The setting is an extended setting, and the extended setting is: fusion information including the setting information and the scene information for setting the vehicle in a corresponding scene.

10. The vehicle-mounted human-computer interaction method based on big data according to claim 1, characterized in that: Step L3 specifically includes the following steps: L31, removing the settings that are repeated in the candidate set and are inconsistent with the current scene information; L32, predicting the current user's preference for each setting in the candidate set remaining after step L31 is removed by dot product calculation; L33, extracting the highest preference level of the current user for each setting in the candidate set remaining after removing step L31, and associating the setting with the highest preference level as the prediction result of vehicle setting for the current user.

11. The vehicle-mounted human-computer interaction method based on big data according to claim 1, characterized in that: In step L1, the vehicle-mounted human-computer interaction habit data is distinguished from discrete information and continuous information and preprocessed, and the preprocessed vector is input into a neural network to output the user feature vector and the setting feature vector; The user discrete information of the user information includes user ID and / or user gender, and the user continuous information includes user age and / or clothing index; the scene discrete information of the scene information includes weather conditions and / or geographical information of the vehicle, and the scene continuous information includes the temperature outside the vehicle and / or the temperature inside the vehicle; the setting discrete information of the setting information includes air outlet mode and / or circulation mode, and the setting continuous information includes set temperature and / or set wind speed.

12. The vehicle-mounted human-computer interaction method based on big data according to claim 11, characterized in that: The discrete and / or continuous user information and / or the scene information are pre-processed and input into the user neural network to obtain the user feature vector; the discrete and / or continuous setting information and / or the scene information are pre-processed and input into the setting neural network to obtain the setting feature vector; The preprocessing method for user discrete information, scene discrete information and setting discrete information is as follows: converting the one-hot vector corresponding to the discrete information into an Embedding matrix, and obtaining the Embedding vector as the discrete information preprocessing vector; The preprocessing method for user continuous information, scene continuous information and setting continuous information is: convert the continuous information into discrete information after bucketing, and then multiply the one-hot vector corresponding to the discrete information by the Embedding matrix to obtain the Embedding vector as the discrete information preprocessing vector; or normalize the continuous information.

13. The vehicle-mounted human-computer interaction method based on big data according to claim 12, characterized in that: The bucketing process is to group the continuous information, and the normalization process is to map the continuous values ​​of the continuous information to a specified numerical range using linear changes.

14. The vehicle-mounted human-computer interaction method based on big data according to claim 12 or 13, characterized in that: The method for the neural network to predict the current user's settings for the vehicle in the current scenario is: The cosine similarity between the user feature vector and the setting feature vector is calculated as the preference of the current user for the setting in the current scenario, and the setting with the highest preference is used for controlling the vehicle.

15. The vehicle-mounted human-computer interaction method based on big data according to claim 1, characterized in that: The user's preference for a setting is calculated as: After the user sets the vehicle in a specified scenario, the value of the setting preference corresponding to the setting of the vehicle by the user in the specified scenario is accumulated by "1"; After the system performs the settings for the vehicle in the specified scenario for the user, if the system detects that the user enters the specified scenario again and the user does not change the settings made by the system after entering the specified scenario this time, the user's preference for the settings of the vehicle in the specified scenario is accumulated by "1"; if the user changes the settings made by the system, the user's preference for the settings in the specified scenario is accumulated by "1"; Or map positive preferences to "1" and negative preferences to "-1".

16. The vehicle-mounted human-computer interaction method based on big data according to any one of claims 11 to 14, characterized in that: Step L3 specifically includes the following steps: L301, predicting the current user's preference for each setting in the candidate set by dot product or cosine similarity calculation; L302, extracting the highest preference level of the current user for each setting in the candidate set, and associating the setting with the highest preference level as a prediction result of vehicle settings for the current user.

17. A vehicle-mounted human-computer interaction system based on big data, characterized in that: The vehicle-mounted human-computer interaction method based on big data as described in any one of claims 1 to 16 can be implemented.

Images