Remote control method, device, equipment and storage medium for cooking equipment

Abstract

The present invention relates to the technical field of remote control of equipment, and discloses a remote control method, device, equipment and storage medium for cooking equipment, wherein the method comprises: collecting an operation instruction sequence executed by a user on a mobile terminal, and performing time feature extraction on the operation instruction sequence to obtain user operation feature data; inputting a cooking operation sequence Transformer model with time-space attention to perform feature matching and identity authentication to obtain a user identity authentication result; obtaining a standardized equipment status data set for cooking equipment; performing data security verification to obtain equipment operation status data that passes security verification; generating a standard format control instruction sequence; performing communication grouping and transmission execution on the standard format control instruction sequence to obtain a corresponding group transmission execution result, wherein the method improves the reliability and stability of remote control of cooking equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of remote control of equipment, and in particular to a remote control method, device, equipment and storage medium for cooking equipment. Background Art

[0002] Traditional remote control methods mainly rely on password authentication or biometric recognition, which are vulnerable to security threats such as brute force attacks and phishing. In addition, biometrics may change due to factors such as environment and age, affecting the reliability of recognition.

[0003] In addition, during the remote control of cooking equipment, the device status data collected from sensors is vulnerable to false data injection attacks. Attackers may tamper with sensor data, causing the control system to make wrong judgments, affecting the safety and reliability of the cooking process. The authenticity and integrity of device status data are difficult to be effectively guaranteed. At the same time, when multiple cooking devices are connected to the remote control system at the same time, the limited communication resources and denial of service attacks will cause communication delays and packet loss. The existing communication mechanism lacks effective resource allocation and dynamic adjustment capabilities, making it difficult to ensure the timely and reliable transmission of remote control commands. Summary of the invention

[0004] The present invention provides a remote control method, device, equipment and storage medium for cooking equipment, which are used to improve the reliability and stability of remote control of cooking equipment.

[0005] In a first aspect, the present invention provides a remote control method for a cooking device, the remote control method for a cooking device comprising:

[0006] Collecting the operation instruction sequence executed by the user on the mobile terminal, and extracting the time feature of the operation instruction sequence to obtain user operation feature data;

[0007] Inputting the user operation feature data into a cooking operation sequence Transformer model with time-space attention for feature matching and identity authentication to obtain a user identity authentication result;

[0008] Obtain sensor data collected by cooking equipment, filter outliers and normalize the data to obtain a standardized equipment status data set;

[0009] Inputting the standardized device status data set into a conditional generative adversarial network for data security verification to obtain device operation status data that passes security verification;

[0010] Generate a control instruction sequence in a standard format based on the user identity authentication result and the device operation status data that has passed the security verification;

[0011] The standard format control instruction sequence is grouped and transmitted to obtain a corresponding group transmission execution result.

[0012] In a second aspect, the present invention provides a remote control device for a cooking device, the remote control device for a cooking device comprising:

[0013] A collection module is used to collect the operation instruction sequence executed by the user on the mobile terminal, and extract the time feature of the operation instruction sequence to obtain user operation feature data;

[0014] An identity authentication module, used for inputting the user operation feature data into a cooking operation sequence Transformer model with time-space attention for feature matching and identity authentication, and obtaining a user identity authentication result;

[0015] A processing module is used to obtain sensor data collected by cooking equipment, filter outliers and normalize the data to obtain a standardized equipment status data set;

[0016] A security verification module, used to input the standardized device status data set into a conditional generative adversarial network for data security verification, and obtain device operation status data that passes the security verification;

[0017] A generation module, used to generate a control instruction sequence in a standard format based on the user identity authentication result and the device operation status data that has passed the security verification;

[0018] The execution module is used to perform communication grouping and transmission execution on the standard format control instruction sequence to obtain corresponding grouping transmission execution results.

[0019] A third aspect of the present invention provides a remote control device for a cooking device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the remote control device of the cooking device executes the above-mentioned remote control method of the cooking device.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned remote control method for a cooking device.

[0021] In the technical solution provided by the present invention, by introducing a user identity authentication mechanism based on operation behavior patterns and combining it with a Transformer model with enhanced spatiotemporal attention, the user's unique operation habit characteristics are effectively extracted, the risk of identity feature leakage is reduced, and the security of remote control is enhanced; an improved conditional generative adversarial network is used to perform security verification on device status data, and through adversarial learning of the generator and the discriminator, abnormal data is accurately identified and repaired, thereby ensuring the authenticity and validity of sensor data; a dynamic time allocation communication mechanism is designed, and control instructions are transmitted in groups based on communication resource occupancy characteristics, and are dynamically adjusted through an average residence time automaton and a time ratio monitor, thereby effectively avoiding communication congestion and instruction loss; through a multi-level data processing and security verification mechanism, full-process security protection is achieved from user authentication, data collection to instruction transmission, and the reliability and stability of remote control of cooking equipment are significantly improved.

[0022] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of an embodiment of a remote control method for a cooking device according to an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of an embodiment of a remote control device for a cooking device according to an embodiment of the present invention;

[0026] Figure 3 Schematic diagram of an embodiment of a remote control device for a cooking device in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0029] To facilitate understanding of this embodiment, a remote control method for a cooking device disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps:

[0030] 101. Collecting an operation instruction sequence executed by a user on a mobile terminal, and extracting time features of the operation instruction sequence to obtain user operation feature data;

[0031] It is understandable that the execution subject of the present invention may be a remote control device of a cooking device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0032] Specifically, the operation instruction sequence executed by the user on the mobile terminal is collected, and the user's operation data, including operation time interval data, operation duration data, operation force data and operation trajectory data, are recorded through hardware sensors (such as touch screen pressure sensors, accelerometers and gyroscopes). These data represent the behavioral details of the user's interaction with the mobile terminal. The operation time interval data and operation duration data are analyzed, and the timestamp features of the operation are extracted using the time series analysis method. The timestamp feature reflects the distribution characteristics of the user's operation behavior in the time dimension by marking the regularity of the occurrence time, end time and time interval of the operation. At the same time, the operation force data and operation trajectory data are processed by multidimensional data analysis technology, for example, the data of different scales are unified into the same range through normalization operation, and then the type identification feature of the operation is calculated using the pattern recognition algorithm. The type identification feature can characterize the category of the user's operation, such as click, slide or long press, by defining the mechanical characteristics and trajectory shapes of different operations. On the basis of the above, the timestamp feature of the operation is combined with the type identification feature to construct a time series feature containing timestamp, type identification and operation parameters. Through the integration of multidimensional feature vectors, each operation behavior of the user is recorded as structured time series data. Compare and analyze the time series features with the user's historical usage habit data, and identify the user's operation habit features by calculating the similarity between the features. For example, by comparing, dig out information such as the user's commonly used operation mode, usage frequency, and operation style. In order to enhance the depth of the analysis, associate the user's operation habit features with their historical cooking experience data. Through statistical learning or machine learning methods, combined with the user's past cooking behavior data, such as dish type, temperature setting habits, etc., dig out the association between operation habits and cooking needs to form the user's cooking operation behavior features. Segment the user's cooking operation behavior features according to the preset time window, and divide the entire behavior data into multiple feature data segments. These data segments describe the user's operation mode and cooking behavior characteristics within a specific time range in a more fine-grained manner. Perform spatiotemporal feature fusion on multiple feature data segments, and integrate the information of the time dimension and the space dimension through the fusion algorithm to generate comprehensive user operation feature data.

[0033] 102. Input the user operation feature data into the cooking operation sequence Transformer model with time and space attention for feature matching and identity authentication, and obtain a user identity authentication result;

[0034] Specifically, the user operation feature data is input into the operation sequence embedding layer in the Transformer model. The embedding layer maps the input data through a linear projection matrix with a dimension of 512 to transform the original features into a higher-dimensional representation. At the same time, combined with the predefined position encoding information, the order of the time series features is embedded into the vector to form an initial feature embedding vector. The model calculates the query matrix, key matrix and value matrix corresponding to the initial feature embedding vector through an 8-head parallel attention mechanism, and the dimension of each matrix is ​​64. These matrices share different attention calculation tasks through a multi-head mechanism, where the query matrix and the key matrix generate an attention score matrix in the time dimension through matrix multiplication. This matrix represents the correlation between different time points in the user operation features and can capture the time dependency of the operation behavior. The attention score matrix is ​​weighted and summed so that the features at different times can be dynamically adjusted according to their importance. In order to improve the stability and generalization ability of the features, the calculation results are normalized by Layer Normalization, and the original input information is retained through the residual connection mechanism to generate a standardized temporal attention feature. In order to capture the correlation in the spatial dimension, the normalized temporal attention features are input into the spatial attention layer. In this layer, the features are enhanced by superimposing the spatial position encoding matrix so that the model can perceive the relative position relationship of the operation features in space. At the same time, the adaptive weight mechanism is used to calculate the spatial attention map, so as to dynamically assign the weights of different features in space and generate spatial dimension feature representations. These spatial dimension feature representations are processed by nonlinear transformation, and deep semantic features are extracted using activation functions such as ReLU or GELU to enhance the model's semantic understanding of user operation behavior. The deep semantic features are input into the spatiotemporal feature fusion layer. In the fusion layer, the model comprehensively processes the features of the temporal and spatial dimensions through an adaptive weighting mechanism to ensure that the correlation of the two features is fully expressed, and the spatiotemporal fusion features of the user's operation are obtained. The user's spatiotemporal fusion features are compared one by one with at least three historical operation template features stored in the user's historical operation feature library, and the similarity between the user's current features and the template features is quantified by calculating the cosine similarity, and the template matching degree with the largest similarity is selected as the final feature matching score. Based on the feature matching score, the system determines the degree of conformity between the user's current operation behavior and the historical operation mode, and generates the user identity authentication result. If the matching score is higher than the preset authentication threshold, the system determines that the user identity is valid, otherwise it triggers additional verification measures.

[0035] 103. Obtain sensor data collected by the cooking equipment, perform outlier filtering and data normalization processing, and obtain a standardized equipment status data set;

[0036] Specifically, the raw data collected by the temperature sensor, humidity sensor, firepower sensor and weight sensor installed in the cooking device are parsed and processed, and the specific content of the data is parsed through the communication protocol of the device. The parsed result includes the timestamp of data collection, the sensor identification and the corresponding sensor value, forming the basic structure of the raw data of the device status. The raw data of the device status are classified and sorted according to the sensor identification in the parsed result, and the data are divided into a temperature data sequence, a humidity data sequence, a firepower data sequence and a weight data sequence. The sliding window algorithm is applied to the temperature data sequence to calculate the sliding mean and standard deviation, and the normal fluctuation range of the data is evaluated by setting a predefined first standard deviation range. For temperature data points beyond this range, it is considered that there are abnormalities, and the results retained after filtering these data are used as the temperature abnormal value filtering results. Similarly, the same processing steps are performed on the humidity data sequence, the sliding mean and standard deviation of the humidity are calculated, and the abnormal values ​​are eliminated based on the preset second standard deviation range as the criterion to obtain the humidity abnormal value filtering results. When processing the firepower data sequence and the weight data sequence, the joint distribution analysis method is adopted. By constructing a firepower-weight joint distribution model, the correlation between the firepower data and the weight data is taken into consideration. The joint distribution model can describe the distribution characteristics of the combination of firepower and weight under normal circumstances, thereby identifying abnormal data points that deviate from the main distribution area. The abnormal combination data is filtered, and the retained data constitutes the firepower-weight outlier filtering results. This process effectively eliminates outliers in the sensor data that deviate from the mainstream trend due to faults or external interference. The temperature outlier filtering results, humidity outlier filtering results, and firepower-weight outlier filtering results are normalized to eliminate the differences between different data dimensions and dimensions, using methods such as Min-Max normalization or Z-Score standardization. By mapping the values ​​of each data sequence to a unified range, such as [0,1], the data of different sensors have a consistent scale, which is convenient for subsequent model input or analysis processing. The normalized temperature, humidity, and firepower-weight filtering results are summarized to form a standardized equipment status data set.

[0037] 104. Input the standardized equipment status data set into the conditional generative adversarial network for data security verification, and obtain the equipment operation status data that has passed the security verification;

[0038] Specifically, the standardized device state dataset is input into the generator of the conditional generative adversarial network, which consists of a 5-layer encoder and a 5-layer decoder. In the encoding stage, each layer of the encoder uses a 3×3 convolution kernel with a stride of 2 and a LeakyReLU activation function to extract the deep features of the input data. Through layer-by-layer dimensionality reduction, the generator can effectively compress the complex features of the device state data in high dimensions. In the decoding stage, each layer of the decoder uses a 3×3 deconvolution kernel with a stride of 2 and a ReLU activation function to gradually restore the compressed features to the resolution of the original data and generate generated data samples with similar feature distribution. The generated data samples are input into the discriminator of the conditional generative adversarial network to evaluate the authenticity of the data. The discriminator consists of 4 convolution layers, each of which uses a 4×4 convolution kernel with a stride of 2 and a LeakyReLU activation function to extract the discriminative features of the input data from different levels. The discriminator compares the distribution of the generated data samples with the standardized device state data and outputs a discriminative probability distribution to indicate the degree of similarity between the generated data samples and the real data. Based on the discriminant probability distribution, the authenticity score of the generated data sample is calculated. The higher the authenticity score, the higher the match between the sample and the real data distribution. Conversely, it indicates that the sample contains abnormal or forged data. For samples with low scores, they are marked as abnormal data, and the corresponding abnormal data marking results are generated. Based on the marking results, feature analysis and data interpolation are performed to repair potential abnormal data. The feature analysis stage mines the potential distribution pattern of abnormal data through a deep learning model, while the data interpolation stage combines the statistical characteristics of normal data to generate reasonable replacement values, thereby effectively repairing abnormal data. After the repair is completed, the abnormal data repair result is input into the reconstruction module of the generator, which consists of 3 residual blocks, each of which contains two 3×3 convolutional layers and a jump connection. Through the design of the residual block, the reconstruction module can further optimize the repair result without losing the original data features and generate more accurate reconstructed data. Data security verification is performed based on the data reconstruction results. By comparing the matching degree between the reconstructed data and the standardized equipment status data, it is confirmed whether the data has passed the security verification. Through this process, the equipment operation status data that has passed the security verification is generated.

[0039] 105. Generate a standard format control instruction sequence based on the user identity authentication result and the equipment operation status data that has passed the security verification;

[0040] Specifically, according to the equipment operation status data that has passed the safety verification, the various operation parameters of the equipment are classified and analyzed, and the equipment operation parameter restrictions including temperature range, humidity range and firepower range are extracted. These parameter restrictions represent the working boundaries of the equipment under specific conditions. By analyzing the equipment status, it is ensured that the user's control request will not exceed the safe operation range of the equipment, thereby effectively avoiding the risk of operation errors or equipment damage. The user's control request is verified for legitimacy based on the equipment operation parameter restrictions. By comparing the control parameters proposed by the user with the equipment's restrictions, it is ensured that the control parameters are in line with the actual operation capabilities of the equipment in terms of temperature, humidity, firepower, etc. If the control request exceeds the restriction range, corresponding adjustments or rejections are made to obtain a control parameter set that meets the equipment operation constraints. The control parameter set that meets the restriction conditions is classified and subdivided according to instruction types such as temperature adjustment, firepower adjustment, timing control and mode switching. Through this classification, different types of control requirements are processed separately to ensure that the execution of each type of instruction can adjust the relevant parameters of the equipment in a targeted manner. For each type of classified control instruction, an instruction that conforms to the standard format is generated. The standard format instruction has a unified structure and parameter form, which can ensure compatibility between different devices and facilitate subsequent execution and management. The generated standard format instructions are time-planned according to the preset equipment operation constraint rules. According to the performance, load and other operating conditions of the equipment, the execution order of the instructions is reasonably arranged to obtain the initial instruction execution sequence. The purpose of timing planning is to ensure that the execution of each instruction does not interfere with each other, and the equipment can perform all tasks in the best state, improve efficiency and reduce errors. Conflict detection and conflict elimination are performed on adjacent instructions in the initial instruction execution sequence. By analyzing the relationship between instructions, potential conflicts are identified and adjusted, such as rearranging the execution order of instructions or optimizing the parameters of instructions, to ensure that the final generated instruction sequence is conflict-free. The instruction sequence after conflict elimination is grouped to form a control instruction data packet sequence. The control instruction data packet sequence is verified for timing integrity to ensure that the order and time interval of the instructions meet the operating requirements of the equipment during the data packet transmission process, thereby avoiding control failures caused by data packet confusion or loss, and finally obtaining a standard format control instruction sequence that meets the requirements.

[0041] 106. Perform communication grouping and transmission execution on the standard format control instruction sequence to obtain corresponding grouping transmission execution results.

[0042] Specifically, the control instruction sequence in the standard format is divided into multiple communication control instruction subgroups according to the instruction data transmission frequency, data packet size and communication priority. Each subgroup corresponds to a type of transmission task, and the number of instructions, data size and priority contained in it will affect the scheduling strategy of the subgroup in the entire communication process. Time segments are allocated to multiple communication control instruction subgroups. Each subgroup needs to occupy a certain time period for data transmission, and the length of the time period needs to be reasonably allocated according to the priority of the subgroup and the size of the data packet. A basic communication time allocation scheme is formulated to preliminarily determine the time segment of each subgroup. In order to improve the utilization efficiency of the communication channel and reduce the transmission delay, the time segments are staggered to form a staggered communication scheduling sequence. The staggered arrangement can effectively avoid resource conflicts between different subgroups and ensure that the communication process is smoother and more efficient when multiple subgroups are transmitting at the same time. Based on the staggered communication scheduling sequence, the actual transmission of the instruction is executed. During the transmission process, the average residence time automaton is used to record the residence time of the data packet of each communication control instruction subgroup in the communication channel. The residence time refers to the length of time each data packet is transmitted in the communication channel. This data helps analyze the occupancy of communication resources and provides a basis for subsequent scheduling optimization. By counting the residence time of all data packets, the data on the occupancy of communication resources is obtained, and the average transmission delay of each communication control instruction subgroup is calculated. The delay calculation can help evaluate the current transmission efficiency and promptly discover subgroups with abnormal delays. The calculation results are used to determine whether there is abnormal transmission. If the transmission delay of some subgroups exceeds the preset threshold, it is marked as transmission abnormality. At this time, the communication control instruction subgroup marked as abnormal is processed and the time segment is re-allocated. Through the time ratio monitor, the time segment allocation ratio of each subgroup is dynamically adjusted to avoid excessive concentration of resources in a certain subgroup, ensuring that the transmission efficiency of all subgroups can be reasonably guaranteed. The dynamically adjusted time allocation scheme can maximize the overall efficiency of data transmission and eliminate the negative effects of transmission delays or conflicts. The instruction transmission order is reorganized based on the dynamically adjusted time allocation scheme, and new packet transmission is performed based on the updated time segment. Through the optimized time allocation, the transmission of the control instruction sequence can be more stable and efficient, and the instruction transmission status record is obtained, recording the transmission of each instruction, including transmission time, transmission status, and abnormal conditions. The instruction transmission status records are sorted in timestamp order to obtain the group transmission execution results.

[0043] In the embodiment of the present invention, by introducing a user identity authentication mechanism based on operation behavior patterns and combining it with a Transformer model with enhanced spatiotemporal attention, the user's unique operation habit characteristics are effectively extracted, the risk of identity feature leakage is reduced, and the security of remote control is enhanced; an improved conditional generative adversarial network is used to perform security verification on device status data, and through adversarial learning of the generator and the discriminator, abnormal data is accurately identified and repaired, thereby ensuring the authenticity and validity of sensor data; a dynamic time allocation communication mechanism is designed, and control instructions are transmitted in groups based on communication resource occupancy characteristics, and are dynamically adjusted through an average residence time automaton and a time ratio monitor, thereby effectively avoiding communication congestion and instruction loss; through a multi-level data processing and security verification mechanism, full-process security protection is achieved from user authentication, data collection to instruction transmission, significantly improving the reliability and stability of remote control of cooking equipment.

[0044] In a specific embodiment, the process of executing step 101 may specifically include the following steps:

[0045] Collecting the operation command sequence executed by the user on the mobile terminal, and processing the operation command sequence to obtain operation time interval data, operation duration data, operation force data and operation trajectory data;

[0046] Analyze the operation time interval data and the operation duration data to obtain the timestamp features of the operation, and perform feature calculation on the operation force data and the operation trajectory data to obtain the type identification features of the operation;

[0047] Based on the timestamp feature of the operation and the type identification feature of the operation, a time series feature containing the timestamp, type identification and operation parameters is constructed;

[0048] Compare and analyze the time series features with the user's historical usage habit data to obtain the user's operation habit features, and perform correlation analysis on the operation habit features and the user's historical cooking experience data to obtain the user's cooking operation behavior features;

[0049] The user's cooking operation behavior characteristics are segmented according to a preset time window to obtain a plurality of characteristic data segments, and the plurality of characteristic data segments are fused with time and space characteristics to obtain user operation characteristic data.

[0050] Specifically, the mobile terminal device records various data when the user performs an operation through built-in sensors (such as accelerometers, touch screen sensors, pressure sensors, etc.). These sensors can accurately capture every operation of the user, such as touching, sliding, clicking, and other actions, and record them in the form of a timestamp. Through these sensors, operation time interval data (i.e., the time difference between each operation), operation duration data (i.e., the execution time of each operation), operation force data (i.e., the force applied by the user during the operation, obtained through the pressure sensor), and operation trajectory data (i.e., the user's finger trajectory on the mobile terminal, obtained through the coordinate changes of the touch screen) are obtained. The operation time interval data and operation duration data are analyzed to obtain the timestamp characteristics of the operation. The operation time interval data represents the time interval between two consecutive operations of the user, recorded as ,in Indicates The timestamp of the operation. The operation time interval is expressed as: ,in For the The timestamp of the operation, and the operation duration data refers to the duration of each user operation, calculated by the start and end time of the touch. The start time is , and the end time is , then the operation duration for:

[0051] ;

[0052] When analyzing this data, we can identify the time patterns between different operations, and then reveal the rhythm characteristics of the user's operation, for example, whether the user tends to operate quickly or prefers to operate for a longer time. We calculate the characteristics of the operation force data and operation trajectory data. In terms of operation force, the pressure value of the touch is recorded by the pressure sensor, and the force characteristics are expressed as , which is directly related to the pressure applied during operation, through the formula:

[0053] ;

[0054] in It is The operation trajectory data records the user's movement path through the coordinate changes of the touch screen, and is obtained by calculating the distance and curvature of the touch trajectory. For example, the operation trajectory is a series of coordinate points. To represent the total length of the trajectory Calculated as:

[0055] ;

[0056] Track length Indicates the total distance the user moves during the operation. Based on the timestamp feature and type identification feature of the operation, a time series feature containing timestamp, type identification and operation parameters is constructed. The timestamp feature indicates the time point when the user operation occurs. These time points can help the system understand the timing rules of the operation. The type identification feature is classified and identified according to the characteristics of each operation. For example, if the user keeps the touch for a long time and the pressure is high when performing an operation, the operation is marked as a "hard press operation". Through comprehensive analysis of timestamps, type identification and operation parameters (such as force, trajectory length, etc.), time series features are constructed. , that is, the comprehensive characteristics of each operation on the timeline. Compare and analyze the time series characteristics with the user's historical usage habit data to obtain the user's operation habit characteristics. The user's historical usage habit data includes information such as the user's previous operation mode, frequency, and duration in similar situations. Through comparative analysis, the user's personalized operation habits are mined. For example, if the user frequently performs a certain type of operation in a specific time period, or prefers to use a certain operation method in certain specific environments, extract the characteristics of these operation behaviors and compare them with the current operation to identify the user's operation habit characteristics. The comparative analysis calculates the similarity between the user's historical operation data and the current operation, for example, using methods such as cosine similarity or Euclidean distance to obtain:

[0057] ;

[0058] in, It is the user's historical operation data. is the time series feature of the current operation. The Similarity() function calculates the similarity between the current operation and historical data. The operation habit features and the user's historical cooking experience data are analyzed in association to obtain the user's cooking operation behavior features. The user's cooking experience data includes information such as the user's preferences, habits, and commonly used operation sequences in the past cooking process. Through association analysis, the user's frequently used equipment settings, time arrangements, and operation processes are revealed. The user's cooking operation behavior features are obtained through collaborative filtering, association rule mining, and other methods. It is expressed as:

[0059] ;

[0060] in, is the user's historical cooking experience data. The Correlation() function is used to calculate the correlation between the user's operating habits and his historical cooking experience. The user's cooking operation behavior characteristics are segmented according to the preset time window to obtain multiple feature data segments. The size of the time window is set according to the specific application scenario. Assume that the time window is , the user's operation behavior characteristics are segmented according to each window time period to obtain multiple feature data segments , each data segment contains the operation features within a period of time. In each data segment, the time and space features are combined through the spatiotemporal feature fusion method to obtain a complete user operation feature data set. Contains all spatiotemporal features to form a complete description of the operational behavior.

[0061] In a specific embodiment, the process of executing step 102 may specifically include the following steps:

[0062] The user operation feature data is input into the operation sequence embedding layer of the cooking operation sequence Transformer model with time-space attention. The operation sequence embedding layer uses a linear projection matrix with a dimension of 512 to transform and map the user operation feature data, and superimposes the position encoding information to obtain the initial feature embedding vector.

[0063] The query matrix, key matrix, and value matrix of dimension 64 corresponding to the initial feature embedding vector are calculated through an 8-head parallel attention mechanism, and the query matrix and the key matrix are matrix multiplied to obtain the attention score matrix of the time dimension.

[0064] The attention score matrix based on the time dimension performs weighted summation on the features at different times, and obtains the normalized temporal attention features through LayerNormalization and residual connection;

[0065] The normalized temporal attention features are input into the spatial attention layer, the spatial dependency of the features is enhanced through the spatial position encoding matrix, and the spatial attention map is calculated using an adaptive weight mechanism to obtain the spatial dimension feature representation;

[0066] Perform nonlinear transformation on the spatial dimension feature representation to obtain deep semantic feature representation, and input the deep semantic feature representation into the spatiotemporal feature fusion layer for adaptive weighted fusion to obtain the spatiotemporal fusion features of user operations;

[0067] The cosine similarity is calculated for the spatiotemporal fusion features of the user operation and at least three historical operation template features stored in the user historical operation feature library, the maximum similarity value is selected as the feature matching score, and the user identity authentication result is generated according to the feature matching score.

[0068] Specifically, the user operation feature data is input into the operation sequence embedding layer of the cooking operation sequence Transformer model with a temporal and spatial attention mechanism. In this layer, a linear projection matrix with a dimension of 512 is used to map the user operation feature data. Assume that the user operation feature data consists of multiple time steps. The corresponding eigenvectors are composed of . Perform a linear projection transformation on each operational eigenvector:

[0069] ;

[0070] in, is the linear projection matrix, is the bias term, the result after projection As an embedding vector. After this transformation, the position encoding information is added so that the model can distinguish the relative positions between each time step in the time series. The position encoding uses a predefined method to The location of the , and added to the embedding vector:

[0071] ;

[0072] Get the initial feature embedding vector containing position encoding information . The initial feature embedding vector is input into the attention mechanism layer in the Transtormer model. Through the 8-head parallel attention mechanism, the query matrix, key matrix and value matrix are calculated respectively. Each attention head has its own independent query matrix , key matrix Sum Matrix ,in is the index of the attention head. The query matrix, key matrix and value matrix are obtained by the following linear transformation:

[0073] ;

[0074] in, , and The weight matrices of the query, key, and value of each attention head are calculated by performing matrix multiplication of the query matrix and the key matrix to obtain the attention score matrix :

[0075] ;

[0076] in, is the dimension of the query matrix and the key matrix, which is equal to 64. By performing a softmax operation on the attention score matrix, the normalized attention weight is obtained , which is then compared with the value matrix Multiply them together to get the weighted features of each attention head:

[0077] ;

[0078] Merge the outputs of the 8 attention heads to get the final temporal attention features:

[0079] ;

[0080] And pass it to the Layer Normalization layer for normalization and residual connection to obtain the normalized temporal attention feature The normalized temporal attention features are input into the spatial attention layer. In this layer, the spatial position encoding matrix Enhance the spatial dependency of features. Spatial position encoding is used to capture the spatial features of user operations, such as the user's touch position and sliding trajectory on the screen. The role of spatial position encoding is to help the system identify the impact of user operations at different positions on the cooking process and provide more contextual information in the spatial dimension. Through the adaptive weight mechanism, the spatial attention map is calculated. :

[0081] ;

[0082] Multiply the spatial attention map with the normalized temporal features to obtain the feature representation of the spatial dimension . A nonlinear transformation of the feature representation of the spatial dimension is performed through a layer of activation function (such as ReLU):

[0083] ;

[0084] The features after nonlinear transformation are input into the spatiotemporal feature fusion layer for adaptive weighted fusion. This layer combines the temporal and spatial features through weighted averaging to obtain the spatiotemporal fusion features of user operations. In this fusion process, different weights are assigned according to the importance of the features. Give time and space dimensions:

[0085] ;

[0086] The weighted features can fully integrate time and space information. In order to authenticate the user identity, the spatiotemporal fusion features of the user operation are compared with at least three historical operation template features in the user's historical operation feature library. Each historical operation template feature (in ) represent a typical user operation behavior pattern. By calculating the cosine similarity , evaluate the similarity of user operations. The cosine similarity formula is as follows:

[0087] ;

[0088] in, and They are user operation features and historical template features. and is their norm. By calculating the similarity of the three template features, the one with the largest similarity is selected as the final match and the user identity authentication result is generated. The final authentication result determines whether the user is a legitimate user based on the similarity threshold. If the maximum similarity value exceeds a preset threshold, the authentication passes, otherwise the authentication fails.

[0089] In a specific embodiment, the process of executing step 103 may specifically include the following steps:

[0090] Parse the raw data collected by the temperature sensor, humidity sensor, fire power sensor and weight sensor in the cooking equipment to obtain the raw data of the equipment status including the data collection timestamp, sensor identification and sensor value;

[0091] Classify and sort the original data of equipment status according to the sensor identification to obtain temperature data sequence, humidity data sequence, firepower data sequence and weight data sequence;

[0092] Calculate the temperature sliding mean and standard deviation based on the temperature data sequence, and filter the data in the temperature data sequence that exceeds the preset first standard deviation range to obtain the temperature outlier filtering result;

[0093] Calculate the humidity sliding mean and standard deviation based on the humidity data sequence, and filter the data in the humidity data sequence that exceeds the preset second standard deviation range to obtain the humidity outlier filtering result;

[0094] The firepower-weight joint distribution is calculated based on the firepower data sequence and the weight data sequence, and the abnormal combination data that deviates from the main distribution area is filtered to obtain the firepower-weight abnormal value filtering result;

[0095] The temperature outlier filtering results, humidity outlier filtering results, and firepower-weight outlier filtering results are normalized to obtain a standardized equipment status data set.

[0096] Specifically, each sensor in the cooking device (such as temperature sensor, humidity sensor, fire power sensor and weight sensor) collects data regularly. The raw data of the sensor includes a timestamp, a sensor identifier and a sensor value. The data is classified and sorted according to the sensor identifier to generate a temperature data sequence, a humidity data sequence, a fire power data sequence and a weight data sequence. For example, the temperature data sequence is expressed as:

[0097] ;

[0098] Similarly, store humidity data series, firepower data series, and weight data series separately. Calculate the sliding mean and standard deviation of each type of sensor data to capture the trend and fluctuation of the data, and then identify and filter out abnormal values ​​that deviate from the normal range. For the temperature data series, calculate the sliding mean of the temperature data and sliding standard deviation . Assume that a window size of The sliding window in At each time point, the sliding mean and standard deviation of the temperature are:

[0099] ;

[0100] According to the preset first standard deviation range Determine whether there is an abnormal value. If a temperature value If the value exceeds this range, it is considered an outlier and filtered. Beyond , then it is considered an outlier:

[0101] ;

[0102] The processing of humidity data is similar to that of temperature data. Calculate the sliding mean of humidity data and standard deviation , and use the corresponding standard deviation range to filter out outliers:

[0103] ;

[0104] Use a similar standard deviation range Filter outliers in humidity data. For each value in the humidity data series, if it is outside the range, it is considered an outlier:

[0105] ;

[0106] For firepower data and weight data, there is a certain relationship between these two data, especially there is a linear or nonlinear joint distribution between firepower and weight. Such abnormal combinations can be identified by calculating the firepower-weight joint distribution. Assume that the firepower data sequence is , the weight data sequence is , calculate their joint distribution, and judge abnormal data according to the degree of deviation of the distribution. The calculation of the joint distribution uses the following method:

[0107] ;

[0108] in, is the Dirac function, which represents the probability of occurrence of a specific combination. If the data point deviates from the main area of ​​the joint distribution, the combined data is considered to be abnormal data and filtered. The temperature anomaly filter results, humidity anomaly filter results, and firepower-weight anomaly filter results are normalized to convert the values ​​of different sensors into a unified standard range. The normalization process is performed using the following formula:

[0109] ;

[0110] in, and are the mean and standard deviation of the data sequence, respectively. After normalization, the data of all sensors are represented under the standard normal distribution with zero mean and unit variance, thus ensuring the comparability of data from different sensors. The standardized device status data set is obtained.

[0111] In a specific embodiment, the process of executing step 104 may specifically include the following steps:

[0112] The standardized device status dataset is input into the generator of the conditional generative adversarial network. The generator includes 5 layers of encoders and 5 layers of decoders. Each encoder layer uses a 3x3 convolution kernel with a step size of 2 and a LeakyReLU activation function. Each decoder layer uses a 3x3 deconvolution kernel with a step size of 2 and a ReLU activation function to obtain generated data samples.

[0113] The generated data samples are input into the discriminator of the conditional generative adversarial network. The discriminator contains 4 convolutional layers. Each convolutional layer uses a 4x4 convolution kernel with a step size of 2 and a LeakyReLU activation function to obtain the discriminant probability distribution.

[0114] The authenticity score of the generated data sample is calculated based on the discriminant probability distribution, and the abnormal data is marked based on the authenticity score to obtain the abnormal data marking result;

[0115] Based on the abnormal data labeling results, feature analysis and data interpolation are performed to obtain the abnormal data repair results, and the abnormal data repair results are input into the reconstruction module of the generator. The reconstruction module contains 3 residual blocks, each of which contains two 3x3 convolutional layers and a skip connection to obtain the data reconstruction results;

[0116] Data security verification is performed based on the data reconstruction results to obtain equipment operation status data that has passed security verification.

[0117] Specifically, the standardized device status dataset is input into the generator of the conditional generative adversarial network to generate pseudo data samples related to the input data, which are similar in appearance to the real data. The generator structure includes 5 layers of encoders and 5 layers of decoders. Each layer of encoder uses a 3x3 convolution kernel with a step size of 2 and adopts the LeakyReLU activation function. The mathematical expression of the LeakyReLU activation function is:

[0118] ;

[0119] in, Set to 0.2. Through the convolution operation, the generator is able to extract low-level features of the input data and convert them into higher-dimensional feature representations layer by layer. After each layer of convolution, the LeakyReLU activation function is used to add nonlinear characteristics so that the network can better fit complex distributions. After being processed by 5 layers of encoders, the data enters the decoder part. The decoder consists of 5 layers, and each layer of the decoder uses a 3x3 deconvolution kernel with a stride of 2 and applies the ReLU activation function. The goal of the deconvolution operation is to restore the low-dimensional feature map to a higher dimension so that the output data is as close to the distribution of the input data as possible. The formula for the ReLU activation function is:

[0120] ;

[0121] This function ensures the sparsity of the network and enhances the nonlinear ability of the model by removing negative values. The output of the decoder is the generated pseudo data sample. The generated data sample is input into the discriminator of the conditional generative adversarial network. The discriminator is a binary classification network that determines whether the input data is real data or forged by the generator. The discriminator consists of 4 convolutional layers, each of which uses a 4x4 convolution kernel with a step size of 2 and uses the LeakyReLU activation function. These convolutional layers judge the authenticity of the data by gradually extracting the spatial features of the data. In each convolutional layer, the LeakyReLU activation function plays a role in enhancing the nonlinear characteristics. The output of the discriminator is a probability distribution, indicating the probability that the input data is real data. Assume is the output of the discriminator, the output probability It is expressed as:

[0122] ;

[0123] in, is the input data, is the probability that the data is true. If Approaching 1, it means the data is true; if When it approaches 0, it means that the data is generated pseudo data. In the adversarial training process between the generator and the discriminator, the generator tries to optimize its weights so that the generated data It can deceive the discriminator and obtain a higher authenticity score. The discriminator improves its ability to distinguish between generated data and real data by optimizing its weights. Through this game, the generator gradually learns to generate more realistic data, and the discriminator becomes more accurate. Authenticity score calculated based on the discriminator , marking abnormal data. If the authenticity score of the generated data sample is low, that is, Less than a preset threshold , then these data will be marked as abnormal data. Assume For the generated data samples, the judgment basis for abnormal data labeling is expressed as:

[0124] ;

[0125] in, Representation data is marked as abnormal, Representation data is considered normal. Perform feature analysis and data interpolation on abnormal data. Feature analysis confirms which data points have problems by detecting the distribution characteristics of the data (such as mean, variance, etc.). Interpolation techniques (such as linear interpolation, Lagrange interpolation, etc.) are used to fill in missing data or repair outliers. Through the interpolation algorithm, the missing data is reasonably estimated based on the pattern of the existing data, so as to restore the data value close to the real value. For example, suppose the data of the temperature sensor at some time Lost, calculate its reasonable value through interpolation algorithm:

[0126] ;

[0127] in, and is the interpolation coefficient derived from historical data, and is the temperature data before and after. The repaired data is re-input into the reconstruction module of the generator. The reconstruction module contains 3 residual blocks, each of which consists of two 3x3 convolutional layers and a jump connection. The introduction of residual blocks helps alleviate the gradient vanishing problem and enables the network to better learn deep features. The working method of residual blocks is as follows:

[0128] ;

[0129] in, is the feature obtained through convolution operation, is the input, is the output of the residual block. The skip connection directly connects the input Add to output In order to improve the training efficiency and performance of the network, the reconstruction module is used to repair the data to a state closer to the real data, completing the data reconstruction process. The reconstruction result is used for data security verification. During the security verification process, the repaired data is further verified through multiple security algorithms (such as encryption verification, data consistency detection, etc.) to ensure that the data has not been tampered with or damaged. The device operation status data that passes the security verification is the final output data.

[0130] In a specific embodiment, the process of executing step 105 may specifically include the following steps:

[0131] Classify and analyze the equipment operation status data that has passed the security verification to obtain equipment operation parameter restriction conditions including temperature range, humidity range, and firepower range, and perform legitimacy verification on the user's control request based on the equipment operation parameter restriction conditions to obtain a control parameter set that meets the restriction conditions;

[0132] According to the temperature adjustment instruction, the firepower adjustment instruction, the timing control instruction and the mode switching instruction, the control parameter set that meets the restriction condition is classified to obtain a classified control instruction group;

[0133] Generate standard format instructions for each type of instruction in the classified control instruction group, and perform execution timing planning for the standard format instructions based on preset device operation constraint rules to obtain an initial instruction execution sequence;

[0134] Performing conflict detection and conflict elimination on adjacent instructions in the initial instruction execution sequence to obtain a conflict-eliminated instruction sequence, and grouping the conflict-eliminated instruction sequence to obtain a control instruction data packet sequence;

[0135] The timing integrity of the control instruction data packet sequence is verified to obtain a control instruction sequence in a standard format.

[0136] Specifically, the equipment operation status data that has passed the safety verification is classified and analyzed to extract the equipment's various operating parameters, including temperature, humidity, and firepower. These parameters are collected and normalized by sensors to obtain a temperature range. , Humidity range and range of fire For temperature data, assume that the current temperature of the device is , then determine whether it is within the set range and verify it through the following formula:

[0137] ;

[0138] Similarly, humidity and firepower Respectively and If the current device operating parameters exceed these restrictions, the corresponding control request will be judged as illegal. The user's control request is verified for legitimacy based on the device operating parameter restrictions. The user's control request includes instructions such as temperature adjustment, firepower adjustment, timing control, and mode switching. By verifying the legitimacy of the parameters of these control requests, a control parameter set that meets the device operating parameter restrictions is obtained. For example, if a user issues a temperature adjustment instruction , first check Is it true. If it is true, the control request is legal, and its temperature parameter is added to the control parameter set that meets the restriction conditions. In this way, a complete control parameter set is gradually built to meet the current operating conditions of the equipment. Based on the verified control parameter set, the control instructions are classified. Control instructions are divided into temperature adjustment instructions, firepower adjustment instructions, timing control instructions and mode switching instructions. Each type of instruction is classified and processed according to its characteristics. For example, the temperature adjustment instruction contains the target temperature , the fire control command includes the target fire , the timing control instructions include timing settings , and the mode switch instruction contains the identifier of the new mode . After classifying these instructions, a classified control instruction group is obtained. Generate standard format instructions for each type of instruction. Generate standard format instructions by setting a unified instruction template. The instruction template formats the parameters input by the user to generate a unified instruction format for subsequent processing and execution. According to the preset equipment operation constraint rules, the execution timing of the standard format instructions is planned to ensure that each instruction can be executed in a reasonable order to avoid abnormal equipment operation due to instruction conflicts. The timing planning is dynamically adjusted based on the actual operation of the equipment. For example, suppose that the temperature adjustment instruction needs to be executed before the firepower adjustment instruction, while the timing control and mode switching instructions need to be executed after the temperature adjustment is completed. Design a scheduling algorithm to optimize the execution order of various instructions so that the timing of instruction execution conforms to the operation law of the equipment and avoids conflicts. Perform conflict detection on the preliminary instruction execution sequence to identify and eliminate instructions that will interfere with the normal operation of the equipment. For example, if the temperature adjustment and firepower adjustment instructions are issued at the same time, it will cause excessive load on the equipment, and reasonable scheduling is required during timing planning. Conflict detection is completed by comparing the execution time window of the instruction and the resource requirements. For example, suppose the execution time window of the temperature adjustment instruction is , the execution time window of the firepower adjustment command is ,if and , then there is a conflict between the two instructions. When a conflict occurs, the conflict is eliminated by adjusting the time window to avoid mutual interference of the instructions. The instruction sequence that eliminates the conflict is grouped for data transmission and execution. A control instruction data packet sequence is generated by dividing adjacent control instructions into data packets in chronological order. For example, in some cases, multiple instructions are combined into one data packet for transmission together, which helps to reduce communication overhead and improve the response speed of the system. The control instruction data packet sequence is verified for timing integrity to ensure that the instructions are not lost or tampered with during execution. Timing integrity verification is performed by setting timestamps and checksums. Through timestamps, the execution order of each control instruction is tracked to ensure that the execution of the instructions is carried out according to the predetermined timing. The checksum is used to ensure that there is no error in the transmission of instruction data. If any instruction is found to be lost or tampered with during the verification process, the exception handling mechanism will be triggered and the instruction will be resent. Through the above steps, a standard format control instruction sequence is finally obtained.

[0139] In a specific embodiment, the process of executing step 106 may specifically include the following steps:

[0140] Dividing the standard format control instruction sequence into a plurality of communication control instruction subgroups based on instruction data transmission frequency, data packet size and communication priority;

[0141] Allocating time segments to multiple communication control instruction subgroups to obtain a basic communication time allocation scheme, and staggering the time segments in the basic communication time allocation scheme to obtain a staggered communication scheduling sequence;

[0142] The instruction transmission is executed based on the staggered communication scheduling sequence, and the average residence time automaton is used to record the residence time of the data packet of each communication control instruction subgroup in the communication channel to obtain the communication resource occupancy data;

[0143] Calculate the average transmission delay and abnormal flag of each communication control instruction subgroup according to the communication resource occupancy data, and obtain the transmission abnormal flag;

[0144] The time slices are reallocated for the communication control instruction subgroup where the transmission abnormality mark is located, and a new time slice allocation ratio is calculated through a time ratio monitor to obtain a dynamically adjusted time allocation plan;

[0145] Reorganize the instruction transmission sequence based on the dynamically adjusted time allocation scheme, group and transmit the standard format control instruction sequence according to the updated time segments, and obtain the instruction transmission status record;

[0146] Arrange the instruction transmission status records in the order of timestamps to obtain the group transmission execution results.

[0147] Specifically, based on the transmission frequency, data packet size and communication priority of the command, the standard format control command sequence is divided into multiple communication control command subgroups. Control commands include various types, such as temperature adjustment, firepower adjustment, timing control, etc. The division of each command subgroup should take into account the characteristics of each command: the transmission frequency determines the sending interval of the command, the data packet size determines the size of each data packet, and the communication priority determines the transmission order of the command. The following formula is used to express the transmission frequency and data packet size of each command subgroup:

[0148] ;

[0149] in, For the The transmission frequency of each instruction, For the The time interval between instructions, is the size of the data packet, is the total data size of the instruction, The size of each packet. Based on these parameters, the instruction sequence is divided into multiple subgroups and prioritized according to the communication needs of each subgroup. Time slices are allocated to the communication control instruction subgroups. Each subgroup is allocated a time window to ensure that they can be transmitted on time without conflict. During the time allocation process, the time window of each instruction subgroup is calculated according to its transmission frequency. The length of each time slice is set to , and then dynamically allocate according to the transmission needs of each subgroup. The transmission frequency of each instruction subgroup is , then its time segment It is expressed as:

[0150] ;

[0151] According to the length of the time segment, a basic communication time allocation scheme is obtained. In order to avoid conflicts among multiple instruction subgroups in the same time window, the time segments are staggered to generate a staggered communication scheduling sequence to ensure that multiple instruction subgroups do not occupy the same time window at the same time, thereby reducing transmission conflicts. After generating the staggered communication scheduling sequence, the transmission of the instructions is executed. The average residence time automaton is used to record the residence time of each communication control instruction subgroup in the communication channel to obtain the communication resource occupancy data. The average residence time automaton model evaluates the resource occupancy by calculating the time that the data packet of each instruction subgroup stays in the communication channel. Assume that The stay time of the data packet of the instruction subgroup in the channel is , then calculate the average residence time of each subgroup:

[0152] ;

[0153] in, For the The number of packets transmitted by each instruction subgroup, For the In the subgroup The residence time of each data packet. Through the calculation process, the resource usage of each instruction subgroup is obtained. Based on the resource usage data, the average transmission delay and abnormal flag of each communication control instruction subgroup are calculated. Delay calculation is a key indicator for evaluating network performance, which can help the system determine whether the instruction reaches the destination on time. Assume that the transmission delay of each subgroup is , the average transmission delay is calculated by the following formula:

[0154] ;

[0155] If the transmission delay of a subgroup exceeds the preset delay threshold, the subgroup will be marked as abnormal. The generation of abnormal marking is based on a threshold judgment. , then The instruction subgroups will be marked as abnormal. For the instruction subgroups with transmission abnormality marks, the time segments are reallocated to optimize transmission efficiency and reduce latency. The time segments of the subgroups with abnormalities are adjusted to better utilize communication resources. In this process, the time ratio monitor is used to calculate the new time segment allocation ratio. The time ratio monitor dynamically adjusts the time segment ratio of each subgroup according to the actual load and communication conditions of the network. Assume that the newly allocated time segment ratio is , then adjust according to the following formula:

[0156] ;

[0157] in, For the The adjusted proportion of the subgroups, is the total number of instruction subgroups. Based on the dynamically adjusted time allocation scheme, the transmission order of instructions is reorganized to ensure that each subgroup is transmitted in the new timing order. By optimizing the transmission order of instructions, conflicts caused by overlapping time segments are reduced, thereby improving the efficiency of the system. After all adjustments are completed, the transmission status of the instructions is recorded. The instruction transmission status record includes information such as the transmission status, delay, and conflict detection results of each subgroup. These records are sorted in the order of timestamps to ensure that the system can trace the execution of each instruction and monitor the operating status of the system in real time. The complete group transmission execution results are obtained through the sorted transmission status records.

[0158] The above describes the remote control method of the cooking device in the embodiment of the present invention. The following describes the remote control device of the cooking device in the embodiment of the present invention. Figure 2 , an embodiment of the remote control device for cooking equipment in an embodiment of the present invention includes:

[0159] The collection module 201 is used to collect the operation instruction sequence executed by the user on the mobile terminal, and extract the time feature of the operation instruction sequence to obtain the user operation feature data;

[0160] The identity authentication module 202 is used to input the user operation feature data into the cooking operation sequence Transformer model with time and space attention for feature matching and identity authentication to obtain a user identity authentication result;

[0161] The processing module 203 is used to obtain the sensor data collected by the cooking equipment, and perform outlier filtering and data normalization processing to obtain a standardized equipment status data set;

[0162] The security verification module 204 is used to input the standardized device status data set into the conditional generation adversarial network to perform data security verification, and obtain the device operation status data that passes the security verification;

[0163] A generation module 205, for generating a control instruction sequence in a standard format based on the user identity authentication result and the device operation status data that has passed the security verification;

[0164] The execution module 206 is used to perform communication grouping and transmission execution on the standard format control instruction sequence to obtain corresponding grouping transmission execution results.

[0165] Through the collaborative cooperation of the above components, by introducing a user identity authentication mechanism based on operation behavior patterns and combining it with the Transformer model with spatiotemporal attention enhancement, the user's unique operation habit characteristics can be effectively extracted, the risk of identity feature leakage can be reduced, and the security of remote control can be enhanced; an improved conditional generative adversarial network is used to securely verify the device status data, and through the adversarial learning of the generator and the discriminator, abnormal data can be accurately identified and repaired, ensuring the authenticity and validity of the sensor data; a dynamic time allocation communication mechanism is designed, and control instructions are transmitted in groups based on the communication resource occupancy characteristics, and dynamically adjusted through the average residence time automaton and time ratio monitor, effectively avoiding communication congestion and instruction loss; through multi-level data processing and security verification mechanisms, the full process security protection from user authentication, data collection to instruction transmission is achieved, which significantly improves the reliability and stability of remote control of cooking equipment.

[0166] above Figure 2 The remote control device for the cooking device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The remote control device for the cooking device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0167] Figure 31 is a schematic diagram of the structure of a remote control device for a cooking device provided by an embodiment of the present invention. The remote control device 300 of the cooking device may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 may be temporary storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the remote control device 300 of the cooking device. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the remote control device 300 of the cooking device to implement the steps of the remote control method of the cooking device described above.

[0168] The cooking device remote control device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 3 The illustrated remote control device structure of the cooking device does not constitute a limitation on the remote control device of the cooking device provided by the present invention, and may include more or fewer components than illustrated, or a combination of certain components, or a different arrangement of components.

[0169] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the remote control method of the cooking device.

[0170] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0171] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0172] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A remote control method for a cooking device, characterized in that: The method comprises: Collecting the operation instruction sequence executed by the user on the mobile terminal, and extracting the time feature of the operation instruction sequence to obtain user operation feature data; Inputting the user operation feature data into a cooking operation sequence Transformer model with time-space attention for feature matching and identity authentication to obtain a user identity authentication result; Obtain sensor data collected by cooking equipment, filter outliers and normalize the data to obtain a standardized equipment status data set; Inputting the standardized device status data set into a conditional generative adversarial network for data security verification to obtain device operation status data that passes security verification; Based on the user identity authentication result and the device operation status data that has passed the security verification, a standard format control instruction sequence is generated; specifically comprising: classifying and parsing the device operation status data that has passed the security verification to obtain device operation parameter restriction conditions including temperature range, humidity range, and firepower range, and performing a legitimacy check on the user's control request based on the device operation parameter restriction conditions to obtain a control parameter set that meets the restriction conditions; classifying the control parameter set that meets the restriction conditions according to temperature adjustment instructions, firepower adjustment instructions, timing control instructions, and mode switching instructions to obtain a classified control instruction group; generating a standard format instruction for each type of instruction in the classified control instruction group, and performing execution timing planning for the standard format instructions based on preset device operation constraint rules to obtain an initial instruction execution sequence; performing conflict detection and conflict elimination on adjacent instructions in the initial instruction execution sequence to obtain a conflict-eliminated instruction sequence, and grouping the conflict-eliminated instruction sequence to obtain a control instruction data packet sequence; performing timing integrity verification on the control instruction data packet sequence to obtain a standard format control instruction sequence; The standard format control instruction sequence is grouped and transmitted to obtain a corresponding group transmission execution result.

2. The remote control method for cooking equipment according to claim 1, characterized in that: The collecting of the operation instruction sequence executed by the user on the mobile terminal and extracting the time feature of the operation instruction sequence to obtain the user operation feature data includes: Collecting an operation instruction sequence executed by a user on a mobile terminal, and processing the operation instruction sequence to obtain operation time interval data, operation duration data, operation force data, and operation trajectory data; Analyze the operation time interval data and the operation duration data to obtain a timestamp feature of the operation, and perform feature calculation on the operation force data and the operation trajectory data to obtain a type identification feature of the operation; Constructing a time series feature including a timestamp, a type identifier and an operation parameter based on a timestamp feature of the operation and a type identifier feature of the operation; Comparing and analyzing the time series features with the user's historical usage habit data to obtain the user's operation habit features, and performing correlation analysis on the operation habit features and the user's historical cooking experience data to obtain the user's cooking operation behavior features; The user's cooking operation behavior characteristics are segmented according to a preset time window to obtain a plurality of characteristic data segments, and the plurality of characteristic data segments are fused with time and space characteristics to obtain user operation characteristic data.

3. The remote control method for cooking equipment according to claim 2, characterized in that: The step of inputting the user operation feature data into a cooking operation sequence Transformer model with time-space attention for feature matching and identity authentication to obtain a user identity authentication result includes: Input the user operation feature data into the operation sequence embedding layer in the cooking operation sequence Transformer model with time-space attention, wherein the operation sequence embedding layer transforms and maps the user operation feature data using a linear projection matrix with a dimension of 512, and superimposes position encoding information to obtain an initial feature embedding vector; The query matrix, key matrix and value matrix of dimension 64 corresponding to the initial feature embedding vector are respectively calculated by an 8-head parallel attention mechanism, and the query matrix and the key matrix are matrix multiplied to obtain an attention score matrix of the time dimension; Based on the attention score matrix of the time dimension, weighted summation is performed on the features at different times, and normalized temporal attention features are obtained through LayerNormalization normalization and residual connection; The normalized temporal attention features are input into the spatial attention layer, the spatial dependency of the features is enhanced by the spatial position encoding matrix, and the spatial attention map is calculated by an adaptive weight mechanism to obtain a spatial dimension feature representation; Performing a nonlinear transformation on the spatial dimension feature representation to obtain a deep semantic feature representation, and inputting the deep semantic feature representation into a spatiotemporal feature fusion layer for adaptive weighted fusion to obtain a spatiotemporal fusion feature of the user operation; The cosine similarity is calculated for the spatiotemporal fusion feature of the user operation and at least three historical operation template features stored in the user historical operation feature library, the maximum similarity value is selected as the feature matching score, and the user identity authentication result is generated according to the feature matching score.

4. The remote control method of cooking equipment according to claim 3, characterized in that: The method of acquiring sensor data of the cooking device, filtering outliers and normalizing the data to obtain a standardized device status data set includes: Parse the raw data collected by the temperature sensor, humidity sensor, fire power sensor and weight sensor in the cooking equipment to obtain the raw data of the equipment status including the data collection timestamp, sensor identification and sensor value; Classify and sort the raw data of the equipment status according to the sensor identification to obtain a temperature data sequence, a humidity data sequence, a firepower data sequence and a weight data sequence; Calculate the temperature sliding mean and standard deviation based on the temperature data sequence, and filter the data of the temperature data sequence that exceeds a preset first standard deviation range to obtain a temperature outlier filtering result; Calculating the humidity sliding mean and standard deviation based on the humidity data sequence, and filtering the data in the humidity data sequence that exceeds a preset second standard deviation range to obtain a humidity outlier filtering result; Calculating the firepower-weight joint distribution based on the firepower data sequence and the weight data sequence, and filtering the abnormal combination data that deviates from the main distribution area to obtain a firepower-weight abnormal value filtering result; The temperature abnormal value filtering result, the humidity abnormal value filtering result and the firepower-weight abnormal value filtering result are normalized to obtain a standardized equipment status data set.

5. The remote control method of cooking equipment according to claim 4, characterized in that: The step of inputting the standardized device status data set into a conditional generative adversarial network for data security verification to obtain device operation status data that passes security verification includes: Input the standardized device state data set into a conditional generative adversarial network generator, wherein the generator includes a 5-layer encoder and a 5-layer decoder, each layer of the encoder uses a 3x3 convolution kernel with a step size of 2 and a LeakyReLU activation function, and each layer of the decoder uses a 3x3 deconvolution kernel with a step size of 2 and a ReLU activation function, to obtain a generated data sample; Input the generated data sample into the discriminator of the conditional generative adversarial network, where the discriminator includes 4 convolutional layers, each of which uses a 4x4 convolution kernel with a step size of 2 and a LeakyReLU activation function to obtain a discriminant probability distribution; Calculating the authenticity score of the generated data sample based on the discriminant probability distribution, and marking the abnormal data based on the authenticity score to obtain an abnormal data marking result; Perform feature analysis and data interpolation based on the abnormal data labeling result to obtain an abnormal data repair result, and input the abnormal data repair result into the reconstruction module of the generator, wherein the reconstruction module includes three residual blocks, each residual block includes two 3x3 convolutional layers and one skip connection, to obtain a data reconstruction result; Data security verification is performed based on the data reconstruction result to obtain device operation status data that passes the security verification.

6. The remote control method of cooking equipment according to claim 1, characterized in that: The communication grouping and transmission execution of the standard format control instruction sequence to obtain the corresponding grouping transmission execution result includes: Dividing the standard format control instruction sequence into a plurality of communication control instruction subgroups based on instruction data transmission frequency, data packet size, and communication priority; Allocating time segments to the multiple communication control instruction subgroups to obtain a basic communication time allocation scheme, and staggering the time segments in the basic communication time allocation scheme to obtain a staggered communication scheduling sequence; Execute instruction transmission based on the staggered communication scheduling sequence, and use an average residence time automaton to record the residence time of the data packet of each communication control instruction subgroup in the communication channel to obtain communication resource occupancy data; Calculate the average transmission delay and abnormal flag of each communication control instruction subgroup according to the communication resource occupancy data to obtain a transmission abnormal flag; Re-allocating the time segments of the communication control instruction subgroup where the transmission abnormality mark is located, calculating the new time segment allocation ratio through the time ratio monitor, and obtaining a dynamically adjusted time allocation scheme; Reorganize the instruction transmission sequence based on the dynamically adjusted time allocation scheme, group and transmit the standard format control instruction sequence according to the updated time segments, and obtain an instruction transmission status record; The instruction transmission status records are sorted in the order of timestamps to obtain the packet transmission execution result.

7. A remote control device for a cooking device, characterized in that: The device is used to execute the remote control method of the cooking device according to any one of claims 1 to 6, comprising: A collection module is used to collect the operation instruction sequence executed by the user on the mobile terminal, and extract the time feature of the operation instruction sequence to obtain user operation feature data; An identity authentication module, used for inputting the user operation feature data into a cooking operation sequence Transformer model with time-space attention for feature matching and identity authentication, and obtaining a user identity authentication result; A processing module is used to obtain sensor data collected by cooking equipment, filter outliers and normalize the data to obtain a standardized equipment status data set; A security verification module, used to input the standardized device status data set into a conditional generative adversarial network for data security verification, and obtain device operation status data that passes the security verification; A generation module, used to generate a control instruction sequence in a standard format based on the user identity authentication result and the device operation status data that has passed the security verification; The execution module is used to perform communication grouping and transmission execution on the standard format control instruction sequence to obtain corresponding grouping transmission execution results.

8. A remote control device for a cooking device, characterized in that: The remote control device of the cooking device comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory to enable the remote control device of the cooking device to execute the remote control method for a cooking device according to any one of claims 1 to 6.

9. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the remote control method for the cooking device according to any one of claims 1 to 6 is implemented.

Images