Operator identity type identification method, system and device and storage medium

By obtaining manual sliding and sensor data, combined with fine-grained auxiliary classification module and denoising technology, the problem of poor child identity recognition in the existing technology is solved, efficient and privacy-protected operator identity type recognition, and improved classification Accuracy and model performance.

CN118965318APending Publication Date: 2024-11-15XI AN JIAOTONG UNIV
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411101397.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing human-machine behavior analysis technology has problems such as poor classification effect, poor generalization of the model and limited application scope in children's identity recognition. In addition, traditional biometric recognition methods violate privacy and are difficult to achieve unsensed continuous authentication.

Method used

By obtaining the operator's manual sliding data and sensor data, using a pre-trained operator identity type identification model, combining fine-grained auxiliary classification authentication module and auxiliary losses for pre-training, the identification of children and adult identities is achieved, and the wavelet base is used. Function denoising and PDF analysis methods improve feature extraction and classification performance.

Benefits of technology

It realizes efficient identification of the operator's identity type without infringing on the operator's privacy, improves classification accuracy and model learning efficiency, optimizes the data acquisition process, and is suitable for practical application scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965318A_ABST
    Figure CN118965318A_ABST
Patent Text Reader

Abstract

The invention belongs to the field of man-machine behavior analysis, and discloses an operator identity type identification method, system and device and a storage medium, and the method comprises the steps: obtaining the manual sliding data of an operator and the sensor data of an operation object; according to the manual sliding data and the sensor data, the identity type of the operator is obtained through a pre-trained operator identity type recognition model; wherein when the operator identity type recognition model is pre-trained, pre-training is carried out by setting an auxiliary classification authentication module which is finer than a classification authentication module of the operator identity type recognition model and combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module. And privacy data of an operator is not involved in a data acquisition process. In the aspect of model training, an auxiliary training method is used, the loss value of a finer-grained label can be obtained, the loss value is further combined with the original loss to guide the overall loss reduction of the model, the learning efficiency of the overall model is improved, the training process is optimized, and finally better classification performance is obtained.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of human-machine behavior analysis and relates to an operator identity type recognition method, system, device and storage medium. Background Art

[0002] In the analysis of human-computer behavior, early research mainly focused on biometric features such as face, iris and fingerprint. With the help of these human-computer interaction behaviors, a series of new identity authentication systems based on biometric features of human-computer interaction behaviors have been developed. For example, the face unlocking and fingerprint unlocking methods of mobile phones are now very common. Compared with the previous password, this method reduces the operator's memory cost and achieves faster identification and authentication. With the help of the face, the operator's gender, age and other attributes can also be recognized. However, this type of identification method must collect important identity information such as the operator's face and fingerprint, and the operator cannot completely protect his or her privacy information. If this system is hacked, a large amount of the operator's real face and fingerprint information will be leaked. In addition, the use of this method for identification and authentication requires external devices such as cameras and fingerprint sensors, which makes it difficult for the operator to feel it and achieve continuous authentication and recognition. Researchers began to pay attention to the inherent physiological and behavioral biometrics of operators, such as touch screen habits and mobile phone holding habits, and developed a series of authentication and identification systems.

[0003] Due to the ease of use of smartphones, people are increasingly inclined to use them for gaming and entertainment. Among these users, there are many children and minors whose cognitive and behavioral abilities are not yet mature and who are easily addicted to games. In real life, since children's guardians are often unable to take care of them in time, the safety hazards and financial risks caused by children's misoperation in individual game scenarios will greatly increase.

[0004] Based on this, the academic community has begun to use the inherent physiological characteristics and behavioral habits of operators when using mobile phones to classify and identify the operator's identity type through machine models. The above work has made initial progress, but there are still problems such as data collection only completed in laboratory scenarios, simple model training methods, resulting in poor classification results, and weak model generalization. Most of the work can only identify individual simple applications. In practical applications, it is difficult to effectively classify the age of operators. Summary of the invention

[0005] The purpose of the present invention is to overcome the above-mentioned shortcomings of the prior art and provide an operator identity type identification method, system, device and storage medium.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] According to a first aspect of the present invention, a method for identifying an operator's identity type is provided, comprising: obtaining manual sliding data of an operator and sensor data of an operation object; obtaining the operator's identity type through a pre-trained operator identity type identification model according to the manual sliding data and the sensor data; wherein, during pre-training, the operator identity type identification model is pre-trained by setting an auxiliary classification authentication module that is finer-grained than the classification authentication module of the operator identity type identification model, and combining the loss of the classification authentication module with the auxiliary loss of the auxiliary classification authentication module.

[0008] Optionally, the operator's manual sliding data is obtained by classifying the sliding coordinates of the same finger number from appearance to disappearance in timestamp order; the sensor data of the operation object includes one or more of the following: screen direction flag, accelerometer data and gyroscope data.

[0009] Optionally, before obtaining the operator's identity type through a pre-trained operator identity type recognition model based on manual sliding data and sensor data, the method also includes: traversing each manual sliding data, retaining only the first sliding coordinate at the beginning for multiple repeated sliding coordinates in a continuous time, and deleting other repeated data; and using a wavelet basis function to remove noise in the sensor data.

[0010] Optionally, the method of obtaining the operator's identity type according to the manual sliding data and the sensor data through a pre-trained operator identity type recognition model includes: obtaining manual sliding feature data based on the manual sliding data and a preset manually designed feature group; dividing the sensor data into a plurality of two-dimensional sensor window data according to a preset sliding window, and splicing the plurality of two-dimensional sensor window data to obtain sensor feature data; inputting the manual sliding feature data and the sensor feature data into a pre-trained operator identity type recognition model to obtain the operator's identity type; wherein the manually designed feature group is obtained by: calculating each physical quantity generated during the sliding process, The mathematical statistical values ​​of the physical quantity are combined with the physical quantities and the mathematical statistical values ​​of the physical quantities to obtain a number of manually designed features; the physical quantities include one or more of the following: position, length, speed, acceleration, angle, attitude angle and sensor sequence, and the mathematical statistical values ​​include one or more of the following: mean, variance, maximum value, minimum value, kurtosis and skewness; based on the PDF analysis method, the distribution frequency of each manually designed feature of different identity types is taken as a distribution frequency vector, and the distance between each distribution frequency vector of different identity types is calculated to obtain the discriminant value of each manually designed feature; the manually designed features corresponding to the discriminant values ​​with a large number of preset values ​​are selected to obtain a manually designed feature group.

[0011] Optionally, the operator identity type recognition model includes a feature extraction module and a classification authentication module; the feature extraction module includes a convolution module, a fully connected module and a splicing module; the convolution module is used to extract data features of manual sliding data, the fully connected module is used to extract data features of sensor data, and the splicing module is used to splice data features of manual sliding data and data features of sensor data to obtain splicing features; the convolution module includes a two-dimensional convolution layer, four residual blocks and a fully connected layer connected in sequence; the fully connected module includes a batch normalization layer and five fully connected layers connected in sequence; the classification authentication module includes two fully connected layers connected in sequence.

[0012] Optionally, the identity types of the classification authentication module of the operator identity type recognition model include children and adults; and the identity types of the auxiliary classification authentication module include four age groups divided according to age.

[0013] Optionally, the pre-training by combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module includes: combining the loss of the classification authentication module main and the auxiliary loss l of the auxiliary classification and authentication module aux , the combined loss l of the operator identity type recognition model is obtained by the following formula total :

[0014] ltotaall=lmain+αlaux

[0015] Among them, α is a preset constant; according to the operator identity type recognition model, the combined loss l total , pre-training the operator identity type recognition model; and after each number of trainings, adjusting the learning rate of the operator identity type recognition model to a previously preset multiple.

[0016] According to a second aspect of the present invention, there is provided an operator identity type recognition system, comprising: a data acquisition module, for acquiring manual sliding data of the operator and sensor data of the operation object; a type recognition module, for obtaining the identity type of the operator according to the manual sliding data and the sensor data through a pre-trained operator identity type recognition model; wherein, during pre-training, the operator identity type recognition model is pre-trained by setting an auxiliary classification authentication module that is finer-grained than the classification authentication module of the operator identity type recognition model, and combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module.

[0017] According to a third aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned operator identity type identification method when executing the computer program.

[0018] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned operator identity type identification method are implemented.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The operator identity type recognition method of the present invention is based on the operator's manual sliding data and the sensor data of the operation object, and realizes the identification and authentication of the operator's identity type through a pre-trained operator identity type recognition model. The operator's privacy data is not involved in the data collection process, and implicit collection in the background can be realized without affecting the operator's experience. Among them, when the operator identity type recognition model is pre-trained, an auxiliary classification authentication module with finer granularity is set compared to the classification authentication module of the operator identity type recognition model, and pre-training is performed in combination with the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module. In terms of model training, an auxiliary training method is used. The auxiliary training can obtain the loss value of finer granularity labels, and further combine with the original loss to guide the overall loss of the model to decrease, which can improve the learning efficiency of the overall model, optimize the model training process, and finally obtain an operator identity type recognition model with better classification performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Flow chart of the operator identity type identification method according to an embodiment of the present invention.

[0022] Figure 2 This is a structural block diagram of an operator identity type recognition model according to an embodiment of the present invention.

[0023] Figure 3 4 is a structural block diagram of an operator identity type recognition system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only 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 should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0027] See also Figure 1 In one embodiment of the present invention, a method for identifying an operator's identity type is provided, which can effectively improve the accuracy of identifying an operator's identity type.

[0028] Specifically, the operator identity type identification method of the present invention includes the following steps:

[0029] S1: Acquire the operator's manual sliding data and the sensor data of the operation object.

[0030] S2: According to the manual sliding data and the sensor data, the operator's identity type is obtained through a pre-trained operator identity type recognition model.

[0031] Among them, during the pre-training of the operator identity type recognition model, an auxiliary classification authentication module with finer granularity than the classification authentication module of the operator identity type recognition model is set, and pre-training is performed by combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module.

[0032] The operator identity type recognition method of the present invention is based on the operator's manual sliding data and the sensor data of the operation object, and realizes the identification and authentication of the operator's identity type through a pre-trained operator identity type recognition model. The operator's privacy data is not involved in the data collection process, and implicit collection in the background can be realized without affecting the operator's experience. Among them, when the operator identity type recognition model is pre-trained, an auxiliary classification authentication module with finer granularity is set compared to the classification authentication module of the operator identity type recognition model, and pre-training is performed in combination with the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module. In terms of model training, an auxiliary training method is used. The auxiliary training can obtain the loss value of finer granularity labels, and further combine with the original loss to guide the overall loss of the model to decrease, which can improve the learning efficiency of the overall model, optimize the training process of the model, and finally obtain an operator identity type recognition model with better classification performance.

[0033] In a possible implementation, the pre-training of the operator identity type recognition model is performed based on the collected raw data, where the raw data includes the operator's manual sliding data and the sensor data of the operation object.

[0034] Specifically, the operator's data on the display screen and the sensor data are collected. Each time the raw data is collected, it includes: timestamp, screen direction flag, finger touch coordinates, finger sliding action type, finger number, accelerometer data, gyroscope data, operator ID, operator age group and operator age. In addition, the collected raw data is preprocessed to remove invalid data rows without finger sliding actions, and the corresponding sensor features such as pitch, roll and raw are calculated based on the acceleration data and gyroscope data at each moment; then, the sequence of sliding coordinates of the same finger number from appearance to disappearance is classified into a sliding action vector in the order of timestamp, and finally a sliding action vector sequence is formed, in which each unit contains the complete sliding action data of a single finger. Finally, the operator's manual sliding data and the sensor data of the operation object are formed.

[0035] Optionally, during pre-training, the collected original operator's manual sliding data is preprocessed. Since the sampling frequency is high when collecting data, there may be multiple repeated coordinate points in the sliding process. Each sliding action is traversed, and for multiple repeated sliding coordinate data in a continuous time, only the first coordinate data is retained, and the remaining duplicate data is deleted to reduce the redundancy of the overall manual sliding data.

[0036] For the original sensor data of the operating object collected, it is necessary to remove the noise in the data and retain the useful information of the signal. First, select the appropriate wavelet basis function Daubechies wavelet, and perform wavelet transform on the data of the three axes collected from the acceleration sensor and gyroscope respectively, decompose the signal into approximate coefficients and detail coefficients of different scales, then perform threshold processing on the detail coefficients in the wavelet domain, set the smaller detail coefficients to zero or scale them, thereby suppressing the noise component, and finally perform inverse wavelet transform on the processed approximate coefficients and detail coefficients to reconstruct the sensor data after removing the noise.

[0037] For manual sliding data, based on the spatiotemporal characteristics of touch operations, we first analyze the original data, and then perform manual feature design on the position, length, speed, acceleration, angle, posture angle, and sensor sequence generated during the operator's sliding process. We calculate mathematical statistical values ​​such as mean, variance, maximum, minimum, kurtosis, and skewness, and finally obtain multi-dimensional manually designed features.

[0038] In this embodiment, in order to select a manual feature combination with high discrimination, a PDF analysis method is used. The distribution frequency of each manually designed feature of each identity type is taken as a vector. Then, the distance between the distribution frequency vectors of different identity types is calculated as a measure of feature discrimination. The measure of feature discrimination is sorted, and the top 36 are selected to obtain the final 36-dimensional manually designed feature group.

[0039] The specific features obtained in this embodiment are: sliding duration, relative X-axis position of sliding end point, relative Y-axis position of sliding end point, X-axis range, Y-axis range, X-axis maximum value, X-axis minimum value, Y-axis maximum value, Y-axis minimum value, end-to-end distance, track distance, track direction, contact area mean, contact area variance, vector velocity mean, track velocity, acceleration mean, acceleration variance, pitch angle mean, yaw angle mean, roll angle mean, pitch angle variance, yaw angle variance, roll angle variance, pitch angle minimum value, yaw angle minimum value, roll angle minimum value, pitch angle maximum value, yaw angle maximum value, roll angle maximum value, pitch angle root mean square, yaw angle root mean square, roll angle root mean square, pitch angle average deviation, yaw angle average deviation and roll angle average deviation. Then, according to the manual sliding data, based on the preset manual design feature group, the manual sliding feature data is obtained. In addition, in order to meet the needs of neural network input, multiple row vectors are merged into a matrix to convert the original multiple one-dimensional hand-designed feature data into a two-dimensional data form suitable for the model: [f0,f1,…,f n ].

[0040] For sensor data, the window sliding step length h is designed. The original sensor data is divided into multiple two-dimensional sensor window data with a length of h according to the length of the window. Finally, multiple two-dimensional sensor window data are spliced ​​into three-dimensional features suitable for model input [[f 0,0 ,……,f 0,x ],……,[f y,0 ,……,f y,x ]].

[0041] In this embodiment, the operator identity type recognition model includes a feature extraction module and a classification authentication module; the feature extraction module includes a convolution module, a fully connected module and a splicing module; the convolution module is used to extract data features of manual sliding data, the fully connected module is used to extract data features of sensor data, and the splicing module is used to splice data features of manual sliding data and data features of sensor data to obtain splicing features; the convolution module includes a two-dimensional convolution layer, four residual blocks and a fully connected layer connected in sequence; the fully connected module includes a batch normalization layer and five fully connected layers connected in sequence; the classification authentication module includes two fully connected layers connected in sequence.

[0042] See also Figure 2 In this embodiment, the convolution module is constructed as follows: Conv16, Resblock64, Resblock128, Resblock256, Resblock512 and fc128 are built layer by layer, and the activation function is relu. The fully connected module is constructed as follows: Normalization (normalization layer), fc256, fc512, fc512, fc256 and fc128 are built layer by layer, and the activation function is relu. The classification authentication module is constructed as follows: fc64 and fc2 are built layer by layer, the fc64 activation function is relu, and the fc2 activation function is softmax. In addition, an auxiliary classification authentication module is set up to realize auxiliary training, and the auxiliary classification authentication module is constructed as follows: fc64 and fc4 are built layer by layer, the fc64 activation function is relu, and the fc2 activation function is softmax.

[0043] Among them, Conv16: is a two-dimensional convolutional layer using 16 convolution kernels (or filters). Resblock64, Resblock128, Resblock256 and Resblock512 are residual blocks with 64, 128, 256 and 512 output channels respectively, and fc128, fc256, fc512, fc64, fc2 and fc4 are fully connected layers with 128, 256, 512, 64, 2 and 4 neurons respectively.

[0044] The outputs of the convolutional module and the fully connected module are concatenated and merged as the input of the classification authentication module and the auxiliary classification authentication module to complete the training environment of the entire operator identity type recognition model.

[0045] In a possible implementation manner, the pre-training of combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module includes: combining the loss of the classification authentication module main and the auxiliary loss l of the auxiliary classification and authentication module aux , the combined loss l of the operator identity type recognition model is obtained by the following formula total :

[0046] total =lmain+αlaux

[0047] Among them, α is a preset constant.

[0048] Combination loss l of the identification model according to operator identity type total , pre-training of operator identity type recognition model.

[0049] In this implementation manner, the identity types of the classification authentication module of the operator identity type recognition model include children and adults, and the identity types of the auxiliary classification authentication module include four age groups divided according to age.

[0050] During specific training, the data is first divided into a training set and a test set. In addition to being divided into two categories of labels, adults and children, the data labels in the training set are divided into four categories of labels, namely, 3 to 8 years old, 9 to 14 years old, 15 to 17 years old, and over 18 years old, according to fine-grainedness. Among them, the two categories of labels, adults and children, are used to calculate the main loss, i.e., the loss of the classification and authentication module, in combination with the output of the classification and authentication module, and the four categories of labels are used to calculate the auxiliary loss in the auxiliary classification and authentication part. Given the training optimizer, initial learning rate, loss function, batch size, and training rounds, the extracted manual sliding feature data is input into the fully connected module, and the sensor feature data is input into the convolution module to train the entire operator identity type recognition model. The specific training process is as follows:

[0051] Step 2-1: Organize the extracted manual sliding feature data and sensor feature data into a tensor structure, set the training batch size to 64, and input the feature extraction part to train the entire model.

[0052] Step 2-2: After each round of training, evaluate the model performance on the test set. If the indicators meet the classification performance requirements, stop training; otherwise, continue training.

[0053] Optionally, after each training, the learning rate of the operator identity type recognition model is adjusted to a preset multiple. In this embodiment, the learning rate is adjusted once every 10 trainings to 0.1 times the previous value.

[0054] In a possible implementation, in combination with the pre-training process of the operator identity type recognition model, before obtaining the operator's identity type based on the manual sliding data and sensor data through the pre-trained operator identity type recognition model, it also includes: traversing each manual sliding data, retaining only the first sliding coordinate at the beginning for multiple repeated sliding coordinates in a continuous time, and deleting the remaining repeated data; using wavelet basis functions to remove noise in the sensor data.

[0055] In a possible implementation, in combination with the pre-training process of the operator identity type recognition model, the operator identity type is obtained based on the manual sliding data and the sensor data through the pre-trained operator identity type recognition model, including: obtaining manual sliding feature data based on the manual sliding data and a preset manually designed feature group; dividing the sensor data into a plurality of two-dimensional sensor window data according to a preset sliding window, and splicing the plurality of two-dimensional sensor window data to obtain sensor feature data; inputting the manual sliding feature data and the sensor feature data into the pre-trained operator identity type recognition model to obtain the operator identity type.

[0056] The operator identity type identification method of the present invention can realize an identification method based on the setting of the identity type label, and realizes the full process of child operator identification through the touch screen data and sensor data generated when the operator interacts with the mobile phone. Compared with the currently available child identification system on the mobile phone. The method of the present invention does not involve the operator's privacy data during the data collection process, and can realize implicit collection in the background without affecting the operator's experience. In data processing, the method of the present invention filters the sensor data and deduplicates the sliding data, thereby improving the availability of the data. In terms of model training, a dual-input model and an auxiliary training method are used. The auxiliary training method proposed in the method of the present invention can not only be used for child identification, but also any deep learning classification task with finer precision can use this method.

[0057] In addition, the full-process method proposed by the method of the present invention can also be applied to tasks such as analyzing operator emotions based on operator interaction data and identifying elderly operators, and only the collected labels need to be changed. This provides more possibilities and opens up more directions for the current field of human-computer interaction behavior analysis.

[0058] At the same time, the wavelet denoising technology can be used to remove the noise information caused by the acquisition error in the sensor while retaining the useful information of the original sensor, thereby improving the quality of the data. It helps the model learn higher-quality deep features, and has a better denoising effect than the currently commonly used median filtering method. The PDF method is used to sort the importance of the proposed manual features, and the feature combination with high importance is selected as the input of the final model. On the one hand, it reduces the memory consumption caused by data storage, and on the other hand, it allows the model to learn a better quality feature group. Furthermore, a dual-input child recognition model is constructed, by extracting the depth information from the sensor features and the manual sliding features respectively and splicing them into the final classifier. The depth information in the sliding data and the sensor data can be mined at the same time, which improves the recognition effect of the overall model. Furthermore, an auxiliary training method is used in the model training stage. The auxiliary training can obtain the loss value of finer-grained labels, and further combined with the main classification loss to guide the overall loss of the network to decrease, which can improve the learning efficiency of the overall model, optimize the training process of the model, and finally obtain a model with better classification performance.

[0059] The following are device embodiments of the present invention, which can be used to implement the method embodiments of the present invention. For details not disclosed in the device embodiments, please refer to the method embodiments of the present invention.

[0060] See also Figure 3 In another embodiment of the present invention, an operator identity type recognition system is provided, which can be used to implement the above-mentioned operator identity type recognition method. Specifically, the operator identity type recognition system includes a data acquisition module and a type recognition module.

[0061] Among them, the data acquisition module is used to obtain the operator's manual sliding data and the sensor data of the operation object; the type recognition module is used to obtain the operator's identity type based on the manual sliding data and the sensor data through a pre-trained operator identity type recognition model.

[0062] Among them, during the pre-training of the operator identity type recognition model, an auxiliary classification authentication module with finer granularity than the classification authentication module of the operator identity type recognition model is set, and pre-training is performed by combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module.

[0063] All relevant contents of each step involved in the embodiment of the operator identity type identification method described above can be referred to the functional description of the functional module corresponding to the operator identity type identification system in the embodiment of the present invention, and will not be repeated here.

[0064] The division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into one processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0065] In another embodiment of the present invention, a computer device is provided, the computer device comprising a processor and a memory, the memory being used to store a computer program, the computer program comprising program instructions, and the processor being used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in a computer storage medium to implement a corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the operator identity type identification method.

[0066] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the operator identity type identification method in the above embodiment.

[0067] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0068] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0069] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying an operator's identity type, characterized in that: include: Acquire manual sliding data of the operator and sensor data of the operation object; According to the manual sliding data and the sensor data, the operator's identity type is obtained through a pre-trained operator identity type recognition model; Among them, during the pre-training of the operator identity type recognition model, an auxiliary classification authentication module with finer granularity than the classification authentication module of the operator identity type recognition model is set, and pre-training is performed by combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module.

2. The operator identification type identification method according to claim 1, characterized in that: The operator's manual sliding data is obtained by classifying the sliding coordinates of the same finger number from appearance to disappearance in timestamp order; the sensor data of the operation object includes one or more of the following: screen direction flag, accelerometer data and gyroscope data.

3. The operator identification type identification method according to claim 1, characterized in that: Before obtaining the operator's identity type according to the manual sliding data and the sensor data through the pre-trained operator identity type recognition model, the method further includes: Traverse each manual sliding data, and only keep the first sliding coordinate at the beginning for multiple repeated sliding coordinates in continuous time, and delete the other repeated data; use wavelet basis function to remove noise in sensor data.

4. The operator identification type identification method according to claim 1, characterized in that: The operator identity type is obtained by using a pre-trained operator identity type recognition model based on the manual sliding data and the sensor data, including: According to the manual sliding data, based on a preset manual design feature group, manual sliding feature data is obtained; According to a preset sliding window, the sensor data is divided into a plurality of two-dimensional sensor window data, and the plurality of two-dimensional sensor window data are spliced ​​to obtain sensor feature data; Inputting the manual sliding feature data and the sensor feature data into a pre-trained operator identity type recognition model to obtain the operator's identity type; Among them, the manually designed feature group is obtained by the following method: Based on the physical quantities generated during the sliding process, the mathematical statistics of each physical quantity are calculated, and a number of manually designed features are obtained by combining the physical quantities and the mathematical statistics of each physical quantity; the physical quantities include one or more of the following: position, length, velocity, acceleration, angle, attitude angle and sensor sequence, and the mathematical statistics include one or more of the following: mean, variance, maximum value, minimum value, kurtosis and skewness; Based on the PDF analysis method, the distribution frequency of each manually designed feature of different identity types is taken as a distribution frequency vector, and the distance between the distribution frequency vectors of different identity types is calculated to obtain the discriminant value of each manually designed feature; The manually designed features corresponding to the discriminant values ​​with a large number of pre-set values ​​are selected to obtain a manually designed feature group.

5. The operator identification type identification method according to claim 1, characterized in that: The operator identity type recognition model includes a feature extraction module and a classification authentication module; The feature extraction module includes a convolution module, a fully connected module and a splicing module; the convolution module is used to extract data features of manual sliding data, the fully connected module is used to extract data features of sensor data, and the splicing module is used to splice data features of manual sliding data and data features of sensor data to obtain splicing features; The convolution module consists of a 2D convolutional layer, four residual blocks, and a fully connected layer connected in sequence; The fully connected module consists of a batch normalization layer and five fully connected layers connected in sequence; The classification authentication module consists of two fully connected layers connected sequentially.

6. The operator identification type identification method according to claim 1, characterized in that: The identity types of the classification authentication module of the operator identity type recognition model include children and adults; The identity type of the auxiliary classification authentication module includes four age groups divided according to age.

7. The operator identification method according to claim 1, characterized in that: The pre-training by combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module includes: Combined with the loss l of the classification authentication module main and the auxiliary loss l of the auxiliary classification and authentication module aux , the combined loss l of the operator identity type recognition model is obtained by the following formula total : l total =l main +αl aux Among them, α is a preset constant; Combination loss l of the identification model according to operator identity type total , pre-training the operator identity type recognition model; and after each number of trainings, adjusting the learning rate of the operator identity type recognition model to a previously preset multiple.

8. An operator identity type recognition system, characterized in that: include: A data acquisition module, used to acquire the operator's manual sliding data and the sensor data of the operation object; A type recognition module is used to obtain the operator's identity type based on the manual sliding data and the sensor data through a pre-trained operator identity type recognition model; Among them, during the pre-training of the operator identity type recognition model, an auxiliary classification authentication module with finer granularity than the classification authentication module of the operator identity type recognition model is set, and pre-training is performed by combining the loss of the classification authentication module and the auxiliary loss of the auxiliary classification authentication module.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the operator identity type identification method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the operator identity type identification method according to any one of claims 1 to 7 are implemented.

Citation Information

Cited By

  • Network access identity auditing method and system based on multi-modal data fusion

    CN122133085A

  • Network access identity auditing method and system based on multi-modal data fusion

    CN122133085B