Terminal location identification method and device, computer readable medium, and electronic device
By extracting the sensor data of the terminal device in time domain and frequency domain features, and using the trained position recognition model, the problem of difficulty in adapting to the differences between different users and terminals in the prior art is solved, and accurate terminal position and status recognition is achieved , improves user experience and product robustness.
Patent Information
- Application Number
- CN202210319172.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-29
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-03-29
AI Technical Summary
The prior art is difficult to adapt to the differences between different users and mobile terminals, which makes it difficult to accurately identify the location and status of the terminal device, resulting in poor user experience and poor product robustness.
By collecting sensor data of the terminal device, the time domain feature data and frequency domain feature data are obtained, and the trained position recognition model is input, and the probability value is calculated using multiple classifiers to determine the recognition result of the terminal position.
It realizes adaptive and accurate identification of the location and status of the terminal device, improves user experience and product robustness, and reduces power consumption, and supports real-time identification and hierarchical power consumption optimization strategies.
Smart Images

Figure CN114828208B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of mobile terminals, and in particular to a method for identifying a terminal position, a device for identifying a terminal position, a computer-readable medium, and an electronic device. Background Art
[0002] With the continuous development of Internet technology, in order to realize personalized services for users, more and more services and products can obtain the location and status of smart mobile terminals such as mobile phones and tablets through technical means, and adjust the operation mode of the terminal accordingly. For example, when a user's mobile phone is placed in a pocket or backpack for a long time, it is not often used. If it is still running in a high-performance model at this time, a lot of power will be wasted. The basic status obtained by detection and identification can support the upper layer of the application to make decisions and judgments to optimize more for users. In the existing technical solutions, most of them identify the physical state of the terminal device by collecting sensor parameters and use pre-set thresholds for judgment; but often different users and different mobile terminals have differences, which creates a serious technical problem, that is, the method of using threshold judgment is difficult to adapt to most users. Therefore, such technical pain points will cause the function to take effect for some people and fail for some people, and will also cause the function to take effect and fail for some users. The time is uncertain. Therefore, many problems such as poor user experience and poor product robustness will arise.
[0003] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention
[0004] The present disclosure provides a method for identifying a terminal location, a device for identifying a terminal location, a computer-readable medium, and an electronic device, which can adaptively and accurately identify the location and status of a terminal device.
[0005] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.
[0006] According to a first aspect of the present disclosure, a method for identifying a terminal location is provided, the method comprising:
[0007] Collecting sensor data of a target sensor; wherein the target sensor includes at least one type of sensor equipped on the terminal device;
[0008] Parsing the sensor data to obtain feature data corresponding to the target sensor; wherein the feature data includes time domain feature data and frequency domain feature data;
[0009] Inputting the feature data into a trained location recognition model, and using each classifier in the location recognition model to calculate the corresponding probability value; wherein the number of classifiers in the location recognition model is the same as the number of predefined terminal location categories;
[0010] The identification result of the terminal position is determined according to the comparison result of the probability values output by each of the classifiers.
[0011] According to a second aspect of the present disclosure, there is provided a device for identifying a terminal position, comprising:
[0012] A data acquisition module, used to acquire sensor data of a target sensor; wherein the target sensor includes at least one type of sensor equipped on the terminal device;
[0013] A data analysis module, used to analyze the sensor data to obtain feature data corresponding to the target sensor; wherein the feature data includes time domain feature data and frequency domain feature data;
[0014] A model processing module, used for inputting the feature data into a trained location recognition model, and using each classifier in the location recognition model to calculate the corresponding probability value; wherein the number of classifiers in the location recognition model is the same as the number of predefined location categories of the terminal;
[0015] The recognition result output module is used to determine the recognition result of the terminal position according to the comparison result of the probability values output by each of the classifiers.
[0016] According to a third aspect of the present disclosure, a computer-readable medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned method for identifying a terminal location is implemented.
[0017] According to a fourth aspect of the present disclosure, there is provided an electronic device, including:
[0018] Processor; and
[0019] A memory, configured to store executable instructions of the processor;
[0020] Wherein, the processor is configured to implement the above-mentioned terminal location identification method by executing the executable instructions.
[0021] The terminal location recognition method provided by an embodiment of the present disclosure collects sensor data of a target sensor, and parses the sensor data to extract corresponding time domain feature data and frequency domain feature data, thereby using these time domain and frequency domain feature data to identify the location and state of the terminal device, and determine which predefined location and state the current terminal device is in, thereby adaptively identifying the location and state of the terminal device. In addition, by pre-training a location recognition model, using the time domain feature data and the frequency domain feature data as inputs to the model, and using multiple classifiers of the model to calculate and input the results respectively, a more accurate recognition result can be obtained; at the same time, the recognition result can be obtained in real time, and power consumption can be reduced.
[0022] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.
[0024] Figure 1 A schematic diagram schematically illustrates a method for identifying a terminal location in an exemplary embodiment of the present disclosure;
[0025] In Figure 2, Figure 2-1 , 2-2 , 2-3, and 2-4 schematically illustrate sensor data diagrams of a mobile phone in a handheld state, a static state, a stand state, and a pocket or backpack state in an exemplary embodiment of the present disclosure, respectively;
[0026] Figure 3 A schematic diagram schematically illustrates a method for training a location recognition model in an exemplary embodiment of the present disclosure;
[0027] Figure 4 A schematic diagram schematically illustrates a method for collecting sensor data in an exemplary embodiment of the present disclosure;
[0028] Figure 5 A schematic diagram schematically illustrates a method for parsing sensor data in an exemplary embodiment of the present disclosure;
[0029] Figure 6 A schematic diagram schematically shows the composition of a terminal location identification device in an exemplary embodiment of the present disclosure;
[0030] Figure 7The diagram schematically shows the components of an electronic device in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION
[0031] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
[0032] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.
[0033] In the related art, when the terminal device obtains the corresponding state, it generally collects sensor data and uses various sensors to make threshold judgments. However, such judgment methods often differ due to different users and different mobile terminals; this causes a serious technical problem, that is, the threshold judgment scheme is difficult to adapt to most users, so that this scheme will cause the function to take effect for some people and fail for some people, and will also cause the function to take effect and fail for some users. The time is uncertain. Therefore, there will be many problems such as poor user experience and poor product robustness. At present, most technologies read sensors at the upper layer of the application, which will bring very large energy consumption to the mobile terminal, making the positive impact of such a solution smaller than the negative impact, and ultimately difficult to bring convenience to the user. Then the upper-layer application and design based on this technology are castles in the air. Therefore, it is impossible to bring real-time detection, and it is impossible to achieve the statistical maintenance time of the mobile terminal in each state, so it is difficult to ensure the optimization strategy of layering and grading.
[0034] In view of the above-mentioned shortcomings and deficiencies of the prior art, this exemplary embodiment provides a method for identifying the location of a terminal, which can be applied to smart mobile terminal devices such as mobile phones and tablet computers to achieve accurate identification of the location of the terminal. Figure 1 As shown in , the above-mentioned method for identifying the terminal location may include:
[0035] Step S11, collecting sensor data of a target sensor; wherein the target sensor includes at least one type of sensor installed in the terminal device;
[0036] Step S12, parsing the sensor data to obtain feature data corresponding to the target sensor; wherein the feature data includes time domain feature data and frequency domain feature data;
[0037] Step S13, inputting the feature data into a trained location recognition model, and using each classifier in the location recognition model to calculate the corresponding probability value; wherein the number of classifiers in the location recognition model is the same as the number of predefined terminal location categories;
[0038] Step S14, determining the recognition result of the terminal location according to the comparison result of the probability values output by each of the classifiers.
[0039] The terminal location identification method provided in this example implementation method predefines the location type of the terminal device and trains the location identification model. After collecting the sensor data of the target sensor, it is analyzed to obtain the corresponding time domain feature data and frequency domain feature data; these feature data are input as input parameters into the trained location identification model, and the corresponding probability values are calculated by each classifier, thereby determining the location identification result, and adaptively identifying the location and state of the terminal device. In addition, by using multiple classifiers of the model to calculate and input the results respectively, more accurate identification results can be obtained; at the same time, real-time identification results can be obtained. By accurately identifying the placement position and stay time of the centering device, a hierarchical suggestion system can be used to shut down or optimize different power-consuming functions to extend the battery life of the mobile phone. Efficiently and accurately reduce power consumption.
[0040] Below, each step of the method for identifying the terminal location in this example implementation will be described in more detail with reference to the accompanying drawings and embodiments.
[0041] In this exemplary implementation, the above method can be applied to smart mobile terminal devices such as mobile phones, tablet computers, smart watches, smart bracelets, etc. In the following embodiments, the method of the present disclosure is described by taking the smart mobile terminal device as a mobile phone as an example.
[0042] In this example implementation, the above-mentioned method for identifying the terminal location may include:
[0043] Step S10, pre-training the location recognition model.
[0044] Specifically, the location type of the terminal device can be defined first, that is, the various state types of the terminal device that need to be detected and identified can be determined. For example, the states of the mobile phone can be pre-defined to include: handheld state, pocket backpack state, static state, and bracket state, four states. Among them, the above-mentioned handheld state can be the state of the user holding the terminal device, including but not limited to the terminal device in horizontal screen state, vertical screen state, screen-on state, screen-off state, etc. In the above states, it includes but is not limited to the user's daily life states such as standing, lying on the side, lying flat, walking, running, riding, and driving. The above-mentioned pocket backpack state can refer to the terminal device being placed in the pocket of the user's clothing or backpack; including but not limited to the mobile phone being located in the upper pocket, lower pocket, front pocket, back pocket, backpack, messenger bag, handbag, etc. The above-mentioned static state refers to the terminal device being placed on a plane or inclined surface, including but not limited to being placed at any angle, such as a desktop, the ground, etc. The above-mentioned bracket state refers to the terminal device being placed on a mobile phone bracket, including but not limited to the positions of a car bracket, a motorcycle bracket, a bicycle bracket, a desktop bracket, etc. for mobile phones and tablets. Among them, in the above-mentioned pocket backpack state, static state, and stand state, the screen usage state of the terminal device can also be distinguished between the screen-on state and the screen-off state.
[0045] Generally speaking, there are certain differences in the sensor data of terminal devices in different states; Figure 2-1 to Figure 2-4 As shown, the sensor data of the mobile phone in the handheld state, static state, bracket state, pocket or backpack have different data in the x-axis, y-axis and z-axis directions.
[0046] In this example implementation, reference Figure 3 As shown, the above-mentioned pre-training of the location recognition model may include:
[0047] Step S101, collecting raw sensor data from a number of different terminal devices, and preprocessing the raw sensor data to construct a raw data set; wherein the raw sensor data is used to describe the location type of the terminal device;
[0048] Step S102, parsing the original data set to extract corresponding time domain feature data and frequency domain feature data;
[0049] Step S103, constructing a feature vector using the time domain feature data and the frequency domain feature data, and constructing sample data based on the feature vector and corresponding label information;
[0050] Step S104, using the sample data to input each classifier in the location recognition model for training, and training each classifier until convergence to obtain the location recognition model.
[0051] In this example implementation, in the above step S101, specifically, when training the position recognition model, the original sensor data of multiple terminal devices of different types, brands, and models can be collected. Through these collected sensor data, the placement and status of the terminal can be accurately described and characterized. Among them, the specific collected sensors can at least be the acceleration sensor, gyroscope sensor, and screen status sensor installed on the terminal device; wherein the screen status sensor is used to identify the screen-on or screen-off status of the terminal device, and the original sensor data corresponding to these sensors can be used as the basic data for constructing the original data set. In addition, the game rotation vector sensor can also be collected as an auxiliary sensor to provide auxiliary sensor information, which is conducive to obtaining more accurate recognition results and obtaining a more efficient and effective model. When collecting the above-mentioned raw sensor data, the state label of the terminal device can also be included; that is, the sample label of one of the above four states.
[0052] Of course, in some other exemplary embodiments, the collected raw sensor data may also include one or more of a light sensor, a distance sensor, a rotation vector sensor, a magnetic sensor, and a gravity sensor. The raw sensor data corresponding to one or more of these sensors may be used as basic data or auxiliary data for constructing the raw data set.
[0053] In addition, the acquired raw sensor data may be preprocessed. Specifically, the preprocessing may include data cleaning of the raw sensor data, including but not limited to one or more of null value filling, nan value removal, duplicate value removal, outlier filtering, etc., to finally form a valid non-empty raw data set.
[0054] In this example implementation, in the above step S102, for each sensor, its acquisition frequency F can be preconfigured in Hz. For different sensors, the value range of the corresponding sensor data can be determined according to the maximum and minimum values of the specific sensor obtained by the specific device.
[0055] For the original sensor data, the time domain related features of each sensor include but are not limited to maximum value, minimum value, average value, standard deviation, root mean square, mean absolute difference, amplitude average value, amplitude standard deviation, number of zero crossings, amplitude mean value, amplitude root mean square, average amplitude absolute difference, time entropy, etc.
[0056] For example, the method for calculating the maximum value in any feature may include:
[0057] SensorNameXmean=mean(X 1 , X 2 ......X N )
[0058] SensorNameYmean=mean(Y 1 , Y 2 ......Y N )
[0059] SensorNameZmean=mean(Z 1 , Z 2 ......Z N )
[0060] Where SensorName is the sensor name, and different sensors have different names; each sensor contains parameters in the X, Y, and Z dimensions; and N is the number of sensor data samples. For example, the average value of N data in a time window on the X axis of an acceleration sensor.
[0061] The standard deviation calculation method in the corresponding feature can include:
[0062]
[0063]
[0064]
[0065] Where SensorName is the sensor name, and each sensor contains parameters in the X, Y, and Z dimensions; N is the number of sensor data samples. For example, the standard deviation of N data in a time window on the X axis of an acceleration sensor.
[0066] The root mean square calculation method in the corresponding feature may include:
[0067]
[0068]
[0069]
[0070] Among them, SensorName is the sensor name, and each sensor contains parameters in the X, Y, and Z dimensions.
[0071] The mean absolute difference calculation method in the corresponding features may include:
[0072]
[0073]
[0074]
[0075] Among them, X i+1Yes X i The value of the X dimension in the i+1th data in the time series; the same applies to the others.
[0076] The amplitude average calculation method in the corresponding feature may include:
[0077] SensorNameX_mag_means=mean(|X 1 |,|X 2 |......|X N |)
[0078] SensorNameY_mag_means=mean(|Y 1 |,|Y 2 |......|Y N |)
[0079] SensorNameZ_mag_means=mean(|Z 1 |,|Z 2 |......|Z N |)
[0080] Among them, SensorName is the name of the sensor, and each sensor contains parameters of the X, Y, and Z dimensions. As shown in the above formula, the amplitude average is obtained by taking the absolute value of each dimension and then calculating the overall average.
[0081] Methods for calculating the standard deviation of amplitudes in corresponding features may include:
[0082]
[0083]
[0084]
[0085] Among them, SensorName is the sensor name, and each sensor contains parameters in the X, Y, and Z dimensions.
[0086] The method for calculating the number of zero crossings in the corresponding feature may include:
[0087]
[0088]
[0089]
[0090] Among them, i∈(1, 2, 3...N), SensorName is the sensor name, and each sensor contains parameters of X, Y, and Z dimensions. By traversing the N data in the time window of each dimension, if the current data is greater than 0 and the next data is less than 0, the number of zero crossings increases by 1. Conversely, if the current data is less than 0 and the next data is greater than 0, the number of zero crossings also increases by 1.
[0091] The amplitude mean calculation method in the corresponding feature includes:
[0092]
[0093] Among them, SensorName is the name of the sensor, and each sensor contains parameters of X, Y, and Z dimensions; the amplitude mean can be the embodiment of calculating the total amplitude of the 3D data.
[0094] The calculation methods of the amplitude standard deviation in the corresponding features include:
[0095]
[0096]
[0097]
[0098] Among them, SensorName is the sensor name, mag_mean represents the amplitude mean, mag i Represents the amplitude of the i-th data.
[0099] The RMS calculation methods for the corresponding features include:
[0100]
[0101] Among them, mag i Represents the amplitude of the i-th data.
[0102] The average absolute difference in magnitude among the corresponding features is calculated by:
[0103]
[0104] Among them, mag i Represents the amplitude of the i-th data.
[0105] The above time window may be expressed in the minimum unit time. For example, the time window may be 500ms, 1s, 2s, etc.
[0106] In this example implementation, in the above step S102, the relevant frequency domain features of the sensor may include five fast Fourier transform modulus coefficients and spectrum energy; wherein the spectrum energy may be calculated by taking the square root of the sum of the squares of the spectrum coefficients and dividing it by the number of samples. Specifically, the calculation method of the five fast Fourier transform modulus coefficients of the sensor may include:
[0107] N_FFT = 256
[0108]
[0109] SensorName_fft1=|FFT[mag i ]| 1
[0110] SensorName_fft2=|FFT[mag i ]| 2
[0111] SensorName_fft3=|FFT[mag i ]| 3
[0112] SensorName_fft4=|FFT[mag i ]| 4
[0113] SensorName_fft5=|FFT[mag i ]| 5
[0114] Among them, mag i Represents the amplitude of the i-th data. After Fourier transformation is performed on the amplitude, the Fourier series (the coefficients of the Fourier expansion) is obtained. Here, only the first 5 items are taken, which represent the main energy components.
[0115] The calculation method of the spectrum energy of the sensor may include:
[0116] B=|FFT[mag i ]
[0117]
[0118] Where B represents the value after Fourier transformation of the amplitude, and N_FFT is the size of the Fourier window, which is usually set to 256.
[0119] In this example implementation, based on the above-mentioned definition of frequency domain feature data and time domain feature data, the acquired original data set can be parsed, and the time domain features and frequency domain features of each original feature data can be extracted. The extracted time domain features can be defined as: Td 1 , Td 2 , Td 3 ...Td n ; The extracted frequency domain features can be defined as: Sd 1 , Sd 2 , Sd 3 ...Sd m ; Where n is the number of features in the time threshold related dimension, and m is the number of features in the frequency domain related dimension.
[0120] Specifically, the above method may also include: the number of dimensions of the time domain feature data and frequency domain feature data collected by different sensors installed on the terminal device is the same. For example, for the sensor data of any sensor, n+m time domain and frequency domain features can be parsed and obtained; for k sensors on the terminal device, k*(n+m) features can be parsed and obtained.
[0121] In this example implementation, in the above step S103, after extracting the time domain feature data and the frequency domain feature data, the time domain feature data and the frequency domain feature data of a specified type are selected from the extracted time domain feature data and the frequency domain feature data according to a preset rule to construct a feature vector.
[0122] For example, for an acceleration sensor and a gyroscope sensor, n-dimensional time domain features and m-dimensional frequency domain features can be extracted respectively. For example, among the time domain features and frequency domain features listed above, the maximum value, minimum value, average value, standard deviation, root mean square and mean absolute difference can be selected as time domain features; in addition, spectrum energy, Fourier coefficient 1, Fourier coefficient 2 and Fourier coefficient 3 can be selected as frequency domain features. In addition, these features can be used as basic features, and the remaining time domain features and frequency domain features can be used as auxiliary features. For example, when constructing a feature vector, only the basic features can be selected to construct the feature vector. Alternatively, on the basis of the basic features, at least one auxiliary feature can be added to construct the feature vector, so as to achieve balanced computing performance and meet generalization requirements. For example, on the basis of selecting 6 time domain features and 4 frequency domain features as basic features, two auxiliary frequency domain features are added, such as Fourier coefficient 4 and Fourier coefficient 5, so as to construct a feature vector using 6 time domain features and 6 frequency domain features. That is, for constructing the feature vector, only basic features can be used, or basic features combined with auxiliary features can be used.
[0123] In addition, for the time domain features and frequency domain features selected for constructing the feature vector, each feature can be normalized separately. Through normalization, the feature data can be standardized. For example, Z-score0 mean normalization, Min-Max linear function normalization and other methods can be used for normalization operations. After the specific values of each time domain feature and frequency domain feature are normalized, these features can be used to construct a feature vector; the feature vector can be used to describe the placement state of the terminal device, such as one of the four predefined states mentioned above.
[0124] The feature vector constructed based on the time domain features and frequency domain features is combined with the label information of the corresponding terminal state to construct sample data as training samples for model training.
[0125] In addition, in some exemplary embodiments, the above method further includes: extracting time domain feature data and frequency domain feature data of a specified number and / or dimension according to a preset weight strategy to construct the feature vector.
[0126] For example, the features of the three dimensions x, y, and z can be selected in a weighted manner. For example, the weight strategy is 0.3Td+0.7Sd, and the weight strategy can be a weight distribution of the number of features; then based on the weight strategy, a corresponding number of time domain features and frequency domain features can be selected to form a new feature vector. Alternatively, the weight strategy can also be a weight distribution of numerical values. For example, when constructing a feature vector using the above-mentioned basic features, the coefficients of the numerical values of the time domain features and the frequency domain features can be configured respectively according to the weight distribution strategy, and the feature vector can be constructed using the numerical values calculated according to the coefficients.
[0127] In some exemplary embodiments, the method may further include: extracting a preset number of features according to preset rules to construct an auxiliary feature vector; the auxiliary feature vector is used to describe the user's usage habits of the terminal device; and using the auxiliary feature vector as sample data.
[0128] For example, all the time domain features and frequency domain features defined above can be selected to construct a feature vector. In addition, i features can be randomly extracted to form a description of the user's usage behavior habits of the terminal. The usage behavior habits may refer to the subtle differences in the placement of the mobile phone to be identified during the use of the terminal by different users. For example, the randomly extracted features may be the basic features of the time domain features and frequency domain features mentioned above, thereby ensuring the final recognition effect. For example, the auxiliary feature vector may be N={Td 1 , Sd 3 , Td 5 , Sd 3Alternatively, a certain number of time domain features and frequency domain features may be selected from the basic features in the time domain features and the frequency domain features according to a preset weight allocation rule to construct an auxiliary feature vector.
[0129] In some exemplary embodiments, the constructed sample data may be composed of a feature vector and label information; or may be composed of a feature vector, an auxiliary feature vector, and label information. The label information may be a label value; for example, the label of the handheld state is 1, the label of the pocket or backpack state is 2, the label of the static state is 3, and the label of the stand state is 4.
[0130] In this example implementation, in the above-mentioned step S104, the initial location recognition model can be a multi-algorithm model, including 4 classifiers, each classifier corresponding to a predefined terminal state type. Among them, each classifier can adopt a different classification algorithm. For example, the classification algorithm adopted can be a random forest model, an XGBoost model, an LR (Logistic Regression) model, and a 1D-CNN (one-dimensional convolutional neural network) model. For example, in the location recognition model, the first classifier corresponds to the recognition of the handheld state, which is constructed based on the random forest model; the second classifier corresponds to the recognition of the pocket backpack state, which is constructed based on the XGBoost model; the third classifier corresponds to the recognition of the silent state, which is constructed based on the LR model; the fourth classifier corresponds to the recognition of the bracket state, which is constructed based on the 1D-CNN model. Of course, in other exemplary embodiments of the present disclosure, each classifier can also be constructed using other classification models.
[0131] For each classifier in the model, the sample data used for model training is the same. For example, a certain number of time domain features and frequency domain features are randomly selected from the basic features to construct a feature vector, and the composition of the feature vector includes: Fe = {Td (acceleration X-axis average value), Td (gyroscope Y-axis zero crossing times), Sd (gyroscope spectrum energy), Sd (gyroscope FFT coefficient)}.
[0132] Taking the logistic regression classifier as an example, the model formula includes:
[0133] h θ (x) = g(θ 0 +θ 1 x 1 +θ 2 x 2 +…+θ n x n )
[0134] The loss function of the model can include:
[0135]
[0136] Among them, y i represents the correct label value; m is the number of training samples; h θ (x i ) represents the probability value of the sample predicted by the model; x n represents the eigenvector; θ represents the coefficient.
[0137] As shown in the above formula, the classifier built based on the LR model is characterized by the g function. For different classifiers, four different g functions can be obtained, including: θ g1 (θ 0 ....θ n ), θ g2 (θ 0 ....θ n ), θ g3 (θ 0 ....θ n ), θ g4 (θ 0 ....θ n ).
[0138] For each classifier of the location recognition model, the sample data constructed above can be used as the input parameter of each classifier, and each classifier is used for calculation; the output parameter of each classifier is the probability value of the sample belonging to the corresponding category; based on the label of the sample, the loss function is used to feedback train each classifier model until the value of the loss function is less than a certain set value, and the coefficients θ(θ 0 ....θ n ), we can get the trained model, which is the setting function with known coefficients.
[0139] In some exemplary embodiments, the above-mentioned model training method can be executed on the terminal device side or the server side. After the model training is completed, it can be run on smart terminal devices such as mobile phones, tablet computers, smart watches, etc.
[0140] In step S11 , sensor data of a target sensor is collected; wherein the target sensor includes at least one type of sensor equipped on the terminal device.
[0141] In this example implementation, reference Figure 4 As shown, the above step S11 may include:
[0142] Step S111, collecting sensor data of each sensor installed in the terminal device according to a preset data collection cycle;
[0143] Step S112: determining the target sensor according to the current power consumption allocation strategy of the terminal device, so as to determine the sensor data to be calculated according to the determined target sensor.
[0144] Specifically, the terminal device can collect sensor data according to a preset data collection cycle; or, it can also collect data according to a predefined sensor data collection frequency. For the terminal device, sensor data corresponding to all sensors in the terminal device can be collected. At the same time, the power consumption allocation strategy of the terminal device for the location recognition model can be identified. Specifically, the power consumption allocation strategy of the terminal device for the location recognition model can include a first power consumption strategy for selecting some sensors and a second power consumption strategy for selecting all sensors. When the second power consumption strategy is executed, all predefined sensors can be used as target sensors, and sensor data of each target sensor can be collected. Alternatively, when the first power consumption strategy is executed, some predefined sensors are selected as target sensors; or, when the first power consumption strategy is executed, some predefined sensors can also be selected as auxiliary sensors. For example, when the first power consumption strategy is executed, an acceleration sensor, a gyroscope sensor, and a mobile phone screen status sensor are selected as target sensors to provide basic sensor data; at the same time, a game rotation vector sensor can be selected as an auxiliary sensor to provide auxiliary sensor data. For the terminal device, the power consumption configuration strategy can be selected according to the current occupancy of the processor and memory.
[0145] Alternatively, in some exemplary embodiments, the terminal device may first determine the current power consumption strategy, then determine the target sensor, and then collect sensor data of the target sensor.
[0146] In step S12, the sensor data is parsed to obtain feature data corresponding to the target sensor; wherein the feature data includes time domain feature data and frequency domain feature data.
[0147] In this example implementation, after obtaining the sensor data corresponding to the target sensor, it can be parsed to extract the time domain feature data and frequency domain feature data corresponding to each sensor. Among them, the current sensor data can be parsed according to the time domain features and frequency domain features selected during the location recognition model training to calculate the same time domain feature data and frequency domain feature data.
[0148] Specifically, refer to Figure 5 As shown, the above step S12 may include:
[0149] Step S121, preprocessing the sensor data of each target sensor;
[0150] Step S122, extracting n-dimensional time domain feature data in the time domain and m-dimensional frequency domain feature data in the frequency domain from each of the preprocessed sensor data; wherein n and m are positive integers;
[0151] Step S123, normalizing the extracted time domain feature data and frequency domain feature data; and
[0152] Step S124 , extracting time domain feature data and frequency domain feature data of a specified number and dimension from the normalized time domain feature data and frequency domain feature data according to a preset weight strategy, so as to construct the feature data.
[0153] Specifically, the currently collected sensor data can be preprocessed, including one or more of the following: filling in null values, removing nan values, removing duplicate values, and filtering outliers. After the preprocessing is completed, the time domain features and frequency domain features of the sensor data corresponding to each target sensor can be calculated and extracted. For each target sensor, n-dimensional time domain feature data and m-dimensional frequency domain feature data can be extracted. For example, time domain features may include: maximum value, minimum value, average value, standard deviation, root mean square, and mean absolute difference; frequency domain features may include: spectral energy, Fourier coefficient 1, Fourier coefficient 2, and Fourier coefficient 3.
[0154] After the time domain features and frequency domain features are calculated, the feature data can be normalized. Then, the same weight strategy as in model training can be used to select the time domain features and frequency domain features and construct the feature vector. For example, the weight strategy can be the weight distribution between the time domain features and the frequency domain features, such as: 0.3Td+0.7Sd; or, 0.6Td+0.4Sd; where Td is the number of time domain features and Sd is the number of frequency domain features.
[0155] In some exemplary embodiments, for the normalized time domain data and frequency domain data, a certain number of features may be randomly extracted to construct a user usage behavior vector. For example, two features may be randomly extracted from the time domain features and two features may be randomly extracted from the frequency domain features to construct a user usage behavior vector. The user usage behavior vector is combined with the feature vector constructed above to form feature data as an input parameter of the model.
[0156] In step S13, the feature data is input into a trained location recognition model, and the corresponding probability values are calculated using each classifier in the location recognition model; wherein the number of classifiers in the location recognition model is the same as the number of predefined terminal location categories.
[0157] In this example implementation, the location recognition model may include four classifiers, each of which corresponds to a predefined terminal state type. Among them, the first classifier may correspond to the identification of the handheld state; the second classifier may correspond to the identification of the pocket backpack state; the third classifier may correspond to the identification of the silent state; and the fourth classifier may correspond to the identification of the stand state.
[0158] After constructing the feature data, it can be used as input data and input into the four classifiers in the trained location recognition model. For the four classifiers, the input feature vectors are the same. And the probability values output by each classifier are obtained respectively.
[0159] In step S14, the recognition result of the terminal location is determined according to the comparison result of the probability values output by each of the classifiers.
[0160] In this example implementation, after obtaining the probability values output by each classifier, the four probability values can be compared, and the result with the largest probability value is selected, and the type of classifier corresponding to the probability value is used as the current position recognition result of the terminal device. For example, when the probability value output by the fourth classification is the largest, the current position recognition result of the terminal device is the stand state.
[0161] In addition, in some exemplary embodiments, if there are two or three identical probability values, the user usage behavior vector constructed above can be input into the location recognition model, and judgment can be made based on the probability values output by each classifier, and the result with the largest probability value can be selected as the recognition result of the current location of the terminal device. For example, if the results output by the first classifier and the fourth classifier are the same, that is, the result of the first judgment is that the terminal device may be in a handheld state or a stand state; the constructed user usage behavior vector can be input into the model for another judgment, and the model can only use the first classifier and the fourth classifier for calculation, and output the probability values calculated by the two classifiers, and select the largest probability value as the final recognition result; for example, if the probability value output by the fourth classifier is greater than the probability value output by the first classifier, then the current location recognition result of the terminal device is judged to be a handheld state.
[0162] Based on the above content, in some exemplary embodiments of the present disclosure, the above method may also include obtaining a plurality of continuous recognition results within a preset time length, determining the duration of the terminal device staying in a certain state according to the plurality of continuous recognition results; and then determining a power consumption optimization strategy for the terminal device according to the duration of the stay. Among them, the power consumption optimization strategy may be a strategy for the terminal device system to hierarchically shut down or optimize different power-consuming functions according to the different durations of stay in a certain state; thereby extending the battery life of the mobile phone and optimizing the power consumption of the terminal device.
[0163] After determining the identification result of the terminal device's location and the length of time the terminal device remains in the location and state, the corresponding power consumption management strategy can be configured.
[0164] The terminal location recognition method provided by the embodiment of the present disclosure can be applied to smart mobile terminal devices such as mobile phones, tablet computers, smart watches, and smart bracelets. Using the pre-trained location recognition model, the terminal device can be judged to be currently in one of the predefined positions, such as the above-mentioned handheld state, bracket state, pocket backpack state, or silent state, according to the time domain features and frequency domain features obtained after the sensor data is parsed. By using the frequency domain features and time domain features corresponding to the sensor data for judgment, data processing based on the underlying multi-sensor data is realized, and the problem that the world mobile phone position and state cannot be accurately determined in the prior art is solved from the feature dimension. The stability and accuracy of the recognition state are guaranteed by using the hierarchical progressive reasoning method. Compared with the method of using sensor data threshold judgment in the prior art, this solution is more robust, the obtained state is more accurate, and there are more sub-states, which is further improved than the previous pocket mode. In addition, the recognition model of the present application can be integrated into the underlying coprocessor, so that the power consumption of the algorithm startup is lower, so that the effect of Always On can be achieved, which makes the application scenarios of the present solution richer, and the results can be returned in time at any time.
[0165] It should be noted that the above figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.
[0166] For further reference, Figure 6 As shown, in the embodiment of this example, a terminal location recognition device 60 is also provided, and the device includes: a data acquisition module 601, a data analysis module 602, a model processing module 603, and a recognition result output module 604. Among them,
[0167] The data acquisition module 601 may be used to acquire sensor data of a target sensor; wherein the target sensor includes at least one type of sensor equipped in a terminal device.
[0168] The data analysis module 602 may be used to analyze the sensor data to obtain feature data corresponding to the target sensor; wherein the feature data includes time domain feature data and frequency domain feature data.
[0169] The model processing module 603 can be used to input the feature data into a trained location recognition model, and use each classifier in the location recognition model to calculate the corresponding probability value; wherein the number of classifiers in the location recognition model is the same as the number of pre-defined terminal location categories.
[0170] The recognition result output module 604 may be used to determine the recognition result of the terminal location according to the comparison result of the probability values output by each of the classifiers.
[0171] In some exemplary embodiments, the data acquisition module 601 may include: collecting sensor data of each sensor installed on the terminal device according to a preset data acquisition cycle; determining the target sensor according to the current power consumption allocation strategy of the terminal device, so as to determine the sensor data to be calculated according to the determined target sensor.
[0172] In some exemplary embodiments, the data analysis module 602 may include: preprocessing the sensor data of each of the target sensors; extracting time domain feature data of n dimensions in the time domain and extracting frequency domain feature data of m dimensions in the frequency domain for each of the preprocessed sensor data; wherein n and m are positive integers; normalizing the extracted time domain feature data and frequency domain feature data; and extracting time domain feature data and frequency domain feature data of a specified number and dimension from the normalized time domain feature data and frequency domain feature data according to a preset weight strategy to construct the feature data.
[0173] In some exemplary embodiments, the feature vectors input to each classifier in the location recognition model are the same.
[0174] In some exemplary embodiments, the apparatus further comprises: a model training module.
[0175] The model training module may include:
[0176] A raw sensor data acquisition module, used to collect raw sensor data from a number of different terminal devices, and pre-process the raw sensor data to construct a raw data set; wherein the raw sensor data is used to describe the location type of the terminal device;
[0177] A feature data extraction module, used to parse the original data set to extract corresponding time domain feature data and frequency domain feature data;
[0178] A sample data construction module, used to construct a feature vector using the time domain feature data and the frequency domain feature data, and to construct sample data based on the feature vector and corresponding label information;
[0179] The training execution module is used to use the sample data to input each classifier in the location recognition model for training, and train each classifier until convergence to obtain the location recognition model.
[0180] In some exemplary embodiments, the apparatus further comprises: a feature selection module.
[0181] The feature selection module can be used to construct a feature vector using the time domain feature data and frequency domain feature data. From the extracted time domain feature data and frequency domain feature data, a specified type of time domain feature data and frequency domain feature data are selected according to a preset rule to construct a feature vector.
[0182] In some exemplary embodiments, the apparatus further comprises: a feature weight configuration module.
[0183] The feature weight configuration module can be used to extract time domain feature data and frequency domain feature data of a specified number and / or dimension according to a preset weight strategy when constructing a feature vector using the time domain feature data and the frequency domain feature data to construct the feature vector.
[0184] In some exemplary embodiments, different sensors equipped on the terminal device have the same number of dimensions of the corresponding time domain feature data and frequency domain feature data collected.
[0185] In some exemplary embodiments, the apparatus further comprises: an auxiliary feature vector construction module.
[0186] The auxiliary feature vector construction module can be used to extract a preset number of features according to preset rules to construct an auxiliary feature vector; the auxiliary feature vector is used to describe the user's usage habits of the terminal device; and the auxiliary feature vector is used as sample data.
[0187] In some exemplary embodiments, the apparatus further comprises: a location type definition module.
[0188] The location type definition module is used to predefine the location type of the terminal device; wherein the location type includes at least one type.
[0189] The specific details of each module in the above-mentioned terminal position identification device 60 have been described in detail in the corresponding terminal position identification method, so they will not be repeated here.
[0190] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.
[0191] Figure 7 A schematic diagram of an electronic device suitable for implementing an embodiment of the present invention is shown.
[0192] It should be noted that Figure 7 The electronic device 1000 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.
[0193] like Figure 7 As shown, the electronic device 1000 includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage part 1008 to the random access memory (RAM) 1003. In the RAM 1003, various programs and data required for system operation are also stored. The CPU 1001, the ROM 1002 and the RAM 1003 are connected to each other through a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.
[0194] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, etc.; an output section 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed, so that a computer program read therefrom is installed into the storage section 1008 as needed.
[0195] In particular, according to an embodiment of the present invention, the process described below with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication part 1009, and / or installed from a removable medium 1011. When the computer program is executed by a central processing unit (CPU) 1001, various functions defined in the system of the present application are executed.
[0196] Specifically, the electronic device may be a smart mobile terminal device such as a mobile phone, a tablet computer or a laptop computer, or may be a smart terminal device such as a desktop computer.
[0197] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.
[0198] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0199] The units involved in the embodiments of the present invention may be implemented by software or hardware, and the units described may also be arranged in a processor. The names of these units do not, in some cases, limit the units themselves.
[0200] It should be noted that, as another aspect, the present application also provides a computer-readable medium, which may be included in an electronic device; or may exist independently without being installed in the electronic device. The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by an electronic device, the electronic device implements the method described in the following embodiments. For example, the electronic device may implement the following Figure 1 The steps shown.
[0201] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.
[0202] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.
[0203] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims
1. A method for identifying a terminal location, characterized in that: Applied to a terminal device, the method comprises: Collecting sensor data of a target sensor; wherein the target sensor includes at least one type of sensor equipped on the terminal device; Parsing the sensor data to obtain feature data corresponding to the target sensor; wherein the feature data includes time domain feature data and frequency domain feature data; Input the feature data into a trained location recognition model, and use each classifier in the location recognition model to calculate the corresponding probability value; wherein the number of classifiers in the location recognition model is the same as the number of pre-defined terminal location categories, and the feature vectors input to each classifier in the location recognition model are the same; Determining the recognition result of the terminal position according to the comparison result of the probability values output by each of the classifiers; Acquire a plurality of continuous recognition results within a preset time period, and determine the duration of a terminal device staying in a certain state according to the plurality of continuous recognition results; A power consumption optimization strategy for the terminal device is determined according to the stay duration.
2. The method for identifying the terminal position according to claim 1, characterized in that: The collecting of sensor data of the target sensor includes: Collect sensor data of each sensor installed in the terminal device according to a preset data collection cycle; The target sensor is determined according to the current power consumption allocation strategy of the terminal device, so as to determine the sensor data to be calculated according to the determined target sensor.
3. The method for identifying the terminal position according to claim 1, characterized in that: The parsing of the sensor data to obtain feature data corresponding to the target sensor includes: Preprocessing the sensor data of each of the target sensors; Extracting n-dimensional time domain feature data in the time domain and m-dimensional frequency domain feature data in the frequency domain from each of the preprocessed sensor data; wherein n and m are positive integers; Normalizing the extracted time domain feature data and frequency domain feature data; and According to a preset weight strategy, time domain feature data and frequency domain feature data of a specified number and dimension are extracted from the normalized time domain feature data and frequency domain feature data to construct the feature data.
4. The method for identifying the terminal position according to claim 1, characterized in that: The method further includes: pre-training the location recognition model, including: Collecting raw sensor data from a number of different terminal devices, and preprocessing the raw sensor data to construct a raw data set; wherein the raw sensor data is used to describe the location type of the terminal device; Parsing the original data set to extract corresponding time domain feature data and frequency domain feature data; Constructing a feature vector using the time domain feature data and the frequency domain feature data, and constructing sample data based on the feature vector and corresponding label information; The sample data is input into each classifier in the location recognition model for training, and each classifier is trained until convergence to obtain the location recognition model.
5. The method for identifying the terminal position according to claim 4, characterized in that: When constructing a feature vector using the time domain feature data and the frequency domain feature data, the method further includes: Among the extracted time domain feature data and frequency domain feature data, time domain feature data and frequency domain feature data of a specified type are selected according to a preset rule to construct a feature vector.
6. The method for identifying the terminal position according to claim 4 or 5, characterized in that: When constructing a feature vector using the time domain feature data and the frequency domain feature data, the method further includes: The time domain feature data and the frequency domain feature data of a specified quantity and / or dimension are extracted according to a preset weight strategy to construct the feature vector.
7. The method for identifying the terminal position according to claim 4 or 5, characterized in that: Different sensors installed in the terminal device have the same number of dimensions of the corresponding time domain feature data and frequency domain feature data collected.
8. The method for identifying the terminal position according to claim 4, characterized in that: The method further comprises: Extracting a preset number of features according to a preset rule to construct an auxiliary feature vector; the auxiliary feature vector is used to describe the user's usage habits of the terminal device; and The auxiliary feature vector is used as sample data.
9. The method for identifying the terminal position according to claim 1, characterized in that: The method further comprises: The location type of the terminal device is predefined; wherein the location type includes at least one type.
10. A terminal position identification device, characterized in that: The device comprises: A data acquisition module, used to acquire sensor data of a target sensor; wherein the target sensor includes at least one type of sensor equipped on the terminal device; A data analysis module, used to analyze the sensor data to obtain feature data corresponding to the target sensor; wherein the feature data includes time domain feature data and frequency domain feature data; A model processing module, used for inputting the feature data into a trained location recognition model, and using each classifier in the location recognition model to calculate the corresponding probability value; wherein the number of classifiers in the location recognition model is the same as the number of predefined location categories of the terminal; The recognition result output module is used to determine the recognition result of the terminal location according to the comparison result of the probability values output by each classifier, obtain multiple consecutive recognition results within a preset time length, judge the length of time that the terminal device stays in a certain state according to the multiple consecutive recognition results, and determine the power consumption optimization strategy for the terminal device according to the length of time.
11. A computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for identifying the terminal position according to any one of claims 1 to 9 is implemented.
12. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to execute the terminal location identification method according to any one of claims 1 to 9 by executing the executable instructions.
Citation Information
Patent Citations
Gait identity identification method and system thereof
CN107016346A
