Mobile phone holding state identification method based on posture and pressing information

Abstract

The invention provides a mobile phone holding state recognition method based on posture and pressing information, and belongs to the technical field of mobile phone holding state recognizing.The method comprises the steps that touch information of a user is collected by installing a capacitive touch button; a carrier coordinate system and a navigation coordinate system are defined, user posture information is updated and corrected in combination with the angular velocity measured by the gyroscope, and posture and pressing information is obtained by extracting Euler angles; the adjacent capacitive touch buttons are fused, and the posture and pressing information is encoded in a segmented mode, so that the posture and pressing information subjected to dimension reduction is obtained; and constructing a classifier for voting based on a preset specific holding state category to obtain a predicted holding state, and smoothing the predicted holding state by using a sliding window and sequential logic to obtain a real mobile phone holding state. According to the invention, the problems of screen mistaken rotation, privacy disclosure and large resource consumption in the lying scene of the existing mobile phone holding state identification method are solved.

Description

Technical Field

[0001] This invention belongs to the field of mobile phone grip state recognition technology, and particularly relates to a method for mobile phone grip state recognition based on posture and pressing information. Background Technology

[0002] As smartphones are ubiquitous electronic devices in today's world, improving user experience has always been a key focus of technological development for major manufacturers. Among these features, automatic screen rotation, a fundamental function of smartphones, automatically adjusts the screen's orientation by sensing the device's spatial posture, providing users with the best visual experience.

[0003] Traditional screen rotation mechanisms primarily rely on gravity sensors to detect the device's tilt angle relative to the direction of gravity. When the sensor detects a change in the device's angle, it triggers screen rotation. However, this simple gravity-sensing mechanism cannot handle flexible usage scenarios. For example, when a user is lying down, the phone's angle relative to gravity often differs from the user's desired screen orientation, leading to incorrect or delayed screen rotation.

[0004] In recent years, with the rapid development of sensor technology and data processing algorithms, research on system architectures based on sensor fusion has gradually become a hot topic. Another feasible approach is to use a camera to recognize and detect the user's facial features, thereby identifying the user's grip and making corresponding adjustments. However, this method still has some problems. Image-based recognition may lead to user privacy leaks; at the same time, the inference process using neural networks consumes a lot of resources, thereby reducing the phone's battery life and performance. Summary of the Invention

[0005] To address the aforementioned shortcomings in existing technologies, this invention provides a mobile phone grip state recognition method based on posture and pressure information, which solves the problems of screen mis-rotation, privacy leakage, and high resource consumption in lying scenarios in existing mobile phone grip state recognition methods.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a method for recognizing the holding state of a mobile phone based on posture and pressure information, comprising the following steps: S1. By installing a preset number of capacitive touch buttons on the four edges of the mobile phone simulation device, user touch information is collected; S2. Define the carrier coordinate system and navigation coordinate system, update the phone's attitude information by combining the angular velocity measured by the gyroscope, and correct the updated phone attitude information by using the gravity vector. By extracting Euler angles and combining them with user touch information, obtain the user's attitude and pressing information. S3. Based on the general characteristics of user behavior, by fusing adjacent capacitive touch buttons, the user's posture and pressing information are segmented and encoded to obtain the dimensionality-reduced posture and pressing information. S4. Based on the preset specific grip state category, a classifier is constructed, and the dimensionality-reduced posture and pressing information are voted on to obtain the predicted grip state. Then, the predicted grip state is smoothed using sliding window and timing logic to obtain the real mobile phone grip state.

[0007] The beneficial effects of this invention are as follows: This invention fuses posture and pressing information through capacitive touch buttons and gyroscope measurements to obtain accurate posture information. It then classifies the information through segmented encoding and a binary classifier, and uses sliding windows and timing logic for smoothing to obtain the true mobile phone holding state. This enables the screen to rotate correctly in lying scenarios, prevents privacy leaks, and reduces resource consumption.

[0008] Further, S2 includes the following steps: S201. Using the mobile device as the carrier, establish a right-handed rectangular coordinate system. Define the carrier coordinate system by taking the front of the mobile device as the horizontal axis, the right side of the mobile device as the vertical axis, and the top of the mobile device as the vertical axis. S202. By taking the front of the mobile device as the horizontal axis, the right side of the mobile device as the vertical axis, and the direction pointing to the ground as the vertical axis, a navigation coordinate system is defined. S203. Using the carrier coordinate system and navigation coordinate system, obtain the mobile phone attitude information, and combine it with the angular velocity measured by the gyroscope to update the mobile phone attitude information, and obtain the updated mobile phone attitude information. S204. Based on the carrier coordinate system, the acceleration is measured by an accelerometer and filtered by a low-pass filter to obtain the gravity vector in the carrier coordinate system. The updated mobile phone attitude information is then corrected using the gravity vector to obtain the corrected mobile phone attitude information. S205. Based on the corrected mobile phone attitude information, extract the Euler angle attitude information; S206. Combine the Euler angle attitude information and user touch information into a vector to obtain the user attitude and pressure information.

[0009] Furthermore, the expression for updating the phone's posture information is as follows: in, Indicates the current posture. It indicates the posture at the previous moment. To represent quaternion multiplication, Indicates the sampling period. Represents the quaternion of angular velocity. .

[0010] The beneficial effects of the above-mentioned further solutions are as follows: the present invention integrates the phone's posture with the user's pressing information, which enables better decision-making compared to information from a single sensor.

[0011] Furthermore, step S3 includes the following steps: S301. Based on the general characteristics of user behavior, by merging adjacent capacitive touch buttons, the capacitive touch buttons are grouped and encoded in a preset number of groups to obtain the first segment of capacitive touch button group. S302. The first segment of capacitive touch button group is grouped and encoded again by a preset number of groups to obtain the second segment of capacitive touch button group. S303. The user's posture and pressing information are segmented and encoded according to the second segment of capacitive touch button group to obtain the dimension-reduced posture and pressing information.

[0012] Furthermore, the reduced posture and pressure information is as follows: in, Indicates the roll angle. Indicates pitch angle, Indicates the heading angle. This indicates the gesture information of the second segment of the capacitive touch button group, which is encoded in a segmented manner.

[0013] The beneficial effects of the above-mentioned further solutions are as follows: the present invention encodes high-dimensional touch signals into low-dimensional features through efficient encoding dimensionality reduction, which greatly improves the computational efficiency.

[0014] Furthermore, step S4 includes the following steps: S401. Based on the preset specific grip state categories, construct a binary classifier for every two categories in the preset specific grip state categories; S402. Use each binary classifier to vote on the dimensionality-reduced posture and pressing information, and obtain the predicted grip state by obtaining the category with the most votes. S403. Based on the predicted grip state, the sliding window steady-state method is used to judge the predicted grip state, and combined with the posture change weight, the changes in the predicted grip state at different times are smoothed to obtain the real mobile phone grip state.

[0015] Furthermore, the expression for predicting the grip state is as follows: in, Indicates the predicted holding state. This represents the function that maximizes the summation function. k The value of , This indicates the preset specific grip state category. Indicates an indicator function, Indicates the type of gripping state. i and holding status category j A binary classifier, This represents the attitude information after dimensionality reduction.

[0016] Furthermore, S403 includes the following steps: S4031. Based on the predicted grip state, a preset time length is obtained through a preset sliding window length, and the predicted grip state at different times is obtained according to the preset time length. S4032. The predicted holding states at different times are placed into a preset sliding window for smoothing. When an unstable state appears in the sliding window, the number of unstable states is recorded, and the sliding window continues to slide. It is determined whether the subsequent state data in the preset sliding window is an unstable state. If not, the current state change is ignored. If so, the subsequent state data in the preset sliding window is recorded, and the number of unstable states is updated. S4033. Set the posture change weight according to the preset specific grip state category; S4034. Based on timing logic, select posture change weights according to the unstable state and the current holding state, and obtain the actual mobile phone holding state in response to the posture change weight ratio when the number of unstable states reaches the time window.

[0017] The beneficial effects of the above-mentioned further solutions are as follows: By selecting the attitude change weight based on the steady state of the sliding window and combined with the timing logic, the present invention can effectively filter out instantaneous misjudgments, prevent the screen from shaking at high frequency between horizontal and vertical orientations, and suppress the generation of misjudgments. Furthermore, this invention does not require processing image data, thus reducing computational load and completely avoiding the risk of user privacy leakage. The power consumption of capacitive sensing is also far lower than that of cameras. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention.

[0019] Figure 2 This is a diagram of the segmented encoding method in this embodiment.

[0020] Figure 3 This is a diagram of the sliding window steady-state method in this embodiment.

[0021] Figure 4 This is a diagram showing the attitude change weight settings based on timing logic in this embodiment. Detailed Implementation

[0022] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0023] Example like Figure 1 As shown, this invention provides a method for recognizing the holding state of a mobile phone based on posture and pressure information, and its implementation method is as follows: S1. By installing a preset number of capacitive touch buttons on the four edges of the mobile phone simulation device, user touch information is collected.

[0024] In this embodiment, touch information is obtained by a capacitive touch button. After the capacitive touch button is powered on, it outputs a high level. After being touched, the indicator light illuminates and the output becomes a low level. A preset number of 37 capacitive touch buttons are installed on the four edges of the mobile phone simulation device, simulating the curved screen side of the mobile phone, to fully sense the user's grip and finger contact, and obtain the current user touch information.

[0025] S2. Define the carrier coordinate system and navigation coordinate system. Combine the angular velocity measured by the gyroscope to update the phone's attitude information. Then, use the gravity vector to correct the updated phone attitude information. By extracting Euler angles and combining them with user touch information, obtain the user's attitude and pressure information. The specific steps are as follows: S201. Using the mobile device as the carrier, establish a right-handed rectangular coordinate system. Define the carrier coordinate system by taking the front of the mobile device as the horizontal axis, the right side of the mobile device as the vertical axis, and the top of the mobile device as the vertical axis. S202. By taking the front of the mobile device as the horizontal axis, the right side of the mobile device as the vertical axis, and the direction pointing to the ground as the vertical axis, a navigation coordinate system is defined.

[0026] In this embodiment, a carrier coordinate system and a navigation coordinate system are defined. The carrier coordinate system is a right-handed Cartesian coordinate system with the mobile phone as the carrier. The X-axis points in front of the device, the Y-axis points to the right side of the device, and the Z-axis points above the device. The navigation coordinate system points to the ground with the Z-axis, and the X and Y axes point in the same direction as the planar components of the X and Y axes of the carrier coordinate system.

[0027] S203. Using the carrier coordinate system and navigation coordinate system, obtain the mobile phone attitude information, and combine it with the angular velocity measured by the gyroscope to update the mobile phone attitude information, and obtain the updated mobile phone attitude information. S204. Based on the carrier coordinate system, the acceleration is measured by an accelerometer and filtered by a low-pass filter to obtain the gravity vector in the carrier coordinate system. The updated mobile phone attitude information is then corrected using the gravity vector to obtain the corrected mobile phone attitude information.

[0028] In this embodiment, the user's attitude information is obtained through the carrier coordinate system and the navigation coordinate system. Specifically, the phone's attitude information is obtained using the carrier coordinate system and the navigation coordinate system, and the phone's attitude information is updated based on the angular velocity measured by the gyroscope to obtain the updated phone attitude information. The state update equation of the attitude quaternion used to update the phone attitude information is as follows: in, Indicates the current posture. It indicates the posture at the previous moment. To represent quaternion multiplication, Indicates the sampling period. Represents the quaternion of angular velocity. ; Because gyroscope integration introduces cumulative errors, correction is needed using the gravity direction measured by the accelerometer. Gravitational acceleration can generally be considered a constant vector with unchanging direction and magnitude; therefore, a low-pass filter is used to filter the accelerometer data to extract the gravity component. The gravity acceleration extraction function is shown below: in, This represents the gravity vector in the carrier coordinate system. , Indicates a low-pass filter. express t The acceleration in the carrier coordinate system read by the accelerometer at all times. express t Accelerometer readings of acceleration in the carrier coordinate system at time -1; gravity vector in the carrier coordinate system is obtained. And using the gravity vector The updated phone posture information is corrected to obtain the corrected phone posture information.

[0029] S205. Based on the corrected mobile phone attitude information, extract the Euler angle attitude information; S206. Combine the Euler angle attitude information and user touch information into a vector to obtain the user attitude and pressure information.

[0030] In this embodiment, to extract Euler angles from quaternions as input features for subsequent mobile phone grip state classification, the Euler angle extraction formula is as follows: in, Indicates the roll angle. Indicates pitch angle, Indicates the heading angle. Represent the real part of a quaternion. Each represents the imaginary part of the quaternion, thus obtaining the Euler angle attitude information. ; The Euler angle attitude information and user touch information are combined into a vector to obtain user attitude and press information, which includes 3D Euler angle attitude information and 37D user touch information collected by capacitive touch sensor.

[0031] S3. Based on the general characteristics of user behavior, by fusing adjacent capacitive touch buttons, the user's posture and pressing information are segmented and encoded to obtain the dimensionality-reduced posture and pressing information. The specific steps are as follows: S301. Based on the general characteristics of user behavior, by merging adjacent capacitive touch buttons, the capacitive touch buttons are grouped and encoded in a preset number of groups to obtain the first segment of capacitive touch button group. S302. The first segment of capacitive touch button group is grouped and encoded again by a preset number of groups to obtain the second segment of capacitive touch button group. S303. The user's posture and pressing information are segmented and encoded according to the second segment of capacitive touch button group to obtain the dimension-reduced posture and pressing information.

[0032] In this embodiment, since the posture and pressing information have 40 dimensions, if classification and judgment are performed directly, the classification reasoning speed will be slowed down due to the high dimension of the input data. Therefore, segmented encoding is adopted to reduce the dimension of posture information while preserving the original information of the input data. like Figure 2 As shown, the segmented encoding specifically involves: based on the general characteristics of user behavior, merging adjacent capacitive touch buttons and grouping them into a preset number (2 or 3) of groups to obtain 14 groups of the first segment of capacitive touch buttons. ; Furthermore, the first segment of capacitive touch button group is grouped and encoded again by a preset number (2 or 3) to obtain the second segment of capacitive touch button group. .

[0033] In this embodiment, the posture and pressure information are segmented and encoded according to the second segment of capacitive touch button group, reducing the impact of encoding weights. At the same time, adjacent capacitive touch buttons are encoded, improving the correlation between adjacent capacitive touch buttons. When a user touches a capacitive button, they often also touch the adjacent capacitive buttons. Touch merging in the side area better reflects touch information and reduces the data complexity of the model, resulting in dimensionality-reduced posture and pressure information. The dimensionality-reduced posture and pressure information can be represented as a 10-dimensional vector, as shown below: in, This indicates the gesture information of the second segment of the capacitive touch button group, which is encoded in a segmented manner.

[0034] S4. Based on the preset specific grip state categories, a classifier is constructed, and the dimensionality-reduced posture and pressing information are voted on to obtain the predicted grip state. Then, using sliding window and timing logic, the predicted grip state is smoothed to obtain the real phone grip state. The specific steps are as follows: S401. Based on the preset specific grip state categories, construct a binary classifier for every two categories in the preset specific grip state categories; S402. Use each binary classifier to vote on the dimensionality-reduced posture and pressing information, and obtain the predicted grip state by obtaining the category with the most votes.

[0035] In this embodiment, specific gripping state categories are preset, including: {left-side lying left horizontal screen, left-side lying right horizontal screen, left-side lying portrait screen, lying flat left horizontal screen, lying flat right horizontal screen, lying flat portrait screen, right-side lying left horizontal screen, right-side lying right horizontal screen, right-side lying portrait screen, standing left horizontal screen, standing right horizontal screen, and standing portrait screen}. To handle the subtle differences between categories, a one-to-one multi-classification strategy is adopted, constructing a classifier for every two categories in the preset specific gripping state categories, for a total of 66 binary classifiers; each binary classifier is specifically responsible for distinguishing two specific gripping states; The binary classifiers are used to vote on the dimensionality-reduced pose and pressure information, and the grip state is finally predicted as the category that receives the most votes, as shown in the following expression: in, Indicates the predicted holding state. This represents the function that maximizes the summation function. k The value of , This indicates the preset specific grip state category. Indicates an indicator function, Indicates the type of gripping state. i and holding status category j A binary classifier, This represents the reduced-dimensional posture and pressure information.

[0036] S403. Based on the predicted grip state, the sliding window steady-state method is used to determine the predicted grip state, and combined with the posture change weight, the changes in the predicted grip state at different times are smoothed to obtain the real mobile phone grip state. The specific steps are as follows: S4031. Based on the predicted grip state, a preset time length is obtained through a preset sliding window length, and the predicted grip state at different times is obtained according to the preset time length. S4032. The predicted holding states at different times are placed into a preset sliding window for smoothing. When an unstable state appears in the sliding window, the number of unstable states is recorded, and the sliding window continues to slide. It is determined whether the subsequent state data in the preset sliding window is an unstable state. If not, the current state change is ignored. If so, the subsequent state data in the preset sliding window is recorded, and the number of unstable states is updated. S4033. Set the posture change weight according to the preset specific grip state category; S4034. Based on timing logic, select posture change weights according to the unstable state and the current holding state, and obtain the actual mobile phone holding state in response to the posture change weight ratio when the number of unstable states reaches the time window.

[0037] In this embodiment, after obtaining the predicted holding state, certain judgments and error removal are required. Smoothing is performed using a sliding window and timing logic to obtain a stable and accurate result of the actual phone holding state. Specifically: like Figure 3 As shown, based on the predicted grip state, the corresponding time length is obtained by setting a sliding window length of 35. Based on a preset time length, the predicted grip state at different times is obtained; The predicted gripping states at different times are placed into a preset sliding window for smoothing. When an unstable state occurs in the system, the number of unstable states is recorded. If the unstable state does not appear again in subsequent state data, the change is ignored. If the unstable state continues to appear, the subsequent state data in the preset sliding window is recorded and the number of unstable states is updated. like Figure 4 As shown, the posture change weight is set according to the preset specific holding state category. If the current state is standing, the recognition algorithm is more likely to be standing in the next recognition process. Figure 4The red arrows indicate that switching is possible freely, while the black dashed arrows indicate that switching is only possible under certain conditions. These conditions are: a red arrow indicates that more than half of the points in the sliding window output a certain pose, at which point the transformation can occur; a black arrow indicates that more than 65% of the points in the sliding window output a certain pose, at which point the transformation can occur. Specifically: Dynamic pose change weights will be used within the time window. State discrimination is performed, and the posture change weights during the transition from standing to other lying positions are determined in the sliding window. It will become larger (65%), and if you switch from standing portrait mode to right horizontal mode, the posture change weight will be affected. It remains at the initial state (50%) to suppress the occurrence of misjudgments; In the recognition method, the output of the previous time step guides the prediction of the system's output at the next time step; for example, in If the system outputs that the user's grip posture is standing and holding the screen in portrait mode, then in the next moment, the probability that the user will still be holding the screen in a standing posture is high, and the probability of switching to a lying posture is low. This is because in a very short time, it is difficult for the user's grip posture to change drastically, such as switching from standing to lying down. Such posture switching usually takes a certain amount of time. Based on temporal logic, attitude change weights are selected according to the unstable state and the current holding state. The attitude change weights respond to the number of unsteady states reaching the time window. The ratio is used to obtain the actual phone holding state; the telephone equipment system can intelligently switch between landscape and portrait screens based on the actual phone holding state.

[0038] In this embodiment, the present invention provides a novel method for recognizing mobile phone grip status based on posture and pressure information, enabling users to correctly switch between landscape and portrait modes while lying flat or on their side. It uses hand grip posture and edge pressure signals as the primary criteria, abandoning the sole reliance on gravity vectors and fundamentally solving the problem of screen mis-rotation in lying scenarios. By integrating a low-power inertial unit and a capacitive touch array, it avoids the privacy risks and high computational overhead associated with image acquisition. Through dimensionality reduction coding technology combined with a support vector machine model, it achieves fine differentiation of multiple grip modes, combining high accuracy and real-time performance, significantly improving the intelligence of landscape and portrait switching and the user experience.

Claims

1. A method for recognizing mobile phone holding state based on posture and pressure information, characterized in that, Includes the following steps: S1. By installing a preset number of capacitive touch buttons on the four edges of the mobile phone simulation device, user touch information is collected; S2. Define the carrier coordinate system and navigation coordinate system, update the phone's attitude information by combining the angular velocity measured by the gyroscope, and correct the updated phone attitude information by using the gravity vector. By extracting Euler angles and combining them with user touch information, obtain the user's attitude and pressing information. S3. Based on the general characteristics of user behavior, by fusing adjacent capacitive touch buttons, the user's posture and pressing information are segmented and encoded to obtain the dimensionality-reduced posture and pressing information. S4. Based on the preset specific grip state category, a classifier is constructed, and the dimensionality-reduced posture and pressing information are voted on to obtain the predicted grip state. Then, the predicted grip state is smoothed using sliding window and timing logic to obtain the real mobile phone grip state.

2. The mobile phone grip state recognition method based on posture and pressure information according to claim 1, characterized in that, S2 includes the following steps: S201. Using the mobile device as the carrier, establish a right-handed rectangular coordinate system. Define the carrier coordinate system by taking the front of the mobile device as the horizontal axis, the right side of the mobile device as the vertical axis, and the top of the mobile device as the vertical axis. S202. By taking the front of the mobile device as the horizontal axis, the right side of the mobile device as the vertical axis, and the direction pointing to the ground as the vertical axis, a navigation coordinate system is defined. S203. Using the carrier coordinate system and navigation coordinate system, obtain the mobile phone attitude information, and combine it with the angular velocity measured by the gyroscope to update the mobile phone attitude information, and obtain the updated mobile phone attitude information. S204. Based on the carrier coordinate system, the acceleration is measured by an accelerometer and filtered by a low-pass filter to obtain the gravity vector in the carrier coordinate system. The updated mobile phone attitude information is then corrected using the gravity vector to obtain the corrected mobile phone attitude information. S205. Based on the corrected mobile phone attitude information, extract the Euler angle attitude information; S206. Combine the Euler angle attitude information and user touch information into a vector to obtain the user attitude and press information.

3. The mobile phone grip state recognition method based on posture and pressure information according to claim 2, characterized in that, The expression for updating the phone's posture information is as follows: in, Indicates the current posture. It indicates the posture at the previous moment. To represent quaternion multiplication, Indicates the sampling period. Represents the quaternion of angular velocity. .

4. The mobile phone grip state recognition method based on posture and pressure information according to claim 1, characterized in that, S3 includes the following steps: S301. Based on the general characteristics of user behavior, by merging adjacent capacitive touch buttons, the capacitive touch buttons are grouped and encoded in a preset number of groups to obtain the first segment of capacitive touch button group. S302. The first segment of capacitive touch button group is grouped and encoded again by a preset number of groups to obtain the second segment of capacitive touch button group. S303. The user's posture and pressing information are segmented and encoded according to the second segment of capacitive touch button group to obtain the dimension-reduced posture and pressing information.

5. The mobile phone grip state recognition method based on posture and pressure information according to claim 4, characterized in that, The reduced posture and pressure information is shown below: in, Indicates the roll angle. Indicates pitch angle, Indicates the heading angle. This indicates the gesture information of the second segment of the capacitive touch button group, which is encoded in a segmented manner.

6. The mobile phone grip state recognition method based on posture and pressure information according to claim 1, characterized in that, S4 includes the following steps: S401. Based on the preset specific grip state categories, construct a binary classifier for every two categories in the preset specific grip state categories; S402. Use each binary classifier to vote on the dimensionality-reduced posture and pressing information, and obtain the predicted grip state by obtaining the category with the most votes. S403. Based on the predicted grip state, the sliding window steady-state method is used to judge the predicted grip state, and combined with the posture change weight, the changes in the predicted grip state at different times are smoothed to obtain the real mobile phone grip state.

7. The mobile phone grip state recognition method based on posture and pressure information according to claim 6, characterized in that, The expression for predicting the grip state is as follows: in, Indicates the predicted holding state. This represents the function that maximizes the summation function. k The value of , This indicates the preset specific grip state category. Indicates an indicator function, Indicates the type of gripping state. i and holding status category j A binary classifier, This represents the reduced-dimensional posture and pressure information.

8. The mobile phone grip state recognition method based on posture and pressure information according to claim 6, characterized in that, S403 includes the following steps: S4031. Based on the predicted grip state, a preset time length is obtained through a preset sliding window length, and the predicted grip state at different times is obtained according to the preset time length. S4032. The predicted holding states at different times are placed into a preset sliding window for smoothing. When an unstable state appears in the sliding window, the number of unstable states is recorded, and the sliding window continues to slide. It is determined whether the subsequent state data in the preset sliding window is an unstable state. If not, the current state change is ignored. If so, the subsequent state data in the preset sliding window is recorded, and the number of unstable states is updated. S4033. Set the posture change weight according to the preset specific grip state category; S4034. Based on timing logic, select posture change weights according to the unstable state and the current holding state, and obtain the actual mobile phone holding state in response to the posture change weight ratio when the number of unstable states reaches the time window.

Images