A method and apparatus for device fingerprinting

Abstract

The application is suitable for the technical field of fingerprint identification, and provides a device fingerprint identification method and device, the device fingerprint identification method comprises: in the process that a user's finger contacts and presses until leaves a sensor, an image frame sequence of a fingerprint to be identified is collected at a fixed frame rate; a static feature vector is extracted based on a frame image corresponding to a maximum contact area in the image frame sequence; a time sequence behavior feature is extracted based on the image frame sequence; whether a user corresponding to the image frame sequence is a device user is identified based on the static feature vector and the time sequence behavior feature. The introduction of the dynamic feature makes the identification process not only depend on static image data, but also consider the actual behavior of the user in the operation process, thereby improving the identification ability of the real user.

Description

[0001] This invention belongs to the technical field of fingerprint recognition, and particularly relates to a method and apparatus for fingerprint recognition in a device. Background Technology

[0002] With the rapid development of information technology and the widespread adoption of smart devices, the demand for user authentication is becoming increasingly urgent. Traditional authentication methods, such as passwords, ID cards, and biometrics including facial and iris recognition, while improving security to some extent, still have many shortcomings. For example, passwords are easily stolen or forgotten, ID cards are easily forged, and facial and iris recognition become less accurate in low-light conditions or when the user's facial features change. Fingerprint recognition, as a mature biometric identification technology, is widely used in various devices due to its uniqueness and stability. However, existing fingerprint recognition technologies mostly focus on extracting static features, such as the patterns and details of the fingerprint image, and are typically performed when the user's finger is stationary. While this method can ensure accuracy to a certain extent, in practical applications, the recognition effect of static features is poor when the user quickly touches or slides the sensor. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a method and apparatus for device fingerprint recognition to solve the technical problem of poor recognition effect of static features when the user quickly touches or slides the sensor. A first aspect of the present invention provides a method for device fingerprint recognition, the method comprising: During the process of the user's finger touching and pressing the sensor until it leaves the sensor, a sequence of image frames of the fingerprint to be identified is acquired at a fixed frame rate; Based on the frame image corresponding to the maximum contact area in the image frame sequence, extract the static feature vector; Based on the image frame sequence, temporal behavioral features are extracted; wherein, the temporal behavioral features are used to characterize the change patterns corresponding to fingerprint pressure diffusion, fingerprint ridge deformation and fingerprint slippage; Based on the static feature vector and the temporal behavior features, it is determined whether the user corresponding to the image frame sequence is a device user. Further, the step of extracting static feature vectors based on the frame image corresponding to the maximum contact area in the image frame sequence includes: Multiple detail points are extracted from the frame image corresponding to the largest contact area; wherein, the detail points include endpoints and bifurcation points; For each detail point, local texture features of the surrounding preset region are extracted to form a local texture descriptor; wherein, the local texture features include at least one of the following: ridge curvature statistical histogram, ridge-valley width ratio variation sequence, and phase information based on Gabor filtering in the preset region surrounding the detail point; The local texture descriptors of all detail points are aggregated and encoded to generate an aggregated local texture feature vector, which is then used as a static feature vector. Furthermore, the step of extracting temporal behavioral features based on the image frame sequence includes: Based on the binarized contact mask of each frame image in the image frame sequence, the curves of the area, centroid position and shape parameters of the contact area changing with time are calculated, and the dynamic parameters are extracted from the curves to form the first sub-feature vector; Fingerprint image temporal sequence is enhanced and tracked for ridges, displacement field of selected feature points is calculated, and displacement pattern of finger skin is analyzed based on the displacement field to form a second sub-feature vector; Calculate the global motion vector between adjacent frame images, filter the global motion vector to separate the high-frequency jitter component, and extract the mode features of the high-frequency jitter component to form a third sub-feature vector; Arrange the first sub-feature vector, the second sub-feature vector, and the third sub-feature vector in chronological order to form a feature time sequence; The time-series feature is input into a time-series neural network, which then extracts and outputs the time-series behavioral features. Further, the step of calculating the curves of the area, centroid position, and shape parameters of the contact region changing over time based on the binarized contact mask of each frame image in the image frame sequence, and extracting dynamic parameters from the curves to form the first sub-feature vector includes: Adaptive threshold binarization is performed on each frame image to generate a contact mask; Calculate the contact mask area and centroid coordinates for each frame image, and fit the circumscribed ellipse of the contact area to obtain its shape parameters; wherein, the shape parameters include the major axis, minor axis and orientation angle; The change curve of the contact mask area over time was obtained by nonlinear fitting. Extract the maximum contact mask area and extract the diffusion rate parameter from the variation curve; Calculate the total displacement of the centroid trajectories corresponding to all frames in the image frame sequence; Calculate the statistics of how the shape parameters of the circumscribed ellipse change over time; The first sub-feature vector is formed by combining the maximum contact mask area, diffusion rate parameter, total displacement, and the statistics. Furthermore, the steps of performing ridge enhancement and tracking on the temporal sequence of the fingerprint image, calculating the displacement field of selected feature points, and analyzing the displacement pattern of the finger skin based on the displacement field to form a second sub-feature vector include: A filter is used to enhance the ridge lines of each frame in the image frame sequence to obtain a ridge-enhanced image. Select a set of feature points on the texture enhancement image corresponding to the first frame image in the image frame sequence; The displacement field is obtained by tracking the positional changes of the feature points in the image frame sequence using optical flow. Based on the displacement field, the radial displacement of each feature point relative to the centroid of the contact area corresponding to the first frame image is calculated. Extract the mean and standard deviation for all radial displacements; Based on the radial displacement corresponding to multiple feature points in each frame image, calculate the kurtosis and skewness of each frame image; The mean, standard deviation, kurtosis, and skewness are used as the second sub-feature vector. Further, the steps of calculating the global motion vector between adjacent frame images, filtering the global motion vector to separate high-frequency jitter components, and extracting the mode features of the high-frequency jitter components to form a third sub-feature vector include: The global translation vector between adjacent frames is calculated using the phase correlation method. High-pass filtering is applied to the time series of the global translation vector to obtain a high-frequency jitter signal; Perform spectral analysis on the high-frequency jitter signal and calculate its spectral entropy in different frequency bands; Calculate the fractal dimension of the motion trajectory on a two-dimensional plane composed of high-frequency jitter signals; The spectral entropy and fractal dimension are combined to form the third sub-feature vector. Furthermore, the step of identifying whether the user corresponding to the image frame sequence is a device user based on the static feature vector and the temporal behavior features includes: Obtain the pre-stored static vector and pre-stored behavioral vector corresponding to historical users; Calculate the first similarity between the static feature vector and the pre-stored static vector; Calculate the second similarity between the temporal behavior features and the pre-stored behavior vectors; When both the first similarity and the second similarity are greater than the preset similarity, the user corresponding to the image frame sequence is determined to be the device user. A second aspect of the present invention provides a device fingerprint recognition apparatus, comprising: The acquisition unit is used to acquire a sequence of image frames of the fingerprint to be identified at a fixed frame rate during the process of the user's finger touching and pressing the sensor until it leaves the sensor; The first extraction unit is used to extract static feature vectors based on the frame image corresponding to the maximum contact area in the image frame sequence; The second extraction unit is used to extract temporal behavioral features based on the image frame sequence; wherein, the temporal behavioral features are used to characterize the change patterns corresponding to fingerprint pressure diffusion, fingerprint ridge deformation and fingerprint slippage; The identification unit is used to identify whether the user corresponding to the image frame sequence is a device user based on the static feature vector and the temporal behavior features. A third aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the device fingerprint recognition method described in the first aspect. A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the device fingerprint recognition method described in the first aspect. The beneficial effects of this invention compared to existing technologies are as follows: By acquiring image frame sequences during the user's finger contact with the sensor at a fixed frame rate, and especially by extracting static feature vectors based on the frame image corresponding to the maximum contact area, the detailed features of the user's fingerprint can be captured more accurately, reducing recognition errors caused by changes in fingerprint position or angle. Extracting temporal behavioral features allows the system to analyze the pressure diffusion, ridge deformation, and slippage patterns generated during fingerprint contact. The introduction of these dynamic features means that the recognition process not only relies on static image data but also considers the user's actual behavior during operation, thereby improving the ability to identify genuine users. Since temporal behavioral features can reflect the changing patterns of the user's fingerprint under different operating states, the system can effectively distinguish between genuine users and counterfeit fingerprints (such as silicone fingerprints). This capability significantly enhances system security and reduces the risk of identity theft. Attached Figure Description To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 A schematic flowchart of a device fingerprint recognition method provided by the present invention is shown; Figure 2A schematic diagram of a device fingerprint recognition apparatus according to an embodiment of the present invention is shown; Figure 3 A schematic diagram of a terminal device provided in an embodiment of the present invention is shown. Detailed Implementation In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail. This invention provides a method and apparatus for device fingerprint recognition to solve the technical problem of poor recognition effect of static features when the user quickly touches or slides the sensor. First, this invention provides a method for device fingerprint recognition. Please refer to [link to relevant documentation]. Figure 1 , Figure 1 A schematic flowchart of a device fingerprint recognition method provided by the present invention is shown. Figure 1 As shown, the fingerprint recognition method of this device may include the following steps: Step 101: During the process of the user's finger touching and pressing the sensor until it leaves the sensor, acquire a sequence of image frames of the fingerprint to be identified at a fixed frame rate; The data acquisition timing is not based on capturing a single fingerprint image, but rather on recording the complete dynamic process from contact and pressure to removal. Using a fixed frame rate ensures the continuity and equal intervals of the data over time, which forms the mathematical basis for subsequent precise time-series analysis. Fake fingerprints (such as silicone fingerprint molds or high-resolution images) can usually only simulate static textures and are difficult to perfectly reproduce the natural physical interaction process of a real finger during pressing. Step 102: Extract static feature vectors based on the frame image corresponding to the maximum contact area in the image frame sequence; The frame with the most complete contact and the highest image quality (maximum contact area frame) is intelligently selected from the sequence. This frame has the clearest texture and the most complete features. The static feature vector refers to the mathematical description extracted from this frame that does not change over time. Specifically, step 102 includes steps 1021 to 1023: Step 1021: Extract multiple detail points from the frame image corresponding to the largest contact area; wherein, the detail points include endpoints and bifurcation points; Minute points are the most crucial feature of fingerprint recognition. An endpoint is where a ridge terminates, and a bifurcation is where a ridge splits into two. The type, location, and orientation of these points constitute the unique topological skeleton of a fingerprint, giving it high distinguishability. Minutiae are stable features, but they are limited in number and susceptible to local damage. Their main function is to provide a localization reference for further, more refined texture analysis. Step 1022: For each detail point, extract the local texture features of the surrounding preset region to form a local texture descriptor; wherein, the local texture features include at least one of the following: ridge curvature statistical histogram, ridge-valley width ratio change sequence, and phase information based on Gabor filtering in the preset region surrounding the detail point; It not only records the minutiae at a single point, but also the complete texture pattern within a small neighborhood centered on that point (the preset region). The preset region is a local block or fan-shaped area centered on the minutiae, which has better robustness to the elastic deformation and partial occlusion of the fingerprint. Histogram analysis of ridge curvature reveals the degree of curvature and directional distribution of ridges around detail points. For example, dividing a 360-degree area into several sector-shaped regions allows for the statistical analysis of the average curvature or curvature distribution of ridge segments within each region. This effectively captures the flow direction and curvature patterns of the texture. The ridge-valley width ratio variation sequence is plotted with the minutiae as the center, along multiple radial directions (e.g., one direction every 45 degrees). The width ratio of the ridge (raised) and valley (depressed) on each line and their changing trends are analyzed. This reflects the stability or variation pattern of the local physiological structure of the fingerprint and has a certain characterizing ability for skin elasticity and pressure intensity. Gabor filters are considered the gold standard for texture analysis due to their similarity to the human visual system. Extracting the filtered phase information (rather than the commonly used amplitude) provides greater invariance to changes in lighting and the dryness or wetness of fingerprints, resulting in a more stable representation of the precise location and continuity of ridges. Quantizing one or more of the above features into a numerical vector is the local texture descriptor of that detail point. Even if the details themselves are slightly offset due to image quality, the texture pattern around them remains stable. Two fingerprints may have details in similar locations, but the microtexture (curvature, ridge-to-valley ratio) around them is almost impossible to be exactly the same. Forged fingerprints (such as prints) have a great difficulty in replicating this complex, natural pattern of real skin in terms of microtexture, especially at the phase information level. Step 1023: Aggregate and encode the local texture descriptors of all minutiae to generate an aggregated local texture feature vector, which is then used as a static feature vector. Aggregate encoding is a key operation that transforms an unordered set of local features into an ordered global vector. Aggregate encoding can employ Fisher vectors or the bag-of-words model; this process is a traditional technique and will not be elaborated upon here. The final output is a compact, highly discriminative mathematical vector. This vector not only contains the location information of the minutiae (implied in the aggregation relationship of the local descriptors), but more importantly, it encodes the rich, multi-layered local texture patterns of the entire fingerprint. In the embodiments corresponding to steps 1021 to 1023, by analyzing minutiae and extracting local texture features, the system can fully utilize the detailed information of the fingerprint, thereby enhancing the accuracy and reliability of fingerprint recognition. This method not only considers the overall structure of the fingerprint but also focuses on the details, providing rich information for the recognition process. Step 103: Based on the image frame sequence, extract temporal behavioral features; wherein, the temporal behavioral features are used to characterize the change patterns corresponding to fingerprint pressure diffusion, fingerprint ridge deformation, and fingerprint slippage; Temporal behavioral features are features extracted from the entire image sequence that describe the process of change. They focus on how the image was pressed, rather than how it was pressed. The pressure diffusion pattern is used to characterize how the finger contact area expands outward from the center over time. This reflects the user's pressing habits and pressure. Fingerprint ridge deformation patterns are the elastic deformations that occur in fingerprint ridges under pressure (such as flattening or widening of the ridges). This is directly related to the soft tissue characteristics of the fingertip and the pressure applied. Fingerprint swipe patterns are used to characterize minute lateral or rotational movements of the finger on the sensor surface. This reflects unconscious muscle twitches and touch habits. The three modes described above are physical and physiological phenomena unique to living fingers, making them extremely difficult to simulate and thus effectively resisting fake fingerprint attacks. It adds a unique pressing behavior signature to each user; even if the static features of fingerprints are similar, different users may have different pressing patterns. Expanding the recognition dimension from two dimensions (planar texture) to three dimensions (texture + time + behavior) greatly enhances the system's distinguishability and security. Specifically, step 103 includes steps 1031 to 1035: Step 1031: Based on the binarized contact mask of each frame image in the image frame sequence, calculate the curves of the area, centroid position and shape parameters of the contact area changing over time, and extract the dynamic parameters from the curves to form the first sub-feature vector; A binarized contact mask is a binary image that separates the area where a finger is in contact (foreground) from the area where it is not in contact (background) in each frame of an image. The area of ​​the contact region refers to the total number of contact pixels, which directly reflects the pressure applied (the greater the pressure, the more complete the contact, and the larger the area is usually). The center of gravity position refers to the coordinates of the center of gravity of the contact area, and its trajectory reflects the overall translational and rotational trend of the finger. From the curves showing the changes of the above parameters over time, dynamic parameters are extracted, such as the maximum rate of area growth / decay, the average velocity / acceleration of the center of mass movement, and the period or stability index of shape changes. These parameters constitute the first sub-feature vector. Specifically, step 1031 includes steps A1 to A7: Step A1: Perform adaptive threshold binarization on each frame image to generate a contact mask; The grayscale fingerprint image is clearly segmented into the area where the finger touched (foreground, usually white) and the background area. Adaptive thresholding differs from fixed thresholding in that it dynamically determines the segmentation threshold based on the grayscale distribution of local regions in the image. Adaptive threshold binarization can handle uneven conditions and effectively process local brightness and darkness variations in images caused by uneven pressure, varying finger dryness and wetness, and differences in sensor sensitivity, resulting in more accurate and stable contact area segmentation results. A precise mask is fundamental to all subsequent area, centroid, and shape calculations. This step improves the robustness of the entire feature extraction process. Step A2: Calculate the contact mask area and centroid coordinates for each frame image, and fit the circumscribed ellipse of the contact area to obtain its shape parameters; wherein, the shape parameters include the major axis, minor axis and orientation angle; The contact mask area is the total number of foreground pixels in the mask. It directly reflects the instantaneous contact area. Centroid coordinates are the average of the coordinates of all pixels in the mask area, i.e., the centroid of the area. They reflect the center position of the contact area. The circumscribed ellipse and shape parameters are the best elliptical approximation of the shape of the contact area. The major axis and minor axis are the lengths of the semi-major axis and semi-minor axis of the ellipse, respectively representing the extension scale of the contact area in the principal direction and the vertical direction. The orientation angle is the angle between the major axis and the horizontal axis, representing the main orientation of the contact area. Step A3: Perform nonlinear fitting on the change of the contact mask area over time to obtain the change curve; The area change A(t) is usually not linear. Fitting it with a nonlinear model (such as an sigmoid growth curve or an exponential growth / decline model) can smooth out noise and reveal its inherent trend. Step A4: Extract the maximum contact mask area and extract the diffusion rate parameter from the variation curve; The maximum contact mask area is the peak value A_max of the contact area throughout the entire pressing process. This is directly related to the user's maximum pressing force and the softness of the fingertip. The diffusion rate parameter is a parameter derived from the fitted curve that characterizes how quickly the area changes. The diffusion rate parameter includes, but is not limited to, the maximum instantaneous growth rate during the diffusion phase (area growth), the time required to reach half of the maximum area (half-race time), or the rate constant of the fitted model itself (such as the exponential coefficient of an exponential model). Step A5: Calculate the total displacement of the centroid trajectories corresponding to all frames in the image frame sequence; The path is formed by connecting the centroid coordinates (Cx(t), Cy(t)) of each frame on a two-dimensional plane. Total displacement refers to the total path length of the trajectory or the straight-line distance between the starting point and the ending point (net displacement). Total displacement quantifies the overall degree of finger slippage during the pressing process. A large total displacement indicates significant sliding or adjustment; a small total displacement indicates relatively stable pressing. Step A6: Calculate the statistics of the change of shape parameters corresponding to the circumscribed ellipse over time; The shape parameters include three sequences: major axis, minor axis, and orientation angle. The statistics are not based on the original sequences, but rather on calculated statistical values ​​that summarize the dynamic characteristics. These statistical values ​​include, but are not limited to, the mean and variance (reflecting the average size and fluctuation of the shape parameters), the mean and range of the aspect ratio (reflecting whether the contact area is nearly circular or clearly elliptical, and the magnitude of its variation), and / or the range or standard deviation of the orientation angle (reflecting whether the finger rotates during pressing and the magnitude of that rotation). Statistics capture the stability and variation patterns of a user's finger posture when pressing. For example, some people maintain a fixed finger posture when pressing (small variance in shape parameters), while others unconsciously make minor adjustments (large variance). Variations in aspect ratio are related to the anisotropy of pressure diffusion and can reflect force application habits. Step A7: Combine the maximum contact mask area, diffusion rate parameter, total displacement, and the statistics to form the first sub-feature vector. All the scalar parameters (maximum contact mask area, one or more diffusion rate parameters, total displacement, and multiple shape statistics) extracted in the above steps are sequentially concatenated into a one-dimensional vector. The original time-series data, which can span tens or even hundreds of frames, is compressed into a few feature values ​​with clear physical meaning. This greatly reduces the amount of data and highlights key dynamic information. In the embodiments corresponding to steps A1 to A7, a rich and representative feature vector is formed by integrating multiple steps such as adaptive binarization, area and centroid calculation, nonlinear fitting, and dynamic parameter extraction. These steps not only improve the accuracy of fingerprint recognition but also better cope with changes under different contact conditions, enhancing the robustness of the system. Step 1032: Perform ridge enhancement and tracking on the temporal sequence of the fingerprint image, calculate the displacement field of the selected feature points, and analyze the displacement pattern of the finger skin based on the displacement field to form a second sub-feature vector; Ridge enhancement and tracking aims to improve ridge contrast and establish inter-frame correspondence. The displacement field is calculated using optical flow or feature point tracking. That is, between consecutive frames, the pixel-level movement of a large number of points (such as ridge intersections and high curvature points) on the fingerprint image is tracked. The second sub-feature vector precisely depicts the microscopic elastic deformation and internal relative movement of the fingerprint skin soft tissue under pressure. This is a unique and extremely difficult-to-forge biomechanical characteristic of a living finger, and it is a direct manifestation of the fingerprint ridge deformation pattern. Specifically, step 1032 includes steps B1 to B7: Step B1: Use a filter to enhance the ridge lines of each frame in the image frame sequence to obtain a ridge-enhanced image; Using Gabor filter banks or similar directional filters enhances frequency components aligned with the fingerprint ridge direction while suppressing noise and valley information. This significantly improves the accuracy and robustness of subsequent optical flow calculations. Step B2: Select a set of feature points on the texture enhancement image corresponding to the first frame image in the image frame sequence; On the enhanced image of the first frame (usually the initial contact phase), select a set of feature points. These points should be located in textured regions (such as ridge intersections or points with high curvature) and their distribution should cover the entire contact area to ensure representative sampling. Provide a high-quality, high-contrast input image for optical flow tracking. Establish a stable set of tracking reference points to ensure that the motion of the same set of physical points is analyzed throughout the sequence. Step B3: Use optical flow to track the position changes of the feature points in the image frame sequence to obtain the displacement field; Optical flow is a computer vision technique used to estimate the motion vector (i.e., displacement) of each pixel in an image sequence. It calculates the movement of points between adjacent frames based on assumptions such as constant brightness. For the selected set of feature points, the optical flow method outputs their displacement vectors (Δx, Δy) in each frame relative to the first frame (or the previous frame). The set of displacement vectors of all feature points constitutes a sparse displacement field. Step B4: Based on the displacement field, calculate the radial displacement of each feature point relative to the centroid of the contact area corresponding to the first frame image; The absolute two-dimensional displacement (Δx, Δy) of each feature point is converted into a radial displacement relative to the centroid of the contact region. The calculation method is as follows: Let the coordinates of the feature point in the first frame be (x_i, y_i) and the coordinates of the centroid be (C_x, C_y). The initial radial direction vector at this point is: r_i = (x_i - C_x, y_i - C_y). In subsequent frames, the point moves to a new position (x_i', y_i'). The radial displacement d_radial is calculated as the projection of the displacement vector at that point onto the initial radial direction. d_radial = (Δx, Δy)·(r_i / ||r_i||), where · is the dot product. d_radial > 0 indicates that the point is far from the centroid (expansion). d_radial < 0 indicates that the point is close to the centroid (contraction). Radial displacement directly reflects the stretching or compression of skin tissue under pressure, which is the core of elastic deformation. It filters out overall translational and rotational movements, which may be caused by finger sliding rather than deformation of the skin itself. Step B5: Extract the mean and standard deviation for all radial displacements; Step B6: Based on the radial displacement corresponding to multiple feature points in each frame image, calculate the kurtosis and skewness of each frame image; Step B7: Use the mean, standard deviation, kurtosis, and skewness as the second sub-feature vector. This step involves statistical analysis of the radial displacement set of all feature points in each frame. The mean is the average of all radial displacements in that frame. Positive values ​​indicate overall expansion, while negative values ​​indicate overall contraction. Its time-varying curve reflects the overall pressure diffusion / contraction pattern. Standard deviation measures the dispersion of radial displacement. A large standard deviation indicates uneven skin deformation in different areas (some areas expand more, while others expand less or even shrink). A small standard deviation indicates uniform deformation. Skewness measures the asymmetry of radial displacement distribution and can reveal local anomalies in deformation modes. Skewness > 0 (positive skewness): The right tail of the distribution is longer, and there are many maximal expansion points that are much larger than the mean. Skewness < 0 (negative skewness): The left tail of the distribution is longer, and there are more maximum contraction points that are much smaller than the mean. Kurtosis measures the sharpness of a radial displacement distribution (compared to a normal distribution). A sharper kurtosis distribution with a heavier tail indicates more extreme displacement values ​​(maximum expansion or contraction). A flatter kurtosis distribution indicates more concentrated displacement values. This reflects the intensity and concentration of the deformation pattern. In the embodiments corresponding to steps B1 to B7, a comprehensive second sub-feature vector is formed through ridge enhancement, feature point tracking, displacement calculation, and statistical feature extraction. This process not only improves the accuracy of fingerprint recognition but also provides a richer data foundation for subsequent identification and verification. Step 1033: Calculate the global motion vector between adjacent frame images, filter the global motion vector to separate the high-frequency jitter component, and extract the mode features of the high-frequency jitter component to form a third sub-feature vector; Global motion refers to the overall motion trend of the entire fingerprint area between adjacent frames (mainly generated by the user's active pressing or swiping intention). High-frequency jitter components are separated by high-pass filtering or similar signal processing techniques to extract low-frequency, conscious motion from the global motion vector, while retaining high-frequency, minute, unconscious jitter. High-frequency jitter originates from muscle tremors, incomplete instability of neural control, and weak heartbeat conduction, and is a powerful marker of a living physiological organism. Extract spectral features (such as energy distribution in a specific frequency band), statistical features (such as the mean and variance of amplitude and frequency), or chaotic features from the high-frequency jitter components to form a third sub-feature vector. It captures unique physiological micro-motion signals that users cannot control voluntarily. This is an extremely powerful feature for liveness detection and user differentiation because counterfeit products (such as silicone fingerprints) completely lack this biological neuromuscular property. Specifically, step 1033 includes steps C1 to C5: Step C1: Calculate the global translation vector between adjacent frames using the phase correlation method; Phase correlation is an image registration technique based on Fourier transform. It estimates the translational offset (Δx, Δy) between two images by calculating the cross-power spectrum of the two images and finding the peak position of their inverse Fourier transform. This method is insensitive to changes in illumination and has high computational efficiency. For a sequence of N frames, this step generates N-1 global translation vectors, forming a two-dimensional time series V(t) = (Δx(t), Δy(t)). It captures the overall movement of the finger on the sensor surface, including intentional movements such as conscious pressing and sliding, as well as unconscious physiological jitters. This provides the raw motion signal for subsequent separation of the pure jitter component. Step C2: Perform high-pass filtering on the time series of the global translation vector to obtain a high-frequency jitter signal; A high-pass filter is a filter that allows high-frequency signals to pass through while suppressing low-frequency signals. Its core principle is to set a cutoff frequency. Conscious movements of the user (such as slow downward pressure, smooth sliding) are typically low-frequency (<1-2 Hz). Physiological micro-movements (such as muscle tremors, nerve twitches) have higher frequencies (typically in the range of 2-30 Hz). High-pass filters can effectively separate the latter. After filtering, a high-frequency jitter signal J(t) = (J_x(t), J_y(t)) is obtained. This is a two-dimensional time-series signal that mainly contains physiological components and has had its low-frequency trends removed. This is a crucial preprocessing step for liveness detection and user behavior recognition. It ensures that the object of subsequent analysis is truly biologically specific micro-motions, rather than user-controllable movement intentions. It avoids interference from low-frequency large-scale movements on the analysis of high-frequency micro-motion patterns. Step C3: Perform spectrum analysis on the high-frequency jitter signal and calculate its spectral entropy in different frequency bands; Perform Fourier transforms on the jitter signals J_x(t) and J_y(t) respectively to transform them from the time domain to the frequency domain, and obtain the power spectral density P(f), which represents the distribution of signal energy at different frequencies. Spectral entropy borrows the concept of information entropy to measure the flatness or complexity of the power spectrum. It is calculated as: H = -Σ[P_norm(f) * log(P_norm(f))], where P_norm(f) is the normalized power spectrum (summing to 1). High spectral entropy has a uniform and flat power spectrum distribution, with energy dispersed across many frequencies, resulting in a complex signal similar to white noise. Low spectral entropy has a concentrated and sharp power spectrum distribution, with energy concentrated at a few frequencies, resulting in a signal with strong regularity and periodicity. It can calculate the spectral entropy of the entire frequency band, or calculate the spectral entropy of typical physiological frequency bands (such as physiological tremor at 2-8 Hz and neurogenic micromotion at 8-30 Hz) to obtain a more refined spectral pattern. Step C4: Calculate the fractal dimension of the motion trajectory on a two-dimensional plane composed of high-frequency jitter signals; The high-frequency jitter signal J(t) = (J_x(t), J_y(t)) is regarded as the trajectory of a point on a two-dimensional plane. Fractal dimension is a measure of the complexity, roughness, or space-filling ability of a geometric shape. For a curve, its fractal dimension D satisfies 1 ≤ D ≤ 2. When D≈1, the curve is very smooth, almost a straight line. When D≈2, the curve is extremely complex, almost filling a two-dimensional region. The fractal dimension can be calculated using the box counting method, which involves covering the trajectory with squares of different side lengths ε and counting the number of squares N(ε). The fractal dimension D = -lim(ε→0) [log N(ε) / logε]. The calculation process of the fractal dimension is a traditional technique and will not be elaborated here. This study quantifies the irregularity and complexity of jitter trajectories from a geometric perspective. Physiological micro-movements are influenced by multiple neural feedback loops and muscular synergy, and their trajectories often exhibit self-similar fractal properties. Fractal dimension can capture this inherent pattern of complexity, while fake, mechanically generated jitters typically lack this natural fractal structure. Step C5: Combine the spectral entropy and fractal dimension to form the third sub-feature vector. The calculated spectral entropy value (which may be one or more frequency bands) and the fractal dimension value (calculated in the X and Y directions or two-dimensional trajectory respectively) are concatenated into a one-dimensional vector. This approach characterizes the complex patterns of physiological jitter from two entirely different perspectives: the spectral domain (spectral entropy) and the spatial geometric domain (fractal dimension). This provides a richer and more robust feature representation, compressing jitter time-series signals lasting hundreds of milliseconds into a few feature values ​​with profound physical and mathematical significance. This makes it ideal as input for machine learning models. Spectral entropy reflects the spectral energy distribution characteristics of the neuromuscular system. Fractal dimension reflects the inherent degree of chaos in its motion control. Both of these are unique and highly individualized characteristics of biological systems, making them extremely difficult to accurately simulate using non-biological means (such as motor-driven prosthetic fingers). In the embodiments corresponding to steps C1 to C5, the third sub-feature vector formed by calculating the global translation vector, filtering, spectrum analysis and fractal dimension can reflect the minute motion changes of the fingerprint during the contact process, thereby improving the capture and recognition effect of fingerprint features under different contact conditions. Step 1034: Arrange the first sub-feature vector, the second sub-feature vector, and the third sub-feature vector in chronological order to form a feature time sequence; Step 1035: Input the feature time sequence into a time-series neural network, and have the time-series neural network extract and output the time-series behavioral features. The three-dimensional sub-features are spliced ​​together at each time point to form a multi-channel time-series signal. This ensures the synchronicity and correlation of macroscopic, microscopic, and physiological signals over time. Temporal neural networks (such as LSTM, GRU, or Transformer) excel at processing sequential data. The network automatically learns the complex temporal correlations and dependencies between the three signals. For example, it might learn that at the instant the centroid begins to shift (macro-slip), a specific shearing pattern appears in the micro-displacement field, while the jitter spectrum briefly suppresses such cross-modal joint modes. The temporal behavioral features output by the network are highly abstract and compact vectors that encode the dynamic nature of the entire pressing process, exhibiting stronger discriminative power and robustness than hand-designed sub-feature vectors. Temporal neural networks are lightweight Transformer networks that separate space and time. They include a spatial feature extraction module, a position encoding module, a lightweight Transformer encoder, a temporal pooling module, and a fully connected output layer. The spatial feature extraction module consists of a weight-shared two-dimensional convolutional layer, used to extract spatial features from each frame of the standardized fingerprint image temporal sequence, obtaining a spatial feature sequence. The positional encoding module adds temporal positional encoding to the spatial feature sequence. The lightweight Transformer encoder contains at least one multi-head self-attention layer employing a local window attention mechanism, where for each temporal position in the sequence, attention computation is limited to its K adjacent temporal positions. The temporal pooling module pools the output of the Transformer encoder to obtain a sequence-level representation. The fully connected output layer generates the temporal micro-behavioral feature vector. In the embodiments corresponding to steps 1031 to 1035, through comprehensive analysis of multiple aspects such as contact area, ridge displacement, and global motion, this method can fully utilize the dynamic information of fingerprints during the contact process, enhancing the accuracy and reliability of fingerprint recognition. This multi-level feature extraction strategy enables the system to better adapt to various contact situations and improve recognition performance. Step 104: Based on the static feature vector and the temporal behavior features, identify whether the user corresponding to the image frame sequence is a device user. The two feature vectors obtained in steps 102 and 103 are fused. The fusion methods include, but are not limited to, feature-level fusion (concatenating the two vectors into a longer mixed feature vector) and score-level fusion (comparing the static features and temporal features with the registered template respectively to obtain two similarity scores, and then merging them according to weights). The fused result is compared with the user's pre-registered feature template in the device. If the similarity exceeds a preset threshold, the user is determined to be a legitimate device user; otherwise, the user is rejected. Even if static features have a slightly lower match rate due to factors like dry or dirty fingers, unique and stable temporal behavioral features can serve as a powerful supplement, improving the authentication success rate. Conversely, if static features match but behavioral features are abnormal, the system will issue a warning about potential spoofing attacks. Specifically, step 104 includes steps 1041 to 1044: Step 1041: Obtain the pre-stored static vector and pre-stored behavioral vector corresponding to the historical user; During the user registration phase, the system collects multiple (or one high-quality) fingerprint presses from the user and uses the feature extraction method of this application to extract and store the user's standard static feature vector and standard temporal behavioral feature vector. These two vectors together constitute the user's biometric template. Step 1042: Calculate the first similarity between the static feature vector and the pre-stored static vector; Step 1043: Calculate the second similarity between the temporal behavior features and the pre-stored behavior vectors; Similarity can be calculated in ways including but not limited to cosine similarity or Euclidean distance. Static similarity primarily assesses the degree of conformity of fingerprint texture; behavioral similarity primarily assesses the degree of conformity of dynamic pressing behavior. The two are evaluated independently from different dimensions. Step 1044: When both the first similarity and the second similarity are greater than the preset similarity, determine that the user corresponding to the image frame sequence is the device user. The preset similarity is a pre-defined threshold. It determines the system's tolerance for matching accuracy. A higher threshold results in higher security but may lower the pass rate (increased false positive rate). A lower threshold results in a higher pass rate but may decrease security (increased false positive rate). The preset similarity can be set to 0.78. Authentication will only pass if both static features and temporal behavioral features achieve sufficiently high similarity. Authentication passes if both the first similarity and the second similarity are greater than the threshold. Otherwise, authentication is rejected if either condition is not met. In the embodiments corresponding to steps 1041 to 1044, by acquiring pre-stored features of historical users, calculating similarity, and performing logical judgments, the system can effectively identify user identities, thereby improving device security and user experience. This process not only considers the user's static features but also integrates their behavioral patterns, making the identification more comprehensive and accurate. In the embodiments corresponding to steps 101 to 104, by acquiring image frame sequences during the user's finger contact with the sensor at a fixed frame rate, and especially by extracting static feature vectors based on the frame images corresponding to the maximum contact area, the detailed features of the user's fingerprint can be captured more accurately, reducing recognition errors caused by changes in fingerprint position or angle. Extracting temporal behavioral features allows the system to analyze the pressure diffusion, ridge deformation, and slippage patterns generated during fingerprint contact. The introduction of these dynamic features means that the recognition process not only relies on static image data but also considers the user's actual behavior during operation, thereby improving the ability to identify genuine users. Since temporal behavioral features can reflect the changing patterns of the user's fingerprint under different operating states, the system can effectively distinguish between genuine users and counterfeit fingerprints (such as silicone fingerprints). This capability significantly enhances system security and reduces the risk of identity theft. like Figure 2 This invention provides a device for fingerprint recognition. Please refer to [link / reference]. Figure 2 , Figure 2 A schematic diagram of a device fingerprint recognition apparatus provided by the present invention is shown, as follows: Figure 2 The device for fingerprint recognition shown includes: The acquisition unit 21 is used to acquire a sequence of image frames of the fingerprint to be identified at a fixed frame rate during the process of the user's finger touching and pressing the sensor until it leaves the sensor. The first extraction unit 22 is used to extract static feature vectors based on the frame image corresponding to the maximum contact area in the image frame sequence; The second extraction unit 23 is used to extract temporal behavioral features based on the image frame sequence; wherein, the temporal behavioral features are used to characterize the change patterns corresponding to fingerprint pressure diffusion, fingerprint ridge deformation and fingerprint slippage; The identification unit 24 is used to identify whether the user corresponding to the image frame sequence is a device user based on the static feature vector and the temporal behavior features. This invention provides a device for fingerprint recognition. By acquiring image frame sequences of a user's finger contacting a sensor at a fixed frame rate, and particularly extracting static feature vectors from the frame images corresponding to the maximum contact area, the device can more accurately capture the detailed features of the user's fingerprint and reduce recognition errors caused by changes in fingerprint position or angle. Extracting temporal behavioral features allows the system to analyze pressure diffusion, ridge deformation, and slippage patterns generated during fingerprint contact. The introduction of these dynamic features means that the recognition process not only relies on static image data but also considers the user's actual behavior during operation, thereby improving the ability to recognize genuine users. Since temporal behavioral features reflect the changing patterns of a user's fingerprint under different operating states, the system can effectively distinguish between genuine users and counterfeit fingerprints (such as silicone fingerprints). This capability significantly enhances system security and reduces the risk of identity theft. Figure 3 This is a schematic diagram of a terminal device provided in an embodiment of the present invention. Figure 3 As shown, a terminal device 3 in this embodiment includes: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30, such as a device fingerprint recognition program. When the processor 30 executes the computer program 32, it implements the steps in the various device fingerprint recognition method embodiments described above, for example... Figure 1 Steps 101 to 104 are shown. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each unit in the above-described device embodiments, for example... Figure 2 The function of the unit shown. For example, the computer program 32 can be divided into one or more units, which are stored in the memory 31 and executed by the processor 30 to complete the present invention. The one or more units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 32 in the terminal device 3. For example, the specific functions of each unit of the computer program 32 can be divided as follows: The acquisition unit is used to acquire a sequence of image frames of the fingerprint to be identified at a fixed frame rate during the process of the user's finger touching and pressing the sensor until it leaves the sensor; The first extraction unit is used to extract static feature vectors based on the frame image corresponding to the maximum contact area in the image frame sequence; The second extraction unit is used to extract temporal behavioral features based on the image frame sequence; wherein, the temporal behavioral features are used to characterize the change patterns corresponding to fingerprint pressure diffusion, fingerprint ridge deformation and fingerprint slippage; The identification unit is used to identify whether the user corresponding to the image frame sequence is a device user based on the static feature vector and the temporal behavior features. The terminal device includes, but is not limited to, a processor 30 and a memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of a terminal device 3 and does not constitute a limitation on a terminal device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc. The processor 30 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The memory 31 can be an internal storage unit of the terminal device 3, such as a hard disk or memory of the terminal device 3. The memory 31 can also be an external storage device of the terminal device 3, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device 3. Furthermore, the memory 31 can include both internal and external storage units of the terminal device 3. The memory 31 is used to store the computer program and other programs and data required by the roaming control device. The memory 31 can also be used to temporarily store data that has been output or will be output. It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above. This invention provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the above-described method embodiments. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention. In the embodiments provided by this invention, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof. It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations. As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [the described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once [the described condition or event] is detected," or "in response to detection of [the described condition or event]." Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized. The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for fingerprint recognition in a device, characterized in that, The method for fingerprint recognition of the device includes: During the process of the user's finger touching and pressing the sensor until it leaves the sensor, a sequence of image frames of the fingerprint to be identified is acquired at a fixed frame rate; Based on the frame image corresponding to the maximum contact area in the image frame sequence, extract the static feature vector; Based on the image frame sequence, temporal behavioral features are extracted; wherein, the temporal behavioral features are used to characterize the change patterns corresponding to fingerprint pressure diffusion, fingerprint ridge deformation and fingerprint slippage; Based on the static feature vector and the temporal behavior features, it is determined whether the user corresponding to the image frame sequence is a device user.

2. The method for device fingerprint recognition as described in claim 1, characterized in that, The step of extracting static feature vectors based on the frame image corresponding to the maximum contact area in the image frame sequence includes: Multiple detail points are extracted from the frame image corresponding to the largest contact area; wherein, the detail points include endpoints and bifurcation points; For each detail point, local texture features of the surrounding preset region are extracted to form a local texture descriptor; wherein, the local texture features include at least one of the following: ridge curvature statistical histogram, ridge-valley width ratio variation sequence, and phase information based on Gabor filtering in the preset region surrounding the detail point; The local texture descriptors of all detail points are aggregated and encoded to generate an aggregated local texture feature vector, which is then used as a static feature vector.

3. The method for device fingerprint recognition as described in claim 1, characterized in that, The step of extracting temporal behavioral features based on the image frame sequence includes: Based on the binarized contact mask of each frame image in the image frame sequence, the curves of the area, centroid position and shape parameters of the contact area changing with time are calculated, and the dynamic parameters are extracted from the curves to form the first sub-feature vector; Fingerprint image temporal sequence is enhanced and tracked for ridges, displacement field of selected feature points is calculated, and displacement pattern of finger skin is analyzed based on the displacement field to form a second sub-feature vector; Calculate the global motion vector between adjacent frame images, filter the global motion vector to separate the high-frequency jitter component, and extract the mode features of the high-frequency jitter component to form a third sub-feature vector; Arrange the first sub-feature vector, the second sub-feature vector, and the third sub-feature vector in chronological order to form a feature time sequence; The time-series feature is input into a time-series neural network, which then extracts and outputs the time-series behavioral features.

4. The device fingerprint recognition method as described in claim 3, characterized in that, The steps of calculating the curves of the area, centroid position, and shape parameters of the contact region changing over time based on the binarized contact mask of each frame image in the image frame sequence, and extracting dynamic parameters from the curves to form the first sub-feature vector include: Adaptive threshold binarization is performed on each frame image to generate a contact mask; Calculate the contact mask area and centroid coordinates for each frame image, and fit the circumscribed ellipse of the contact area to obtain its shape parameters; wherein, the shape parameters include the major axis, minor axis and orientation angle; The change curve of the contact mask area over time was obtained by nonlinear fitting. Extract the maximum contact mask area and extract the diffusion rate parameter from the variation curve; Calculate the total displacement of the centroid trajectories corresponding to all frames in the image frame sequence; Calculate the statistics of how the shape parameters of the circumscribed ellipse change over time; The first sub-feature vector is formed by combining the maximum contact mask area, diffusion rate parameter, total displacement, and the statistics.

5. The method for device fingerprint recognition as described in claim 3, characterized in that, The steps of performing ridge enhancement and tracking on the temporal sequence of fingerprint images, calculating the displacement field of selected feature points, and analyzing the displacement pattern of the finger skin based on the displacement field to form a second sub-feature vector include: A filter is used to enhance the ridge lines of each frame in the image frame sequence to obtain a ridge-enhanced image. Select a set of feature points on the texture enhancement image corresponding to the first frame image in the image frame sequence; The displacement field is obtained by tracking the positional changes of the feature points in the image frame sequence using optical flow. Based on the displacement field, the radial displacement of each feature point relative to the centroid of the contact area corresponding to the first frame image is calculated. Extract the mean and standard deviation for all radial displacements; Based on the radial displacement corresponding to multiple feature points in each frame image, calculate the kurtosis and skewness of each frame image; The mean, standard deviation, kurtosis, and skewness are used as the second sub-feature vector.

6. The method for device fingerprint recognition as described in claim 3, characterized in that, The steps of calculating the global motion vector between adjacent frame images, filtering the global motion vector to separate high-frequency jitter components, and extracting the mode features of the high-frequency jitter components to form a third sub-feature vector include: The global translation vector between adjacent frames is calculated using the phase correlation method. High-pass filtering is applied to the time series of the global translation vector to obtain a high-frequency jitter signal; Perform spectral analysis on the high-frequency jitter signal and calculate its spectral entropy in different frequency bands; Calculate the fractal dimension of the motion trajectory on a two-dimensional plane composed of high-frequency jitter signals; The spectral entropy and fractal dimension are combined to form the third sub-feature vector.

7. The method for device fingerprint recognition as described in claim 1, characterized in that, The step of identifying whether a user corresponding to an image frame sequence is a device user based on the static feature vector and the temporal behavior features includes: Obtain the pre-stored static vector and pre-stored behavioral vector corresponding to historical users; Calculate the first similarity between the static feature vector and the pre-stored static vector; Calculate the second similarity between the temporal behavior features and the pre-stored behavior vectors; When both the first similarity and the second similarity are greater than the preset similarity, the user corresponding to the image frame sequence is determined to be the device user.

8. A device for fingerprint recognition, characterized in that, The device for fingerprint recognition includes: The acquisition unit is used to acquire a sequence of image frames of the fingerprint to be identified at a fixed frame rate during the process of the user's finger touching and pressing the sensor until it leaves the sensor; The first extraction unit is used to extract static feature vectors based on the frame image corresponding to the maximum contact area in the image frame sequence; The second extraction unit is used to extract temporal behavioral features based on the image frame sequence; wherein, the temporal behavioral features are used to characterize the change patterns corresponding to fingerprint pressure diffusion, fingerprint ridge deformation and fingerprint slippage; The identification unit is used to identify whether the user corresponding to the image frame sequence is a device user based on the static feature vector and the temporal behavior features.

9. A terminal device, characterized in that, The terminal device includes: a memory, a processor, and a device fingerprint recognition program stored in the memory and executable on the processor, the device fingerprint recognition program being configured to implement the steps of the device fingerprint recognition method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the device fingerprint recognition method as described in any one of claims 1 to 7.

Images