Multi-dimensional biological characteristic visualization method and system of signature sequence

Abstract

The invention discloses a multi-dimensional biological characteristic visualization method and system for a signature sequence, and the method comprises the steps: firstly, obtaining online signature time sequence data through recording key point information in an electronic signature writing process; then, performing multi-dimensional feature decoupling calculation on the signature sequence data, and extracting implicit dynamic biological feature vectors; then, defining a visual mapping rule base, and respectively mapping the multi-dimensional biological characteristics into corresponding visual channel parameters; and finally, according to the visual channel parameters, rendering to generate a signature feature image set containing dynamic information of different dimensions. According to the method, through a regularized mapping mechanism, a dynamic pen wielding behavior which is difficult to perceive visually in a signature is converted into a high-resolution static image representation, so that the problem of isomerism of online time sequence data and offline image data in a feature space is effectively solved; and an interpretable multi-channel data basis is provided for signature authenticity identification algorithm research and artificial judicial handwriting identification.

Description

A method and system for visualizing multidimensional biometric signature sequences Technical Field

[0001] This invention belongs to the fields of computer vision, biometric recognition and data processing technology, and relates to a method and system for visualizing multi-dimensional biometric features of signature sequences. Background Technology

[0002] Handwritten signatures possess both legal validity and social recognition, exhibiting unique advantages in behavioral biometrics. Compared to traditional paper signatures, electronic signature sequence data not only records the spatial morphology of strokes but also implies richer dynamic features. These dynamic features are invaluable for revealing the signer's individual writing habits, distinguishing genuine from forged signatures, and assisting handwriting experts in forming expert opinions. However, in practical applications, methods for visualizing the dynamic features of electronic signature handwriting remain relatively limited, making it difficult to present multidimensional dynamic information completely and in detail, resulting in many features being difficult to utilize intuitively.

[0003] Electronic signature data typically records the signing process as a sequence of pen strokes, storing only dynamic information such as pen tip coordinates, pressure, and time in a time-series format. Only through echo technology can this sequence data be transformed into a handwriting image that can be visually displayed on a screen. Differences in echo algorithms and parameter settings directly affect the display effect and detail fidelity of the handwriting image. Current technologies mostly use grayscale images for handwriting echo, only showing the spatial outline of the strokes. It is difficult to simultaneously present dynamic biometric features such as pressure, speed, acceleration, and writing order in the image, limiting the full utilization of writing behavior features by handwriting identification, signature verification, and authentication algorithms. In other words, existing electronic signature visualizations suffer from the loss of dynamic features and difficulty in intuitive identification. Summary of the Invention

[0004] To address the aforementioned technical problems in the existing technology, this invention proposes a method and system for visualizing multi-dimensional biometric features of signature sequences, the specific technical solution of which is as follows:

[0005] A multi-dimensional biometric visualization rule method for signature sequences, the core of which lies in the feature decoupling and customized visual mapping of online signature sequences, including:

[0006] Step 1: Obtain online signature sequence data to be processed by recording key point information during the writing process. The signature sequence data includes a sequence of pen tip horizontal and vertical coordinate points sampled in chronological order and the original sensor signals associated with each sampled coordinate point in the sequence.

[0007] Step 2: Perform multi-dimensional feature decoupling calculation on the signature sequence data to extract implicit dynamic biometric vectors, where the dynamic biometric vectors include writing pressure features, kinematic features, and writing order features;

[0008] Step 3: Invoke a preset visual mapping rule library, where the rule library defines the conversion logic from the numerical feature space to the visual image space;

[0009] Step 4: Based on the visual mapping rule library, map the extracted dynamic biometric vectors into corresponding image visual parameters respectively;

[0010] Step 5: According to the image visual parameters, parallelly render and generate a signature feature image set containing dynamic information of different dimensions.

[0011] Further, in the above Step 1, a terminal such as a signature board or a digital tablet is used to record the writing trajectory of the signer, and a signature stroke sequence is obtained by real-time recording of the key stroke point information of the electronic signature. The sequence data at least includes dynamic signals such as the two-dimensional trajectory coordinates of the pen tip, the pen tip state, the writing pressure, and the time stamp.

[0012] Further, the sequence data is stored and represented as {(x i ,y i ,s i ,t i ,p i ,ts i ), 0 < i < n}, where (x i ,y i ) are the coordinate information of the stroke key points, n is the number of all key points, s i and t i represent the pen tip state. If s i = 1, t i = 0, it means that the pen tip is in the down state, and this drawing point is connected to the next point, that is, a continuous drawing operation; if s i = 0, t i = 1, it means that the pen tip is in the up state, and this drawing point is not connected to the next point, that is, the drawing stops.

[0013] Further, in the above Step 2, based on the original sensor signals, the pressure values of each sampling point in the sequence are extracted and normalized as the writing pressure feature; based on the first-order difference and second-order difference of the pen tip horizontal and vertical coordinate point sequences, the instantaneous horizontal velocity, instantaneous horizontal acceleration, instantaneous vertical velocity, instantaneous vertical acceleration, centripetal acceleration, tangential angle, and angular velocity of each sampling point are calculated as the kinematic features; based on the time stamp index of the sampling points, a time series encoding is generated as the writing order feature.

[0014] Furthermore, in step 3, the visualization mapping rule base includes the following rules:

[0015] Pressure feature mapping rule: Map the pen pressure feature of the sampling point to the line width or opacity of the stroke rendering, where the greater the pressure value, the greater the line width or the darker the color;

[0016] Kinematic feature mapping rules: Establish the correspondence between instantaneous velocity, acceleration, angular velocity and thermogram color, map low-speed regions to light colors and high-speed regions to dark colors, and generate thermal distribution parameters;

[0017] Stroke writing order feature mapping rule: Establish the correspondence between the time index of each sampling point in the signature sequence and the HSV color model.

[0018] Furthermore, the stroke writing order mapping rule is specifically as follows:

[0019] Mapping the hue channel of the HSV color model: mapping the time index to the hue circle angle range of 0° to 360°, so that the continuous change of hue in 0° to 360° represents the writing start and end order of strokes;

[0020] The saturation channel of the HSV color model is mapped: the length of the line connecting adjacent key points of the stroke is inversely correlated with the saturation value, so as to highlight the slow-moving areas, i.e. small and fine local strokes, with high saturation.

[0021] The luminance channel of the HSV color model is mapped: the overall features of the signature sequence are mapped to luminance values ​​to determine the global luminance benchmark for rendering the handwriting, so as to ensure the overall visual clarity of the handwriting.

[0022] Further, step 4 specifically involves: for the i-th sampling point in the sequence, where pressure features, kinematic features, and writing order features have been extracted, the corresponding steps are:

[0023] Call the pressure feature mapping rules to obtain the pixel grayscale values ​​used for the pressure rendering map;

[0024] The kinematic feature mapping rules are invoked to obtain the color parameters used for the heatmap;

[0025] The stroke writing order feature mapping rule is invoked to map the time index to parameters in the HSV color space (H). i ,S i V i ), and further converted into RGB color parameters for use in writing the coloring of the sequential coding diagram.

[0026] Further, in step 5, based on the image visual parameters, image rendering is performed on each decoupled biometric dimension to generate multiple independent signature feature images, resulting in a signature feature image set; the generated signature feature image set includes, but is not limited to: pressure rendering map, velocity heatmap, acceleration heatmap, angular velocity heatmap, centripetal acceleration heatmap, and writing order encoding map; the signature feature image set is output for feature analysis of signature verification, input to automatic verification models, or manual assisted identification.

[0027] A multi-dimensional biometric visualization system for signature sequences includes:

[0028] The data acquisition module is used to acquire online signature sequence data;

[0029] The feature calculation module is used to perform multi-dimensional feature decoupling calculation on the signature sequence data and extract the implicit dynamic biometric feature vector.

[0030] The rule mapping module is used to call a preset visualization mapping rule library, which defines the mapping relationship between the numerical feature space and the visual image space, and maps the dynamic biological feature vector to the corresponding image visual parameters.

[0031] The rendering and generation module is used to render and generate a set of signature feature images containing dynamic information in different dimensions based on the image visual parameters.

[0032] The advantages and beneficial effects of this invention are as follows: it decouples the temporal signal of online signatures into multi-dimensional dynamic biometric features, and generates a structurally unified feature image set through a regularized mapping strategy. This achieves a natural transformation of signature behavior from sequential data that is difficult to use intuitively to image data that can be directly used for visual analysis and deep learning modeling. The signature feature images corresponding to each biometric dimension can be generated independently and combined for use, providing a unified input format for image convolutional networks and intuitive behavioral clues for handwriting identification. This is beneficial for improving the accuracy and transparency of signature verification in scenarios such as forensic identification and financial risk control. This system has low deployment costs and is suitable for different terminal devices and various signature collection environments. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0034] Figure 1 is a structural block diagram of a multi-dimensional biometric visualization system for signature sequences provided in an embodiment of the present invention;

[0035] Figure 2 is an example diagram of the visualization image set of partial features of the signature sequence according to an embodiment of the present invention;

[0036] Figure 3 is an example diagram of the annotation of the turning points of the signature strokes according to an embodiment of the present invention. Specific Embodiment

[0037] In order to make the purpose, technical solution and technical effects of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings of the specification and embodiments.

[0038] This embodiment discloses a multi-dimensional biometric visualization method for signature sequences, providing a general and regular signature time-series data feature mapping and rendering mechanism, decoupling the time-series signals obtained from online signature collection into multi-dimensional dynamic biometric features, and mapping them into a set of signature feature image sets with unified structure and clear semantics according to preset visualization rules, so as to improve the expression ability and interpretability of signature behavior features and provide an intuitive basis for subsequent automatic verification and manual identification. The method mainly includes the following steps:

[0039] Step 1: Online signature sequence data collection.

[0040] For the user signature input terminal, the system first obtains the online signature sequence data to be processed by recording the key point information during the signature writing process. The signature sequence data includes a sequence of pen tip horizontal and vertical coordinate points sampled in chronological order and the original sensor signals associated with the coordinate points.

[0041] In this embodiment, a signature board, a digital drawing tablet or a tablet computer with a handwriting function is used as the terminal. The terminal records the key point information of the electronic signature in real time to obtain the signature stroke sequence, and the sequence data is stored and represented as a set D = {(x i , y i , s i , t i , p i , ts i ) | 0 < i < n}. Where:

[0042] (x i , y i ) are the two-dimensional coordinates of the stroke key points in the screen coordinate system;

[0043] n is the number of all key points;

[0044] p i is the original pressure value captured by the sensor;

[0045] ts i is the sampling timestamp;

[0046] si and t i Jointly representing the pen tip state: If s i =1,t i =0 indicates that the pen tip has not been lifted, and the drawing point is connected to the next point, i.e., continuous drawing operation; if s i =0,t i =1 indicates that the pen tip is lifted, meaning that the point being drawn is not connected to the next point by a line in space, but is still recorded as air-stroke data.

[0047] Step 2: Decoupling and computation of multi-dimensional features.

[0048] After preprocessing the collected raw sequence data, multi-dimensional feature decoupling calculations are performed to extract implicit dynamic biological feature vectors. Specifically, this includes:

[0049] Step 201: Extraction of pen pressure features. In the following formulas, ε represents a very small positive number to prevent division by zero.

[0050] Read the raw pressure signal p i By mapping it to the [0,1] interval using a normalization formula, the normalized pressure feature p' is obtained. i The formula is as follows:

[0051]

[0052] normalized p' i The values ​​∈[0,1] constitute the pen pressure feature in this embodiment, which is used to control the line width or opacity later.

[0053] Step 202: Kinematic feature extraction.

[0054] In this embodiment of the invention, the kinematic features of the pen tip are calculated based on a time-series coordinate sequence. To improve the numerical stability of the features and reduce the influence of sampling noise, the derivative is calculated using the central difference method for intermediate sampling points; for the first and last sampling points, forward and backward differences are used for supplementary calculations to ensure that effective kinematic features can be obtained for the entire sequence. The application of central difference and forward / backward difference is illustrated using the calculation of instantaneous velocity and acceleration as an example; the application in the calculation of pen movement tangential angle and angular velocity will not be explained separately.

[0055] Specifically, within the continuous writing interval excluding stroke endpoints, the instantaneous velocity and acceleration in the horizontal and vertical directions of the i-th sampling point are calculated as follows:

[0056]

[0057] The aforementioned center difference utilizes information from two sampling points, resulting in higher accuracy and smoothness compared to single-point difference, and effectively suppressing high-frequency jitter.

[0058] For the sequence termination point i = n-1 and the pen break point i = b, since there is no subsequent sampling point, backward difference estimation is used:

[0059]

[0060] For the sequence starting point i = 0 and the first point i = b + 1 after the break, since there is no previous sampling point, forward difference estimation is used:

[0061]

[0062] Similarly, calculate the angle of the pen stroke direction. and angular velocity w i :

[0063]

[0064] Calculate centripetal acceleration:

[0065]

[0066] Step 203: Extraction of stroke writing order features.

[0067] Traverse the sequence and find all sequences that satisfy s i The set of indices equal to 0 is denoted as:

[0068] I lift ={i|s i =0}

[0069] If there is no stopping point, i.e., I lift If the value is empty, it is assumed that the entire sequence contains only one entry, and a default set of indices containing the beginning and end is constructed for subsequent processing.

[0070] If a stroke stops, increment the index of each stopping point by 1 to obtain the corresponding next sampling point, which serves as the starting index for the next stroke. Then, insert 0s at the beginning of the index set to obtain the stroke starting index sequence:

[0071] S = {s0, s1, ..., s} M}

[0072] Where: s0 = 0 is the starting point of the first stroke, and the remaining elements are the starting point indices of each stroke.

[0073] The points between two adjacent starting indices form a writing trajectory, which is used for subsequent line segment calculations and visualization.

[0074] Step 3: Building a visual mapping rule base.

[0075] A pre-defined visualization mapping rule library is introduced to define the transformation logic from numerical feature space to visual image space. A unified specification is established for the correspondence between numerical feature space and visual image space, used to map dynamic biometric features of different dimensions into image visual parameters such as line width, opacity, color, and brightness.

[0076] This embodiment specifically adopts the following mapping rules:

[0077] Step 301: Pressure Feature Mapping Rules.

[0078] In this embodiment, the normalized pressure feature is mapped to grayscale values, and lines are drawn segment by segment in stroke order to achieve the visualization effect that the greater the pressure, the darker the stroke. For each stroke segment, the average pressure of two adjacent sampling points is calculated:

[0079]

[0080] The average pressure is mapped to a grayscale value in the range of 0–255. To preserve some contrast, this embodiment introduces a scaling factor α∈(0,1), for example, α=0.9, resulting in:

[0081]

[0082] Step 302: Kinematic feature mapping rules.

[0083] In this embodiment, various kinematic features calculated from the signature sequence are converted into pseudo-color thermal distributions on the stroke path using a preset color mapping rule, thereby achieving intuitive visualization of the signature's dynamic behavior.

[0084] For non-symbolic kinematic features such as velocity, angular velocity, and centripetal acceleration, this embodiment employs a linear mapping normalized to the [0,1] interval and uses a monotonically increasing sequential coloring spectrum. For the normalized feature u... i The preset yellow-orange-red thermochromatic color lookup table is invoked and mapped to RGB color values:

[0085] color i =cmap yellow→red (u i ),u i ∈[0,1]

[0086] The color spectrum gradually transitions from bright yellow, golden yellow, and orange to bright red and deep red, achieving a visual effect where the larger the feature value, the redder and deeper the color.

[0087] Color mapping is performed based on symbolic kinematic characteristics such as horizontal / vertical velocity and horizontal / vertical acceleration. This embodiment employs zero-centered symmetric normalization and divergent chromatogram mapping. A zero-centered divergent chromatogram mapping rule is defined, mapping negative values ​​to the blue family and positive values ​​to the yellow-red family. The area near zero corresponds to neutral cyan or light colors to highlight the difference between positive and negative directions.

[0088] color i =cmap blue→cyan→yellow→red (u i ),u i ∈[-1,1]

[0089] The cold end of the color spectrum is dark blue / blue, the middle region is cyan, and the warm end is yellow, orange, red, and dark red.

[0090] Step 303: Stroke writing order feature mapping rules.

[0091] In this embodiment, to explicitly encode the writing order information of online signatures in image form, a writing order feature mapping rule based on the HSV color space is designed. This rule maps the time sequence, line segment length, and position of the last stroke in the handwriting trajectory to the three channels of hue, saturation, and value in the HSV color space, respectively, thereby generating a feature image containing dynamic writing information.

[0092] In the stroke segment, adjacent sampling points p i With p i+1 This forms a trajectory line segment, and the square of the length of the i-th line segment is defined as:

[0093]

[0094] Take the maximum value among the squares of all line segment lengths:

[0095]

[0096] First, we introduce the relative position of the last stroke within the entire writing sequence—this global writing order information. In this embodiment, we set the brightness channel to a uniform constant across the entire image, controlling the visual clarity of the entire handwriting through a global brightness benchmark. Let the starting index of the last stroke be:

[0097] k last =s M-1

[0098] The brightness V of the entire image is then defined as:

[0099]

[0100] According to the above formula, if the last stroke appears early, the V value will be low and the overall image will be dark; if the last stroke appears late, the V value will be high and the overall image will be brighter.

[0101] Secondly, to achieve visual encoding of the writing order, this embodiment uses segment indices as the timeline, mapping them linearly to the hue channel. Let the index of the i-th line segment be i, then the hue value is defined as:

[0102]

[0103] Among them, hue H i The value range is [0, 255], used to match the encoding format under HSV full hue mode (HSV_FULL). Through this linear mapping, the line segments at the beginning of writing correspond to lower hue values, and the line segments at the end of writing correspond to higher hue values, thereby realizing the gradual change of color along the writing order and intuitively expressing time information.

[0104] Finally, to reflect the geometric motion characteristics of each line segment, this embodiment maps the relative lengths of the line segments to a saturation channel. The saturation of the i-th line segment is defined as:

[0105]

[0106] According to the above definition, when the length of a certain line segment is relatively short, Smaller, S i A value close to 255 corresponds to high saturation and vibrant colors; when the line segment length approaches its maximum value... S i A value close to 0 corresponds to low saturation, and the color tends towards grayscale. This allows small, fine details to be highlighted with high saturation, while long, fast-moving strokes are presented with low saturation.

[0107] Step 4: Mapping feature vectors to image visual parameters.

[0108] Based on the constructed visualization mapping rule library, this embodiment queries the visualization mapping rule library according to the dynamic biometric vector corresponding to each sampling point, and converts the numerical features into image visual parameters one by one.

[0109] Specifically, for the i-th sampling point in the sequence, information such as pressure features, velocity / acceleration features, and time index has been extracted. Then, the following steps are taken:

[0110] By applying the pressure feature mapping rules, the pixel grayscale values ​​g used for the pressure rendering map are obtained. j ;

[0111] By applying the kinematic feature mapping rules, the color parameter (color) for the heatmap is obtained. i ;

[0112] The stroke writing order feature mapping rule is invoked to map the time index to parameters in the HSV color space (H). i ,S i V i ), and further converted into RGB color parameters for use in writing the coloring of the sequential coding diagram.

[0113] The above mapping results can be organized into a parameter list for each type of feature image, for example: stress map parameter list {(x i ,y i ,g j List of parameters for velocity heatmap Writing order encoding graph parameter list {(x i ,y i H i ,S i V i Then these parameters are output for rendering to drive the subsequent image generation process.

[0114] Step 5: Parallel rendering and image set generation.

[0115] An image set is generated on a white canvas of a specified size, such as a 448×448 pixel white canvas. Based on the image visual parameters, each decoupled biometric dimension is rendered independently to generate multiple independent signature feature images, resulting in a signature feature image set.

[0116] The specific rendering and generation process is as follows:

[0117] Create an image of a specified size, initialize its pixel values ​​to a white background, and draw line segments one by one using a line segment-level coloring strategy based on the stroke segment index list obtained in step 203. Combine adjacent sampling point pairs into a line segment set, and draw line segments on the canvas according to the image visual parameters with a uniform line width.

[0118] To help observers understand the correspondence between color and numerical value, a color bar is placed next to the heatmap, with the scale of the color bar corresponding to the normalization range. The color change order is marked below the stroke writing sequence diagram, and statistical quantities such as the average value of kinematic features can also be marked above the feature diagram, thus providing an intuitive and quantitative visual reference for the dynamic behavior of signature writing.

[0119] The generated signature feature image set includes, but is not limited to: pressure rendering map, velocity heat map, acceleration heat map, angular velocity heat map, centripetal acceleration heat map, and writing order encoding map; the signature feature image set is output for feature analysis of signature verification, input of automatic verification model, or manual assisted identification.

[0120] Through the parallel rendering mechanism described above, this invention can visualize online signature time-series trajectory data into various feature images in a short time, providing a high-quality data foundation for subsequent signature authentication, feature learning, and model training.

[0121] In some embodiments, as shown in Figure 2, in addition to generating the aforementioned feature images, this method can also draw corresponding time-domain waveforms and frequency-domain spectra based on Fast Fourier Transform (FFT) for feature sequences such as velocity, acceleration, angular velocity, and pressure. This allows for a visual representation of the distribution of kinematic features in the time and frequency domains in image form, providing an auxiliary basis for manual inspection and subsequent feature analysis.

[0122] In a further embodiment, as shown in Figure 3, this method can also generate specialized labeled images for turning points in the stroke trajectory: First, turning points are detected based on the angle of change in stroke direction and a preset threshold; then, using these turning points as boundaries, the trajectory is divided into several adjacent stroke segments, and the turning points are explicitly marked during the rendering process. For adjacent stroke segments, a first color and a second color are used alternately, for example, red and blue are used to distinguish them, so that the stroke segments before and after each turning point form a clear visual contrast, making it easier for the observer to quickly identify the structural changes and dynamic characteristics during the writing process.

[0123] Corresponding to the aforementioned embodiments of the multi-dimensional biometric visualization rule method for signature sequences, this embodiment also provides a multi-dimensional biometric visualization rule system for signature sequences. As shown in Figure 1, the system includes: a data acquisition module for acquiring online signature sequence data to be processed; a feature calculation module for performing multi-dimensional feature decoupling calculation on the signature sequence data to extract implicit dynamic biometric vectors; a rule mapping module for calling a preset visualization mapping rule library to map the dynamic biometric vectors to corresponding image visual parameters; and a rendering generation module for rendering and generating signature feature image sets containing dynamic information of different dimensions on their respective canvases according to the image visual parameters, and outputting the image sets to a display terminal or a downstream analysis module.

[0124] The above modules can be deployed as software functional modules in the same electronic device, or they can be deployed in a distributed manner among multiple devices to achieve collaborative work through network communication.

[0125] The above 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 or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for visualizing multi-dimensional biometric features of signature sequences, characterized in that, The process includes: Step 1: Obtaining online signature sequence data to be processed by recording key point information during the writing process. The signature sequence data includes a sequence of pen tip horizontal and vertical coordinate points sampled in chronological order and the original sensor signals associated with each sampled coordinate point in the sequence; Step 2: Performing multi-dimensional feature decoupling calculation on the signature sequence data to extract implicit dynamic biometric vectors. The dynamic biometric vectors include pen pressure features, kinematic features, and writing order features. Step 3: Call the preset visualization mapping rule library, which defines the conversion logic from numerical feature space to visual image space; Step 4: Based on the visualization mapping rule library, map the extracted dynamic biometric vectors to the corresponding image visual parameters; Step 5: Based on the image visual parameters, generate a signature feature image set containing dynamic information of different dimensions in parallel rendering.

2. The method for visualizing multi-dimensional biometric features of signature sequences according to claim 1, characterized in that: In step 1, the terminal records the signer's writing trajectory, and obtains the signature stroke sequence by recording the key stroke information of the electronic signature in real time. The sequence data includes at least the two-dimensional trajectory coordinates of the pen tip, the pen tip state, the writing pressure, and the timestamp signal.

3. The method for visualizing multi-dimensional biometric features of signature sequences according to claim 2, characterized in that: The pen tip state is divided into pen tip down state and pen tip up state. In the pen tip down state, the drawing point is connected to the next point, that is, the current drawing operation is continuous; in the pen tip up state, the drawing point is not connected to the next point, that is, the current drawing is stopped.

4. The method for visualizing multi-dimensional biometric features of signature sequences according to claim 1, characterized in that: In step 2, based on the original sensor signal, the pressure value of each sampling point in the sequence is extracted and normalized as the pen pressure feature. Based on the first and second differences of the pen tip's horizontal and vertical coordinate point sequence, the instantaneous horizontal velocity, instantaneous horizontal acceleration, instantaneous vertical velocity, instantaneous vertical acceleration, centripetal acceleration, tangential angle, and angular velocity of each sampling point are calculated as the kinematic features; based on the timestamp index of the sampling points, a time series code is generated as the writing order feature.

5. The method for visualizing multi-dimensional biometric features of signature sequences according to claim 4, characterized in that: In step 3, the visualization mapping rule base includes the following rules: Pressure feature mapping rule: Mapping the pen pressure features of the sampling points to the line width and opacity of the stroke rendering, where the greater the pressure value, the greater the line width and the darker the color; Kinematic feature mapping rule: Establishing the correspondence between instantaneous velocity, acceleration, angular velocity and heat map color, mapping low-speed areas to light colors and high-speed areas to dark colors, generating thermal distribution parameters; Stroke writing order feature mapping rule: Establishing the correspondence between the time index of each sampling point in the signature sequence and the HSV color model.

6. The method for visualizing multi-dimensional biometric features of signature sequences according to claim 5, characterized in that: The stroke writing order mapping rule is specifically as follows: the hue channel of the HSV color model is mapped: the time index is mapped to the hue circle angle range of 0° to 360°, and the continuous change of hue in 0° to 360° represents the writing start and end order of the strokes. The saturation channel of the HSV color model is mapped: the length of the line connecting adjacent key points of the stroke is inversely correlated with the saturation value, so as to highlight the slow stroke area with high saturation. The lightness channel of the HSV color model is mapped: the overall features of the signature sequence are mapped to lightness values ​​to determine the global lightness reference for rendering the handwriting.

7. The method for visualizing multi-dimensional biometric features of signature sequences according to claim 5, characterized in that: Step 4 specifically involves: for the pressure features, kinematic features, and writing order features extracted from the sampling points in the sequence, the following steps are taken: calling the pressure feature mapping rule to obtain the pixel grayscale values ​​for the pressure rendering map; calling the kinematic feature mapping rule to obtain the color parameters for the heatmap; and calling the stroke writing order feature mapping rule to map the time index to parameters in the HSV color space and further convert them to RGB color parameters.

8. The method for visualizing multi-dimensional biometric features of signature sequences according to claim 5, characterized in that: In step 5, based on the image visual parameters, image rendering is performed on each decoupled biometric dimension to generate multiple independent signature feature images. The signature feature images include: pressure rendering map, velocity heat map, acceleration heat map, angular velocity heat map, centripetal acceleration heat map, and writing order encoding map.

9. A system employing the multi-dimensional biometric visualization method for signature sequences according to any one of claims 1 to 8, characterized in that, include: The data acquisition module is used to acquire online signature sequence data; The feature calculation module is used to perform multi-dimensional feature decoupling calculation on the signature sequence data and extract the implicit dynamic biometric feature vector; the rule mapping module is used to call the preset visualization mapping rule library, which is used to define the mapping relationship between the numerical feature space and the visual image space, and map the dynamic biometric feature vector to the corresponding image visual parameters. The rendering and generation module is used to render and generate a set of signature feature images containing dynamic information in different dimensions based on the image visual parameters.

Images