A Mongolian horse gait cycle segmentation method, device, equipment and medium

Abstract

The application discloses a Mongolian horse gait cycle segmentation method, device, equipment and medium, including: constructing a three-dimensional skeleton model of a target Mongolian horse; according to the three-dimensional skeleton model of the target Mongolian horse, performing characteristic signal extraction and selection; taking a standard gait cycle set as a template sequence, taking continuous characteristic signals as a continuous test sequence, and aligning the template sequence and the continuous test sequence through local distance calculation, cumulative cost matrix construction and minimum regular path search; introducing a multi-template matching strategy and an improved constraint window strategy to construct an improved dynamic time warping algorithm, which is used for optimizing the alignment process of the template sequence and the continuous test sequence; and determining the start and end positions of a single gait cycle according to an optimal matching interval determined by the improved dynamic time warping algorithm and in combination with local extreme value characteristics. The application belongs to the field of gait segmentation. The application can accurately segment the movement cycle of the Mongolian horse gait.

Description

Technical Field

[0001] This invention relates to the field of gait segmentation, and more particularly to a method, apparatus, device, and medium for segmenting the gait cycle of Mongolian horses. Background Technology

[0002] The Mongolian horse, an important local breed in my country, is characterized by its strong endurance, high adaptability, and good movement stability. Its gait characteristics can reflect an individual's athletic ability, health status, and training effectiveness. Therefore, accurately segmenting the gait cycle during the Mongolian horse's movement is a crucial foundation for gait feature extraction, kinematic parameter analysis, and intelligent health assessment.

[0003] Currently, animal gait analysis mainly relies on manual observation or signal segmentation methods based on fixed time windows. These methods are easily affected by subjective factors of the observer, changes in movement speed, and individual differences, making it difficult to accurately determine the boundaries of the gait cycle during continuous movement. Therefore, how to accurately delineate the boundaries of the gait cycle is an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and medium for segmenting the gait cycle of Mongolian horses, which solves the technical problem of inaccurate gait cycle boundary segmentation in the prior art and achieves the technical effect of accurately segmenting the gait cycle boundary.

[0005] In a first aspect, the present invention provides a method for segmenting the gait period of Mongolian horses, comprising: Construct a three-dimensional skeletal model of the target Mongolian horse; Based on the three-dimensional skeletal model of the target Mongolian horse, characterization signals are extracted and selected, which are used to reflect the gait rhythm change characteristics and phase transition characteristics of the target Mongolian horse. The standard gait cycle set is used as the template sequence, and the continuous characterization signal is used as the continuous test sequence. The template sequence is aligned with the continuous test sequence through local distance calculation, cumulative cost matrix construction and minimum warping path search. The standard gait cycle set contains multiple complete standard gait cycles. A multi-template matching strategy and an improved constraint window strategy are introduced to construct an improved dynamic time warping algorithm, which is used to optimize the alignment process between the template sequence and the continuous test sequence. The optimal matching interval is determined by the improved dynamic time warping algorithm, and the start and end positions of a single gait cycle are determined by combining local extreme value features. The start and end positions of the gait cycle are used for gait cycle segmentation of the target Mongolian horse.

[0006] Furthermore, a three-dimensional skeletal model of the target Mongolian horse is constructed, including: Obtain the skeletal data and body size parameters of the target Mongolian horse, and construct an initial three-dimensional skeletal model; Based on the spatial location and connection relationship of each bone in the skeletal data, the initial three-dimensional skeletal model is divided into bone segments to obtain multiple independent bone segments; The skeletal parameters of each independent bone segment are adjusted to match the initial three-dimensional skeletal model with the actual skeletal structure of the target Mongolian horse, resulting in a three-dimensional skeletal model. The skeletal parameters include length, angle, joint center position, mass, and center of gravity.

[0007] Furthermore, based on the three-dimensional skeletal model of the target Mongolian horse, representational signals are extracted and selected, including: Extract the 3D coordinate information corresponding to the marker points of the metatarsophalangeal joints of the hind limb in the 3D skeletal model, including: , in, Mark the point at Three-dimensional coordinate information at time, The marked points are respectively Always axis, shaft and Displacement of the axis; Differentiating the three-dimensional coordinate information yields... The velocity and acceleration signals in the axial direction include: , , in, For three-dimensional coordinate information in Moment Velocity signal in the axial direction For three-dimensional coordinate information in Moment Acceleration signal in the axial direction; The acceleration signal is low-pass filtered and normalized to obtain the characterization signal, including: , in, For the first Each sampling point is Moment Acceleration signal in the axial direction, This represents the number of sampling points; , in, Angular frequency, The cutoff angular frequency, Let the filter order be . It is the frequency response function. The imaginary unit; , in, The cutoff frequency, The sampling frequency of the marker points, For the first Normalized cutoff frequency for each sampling point; , in, For filter molecules, These are the denominator coefficients of the filter. For the first The acceleration signal after low-pass filtering of each sampling point It is a bidirectional zero-phase filter function; , in, For the standardized first Acceleration signal at each sampling point The mean of the signal. Let be the standard deviation of the signal, where the standardized acceleration signal is used as the characterization signal.

[0008] Furthermore, using a standard gait cycle set as a template sequence and continuous representation signals as continuous test sequences, the template sequence is aligned with the continuous test sequences through local distance calculation, cumulative cost matrix construction, and minimum warping path search, including: Define continuous test sequences and template sequences, including: , , in, This is a vector representation of a continuous test sequence, which is a truncated and standardized sequence. The acceleration signal at each sampling point may contain part or all of the acceleration signal. , The vector representation of the template sequence. For the first test in a continuous test sequence The representation signal corresponding to each sampling point For the template sequence of the first The representation signal corresponding to each sampling point; Calculating the local distance between the template sequence and the test sequence includes: , , , in, It is a set of temporal and representative signals of the template sequence. The time series and representation signal set of continuous test sequences. For the time of the template sequence, For the time of continuous test sequences, Local distance; Constructing a cumulative cost matrix based on the local distance includes: , in, For the cumulative cost matrix, For local distance, This is the cumulative cost matrix function; The optimal matching path between the template sequence and the test sequence is determined through minimum regularity path search, wherein boundary constraints, continuity constraints, and monotonicity constraints are satisfied during the search, including: , in, To standardize the path, For the first One point element; , in, This is the starting element.

[0009] Furthermore, including: Boundary constraints include: , Continuity constraints include: , in, For the first in the normalized path The coordinates of each element; Monotonicity constraints include: .

[0010] Furthermore, a multi-template matching strategy and an improved constraint window strategy are introduced to construct an improved dynamic time warping algorithm. This improved dynamic time warping algorithm optimizes the alignment process between the template sequence and the continuous test sequence, including: The matching cost between the template sequence and the continuous test sequence is calculated, and the optimal template is determined based on the minimum matching cost. The optimal template is used for gait cycle localization, boundary correction, and determination of the optimal matching interval. The constraint window range is adjusted based on the length difference between the template sequence and the continuous test sequence, where the window range is used to limit the search area for regularized paths to optimize the alignment process.

[0011] Furthermore, based on the optimal matching interval determined by the improved dynamic time warping algorithm, and combined with local extremum features, the start and end positions of a single gait cycle are determined, including: Based on the optimal matching interval and local extreme value characteristics, the boundary position of the optimal matching interval is corrected; The start and end positions of a single gait cycle are determined based on the corrected boundary positions.

[0012] Secondly, the present invention provides a Mongolian horse gait period segmentation device, comprising: The model building module is used to build a three-dimensional skeletal model of the target Mongolian horse; The characterization signal extraction module is used to extract and select characterization signals based on the three-dimensional skeletal model of the target Mongolian horse. The characterization signals are used to reflect the gait rhythm change characteristics and phase transition characteristics of the target Mongolian horse. The template alignment module is used to take the standard gait cycle set as the template sequence and the continuous characterization signal as the continuous test sequence, and align the template sequence with the continuous test sequence through local distance calculation, cumulative cost matrix construction and minimum warping path search. The standard gait cycle set contains multiple complete standard gait cycles. The alignment improvement module is used to introduce a multi-template matching strategy and an improved constraint window strategy to build an improved dynamic time warping algorithm. The improved dynamic time warping algorithm is used to optimize the alignment process between the template sequence and the continuous test sequence. The period segmentation module is used to determine the start and end positions of a single gait cycle based on the optimal matching interval determined by the improved dynamic time warping algorithm and combined with local extreme value features. The start and end positions of the gait cycle are used for gait cycle segmentation of the target Mongolian horse.

[0013] Thirdly, the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to execute a Mongolian horse gait period segmentation method as provided in the first aspect.

[0014] Fourthly, the present invention provides a non-transitory computer-readable storage medium that, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to perform a Mongolian horse gait period segmentation method as provided in the first aspect.

[0015] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention establishes template samples under different gait conditions, combines local distance calculation, cumulative cost matrix construction, and minimum regularization path search to achieve optimal alignment between template sequences and continuous test sequences. At the same time, it introduces a multi-template matching strategy and an improved constraint window to enhance the algorithm's adaptability to gait sequences of different lengths and local rhythm changes, and combines the local extremum features of the signal to complete periodic boundary correction.

[0016] This invention can accurately identify the periodic boundaries of continuous gait signals of Mongolian horses, providing reliable data for subsequent gait feature extraction, gait recognition, and mining of gait features related to movement performance. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a method for segmenting the gait period of a Mongolian horse provided by the present invention; Figure 2 A schematic diagram of the original acceleration signal provided by this invention; Figure 3 A schematic diagram of the filtered acceleration signal provided by the present invention; Figure 4 A flowchart illustrating the improved gait cycle segmentation method provided by the present invention; Figure 5 This is a schematic diagram of template samples under different motion states provided by the present invention; Figure 6 A comparative diagram showing the difference before and after the introduction of the improved constraint window strategy provided by this invention; Figure 7 This is a schematic diagram illustrating the determination of the start and end points of the gait cycle and the result of the cycle based on the improved DTW provided by the present invention. Detailed Implementation

[0019] This invention provides a method for segmenting the gait cycle of Mongolian horses, which solves the technical problem of inaccurate gait cycle boundary division in the prior art.

[0020] The technical solution of this invention is to solve the above-mentioned technical problems, and the overall idea is as follows: A method for segmenting the gait cycle of Mongolian horses includes: constructing a three-dimensional skeletal model of a target Mongolian horse; extracting and selecting representation signals based on the three-dimensional skeletal model of the target Mongolian horse, wherein the representation signals are used to reflect the gait rhythm variation characteristics and phase transition characteristics of the target Mongolian horse; using a set of standard gait cycles as a template sequence and continuous representation signals as continuous test sequences, and aligning the template sequence with the continuous test sequences through local distance calculation, cumulative cost matrix construction, and minimum warping path search, wherein the set of standard gait cycles contains multiple complete standard gait cycles; introducing a multi-template matching strategy and an improved constraint window strategy to construct an improved dynamic time warping algorithm, which is used to optimize the alignment process between the template sequence and the continuous test sequences; determining the start and end positions of a single gait cycle based on the optimal matching interval determined by the improved dynamic time warping algorithm and combined with local extremum features, wherein the start and end positions of the gait cycle are used for gait cycle segmentation of the target Mongolian horse.

[0021] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0022] First, it should be clarified that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0023] This invention provides, for example Figure 1 The method for segmenting the gait period of Mongolian horses shown includes steps S11-S15: Step S11: Construct a three-dimensional skeletal model of the target Mongolian horse.

[0024] The process includes: acquiring the skeletal data and body size parameters of the target Mongolian horse, and constructing an initial three-dimensional skeletal model; dividing the initial three-dimensional skeletal model into multiple independent skeletal segments based on the spatial position and connection relationship of each bone in the skeletal data; adjusting the skeletal parameters of each independent skeletal segment to match the initial three-dimensional skeletal model with the actual skeletal structure of the target Mongolian horse, and obtaining a three-dimensional skeletal model. The skeletal parameters include length, angle, joint center position, mass, and center of gravity.

[0025] Skeletal data is used to describe the spatial morphology, size, and interconnections of the various skeletal structures of the target Mongolian horse. It can be obtained by scanning the complete skeleton of the target Mongolian horse with a handheld 3D scanning device. The scanning results are then processed into point cloud, surface reconstruction, and 3D model generation to obtain an initial 3D skeletal model containing major skeletal structures such as the skull, cervical vertebrae, thoracic vertebrae, lumbar vertebrae, pelvis, femur, tibia, and metatarsals.

[0026] Simultaneously, body size parameters such as height, length, chest circumference, and limb length of the target Mongolian horse are collected. These body size parameters are used as model size constraint information to calibrate the overall proportion of the initial three-dimensional skeletal model, ensuring that the model is consistent with the individual characteristics of the target Mongolian horse in terms of spatial scale.

[0027] By combining skeletal spatial geometry information and body size parameters, an initial three-dimensional skeletal model with a realistic morphological basis can be established.

[0028] Based on the anatomical structure, joint connections, and motor function relationships of the Mongolian horse skeleton, multiple bones in the initial three-dimensional skeletal model were hierarchically divided and functionally classified. Bones with related motor characteristics were merged into independent bone segments. For example, the skull and mandible were merged to form the head bone segment, the thoracic vertebrae, ribs, sternum, and scapula were merged to form the trunk bone segment, and the lumbar vertebrae, sacrum, and coccyx were merged to form the tail bone segment. At the same time, the independence of the limb bones such as the femur, tibia, and metatarsals was maintained to ensure the accurate expression of joint movement relationships in subsequent gait analysis.

[0029] After completing the skeletal segment division, the skeletal parameters of each independent skeletal segment were further adjusted to match the initial three-dimensional skeletal model with the actual skeletal structure of the target Mongolian horse. The length, spatial orientation angle, and joint center position of each bone segment can be adjusted according to actual measurement data. Mass parameters and center of gravity coordinates can be configured for each bone segment in Visual3D, so that the model has a mass distribution and motion constraint relationship that conforms to the real biomechanical characteristics.

[0030] Ultimately, a three-dimensional skeletal model was obtained that reflects the true skeletal structure, spatial connections, and kinematic characteristics of the target Mongolian horse.

[0031] Step S12: Based on the three-dimensional skeletal model of the target Mongolian horse, characterization signals are extracted and selected, wherein the characterization signals are used to reflect the gait rhythm change characteristics and phase transition characteristics of the target Mongolian horse.

[0032] Specifically, it includes: Extract the 3D coordinate information corresponding to the marker points of the metatarsophalangeal joints of the hind limb in the 3D skeletal model, including: , in, Mark the point at Three-dimensional coordinate information at time, The marked points are respectively Always axis, shaft and Displacement of the axis; Differentiating the three-dimensional coordinate information yields... The velocity and acceleration signals in the axial direction include: , , in, For three-dimensional coordinate information in Moment Velocity signal in the axial direction For three-dimensional coordinate information in Moment Acceleration signal in the axial direction; The acceleration signal is low-pass filtered and normalized to obtain the characterization signal, including: , in, For the first Each sampling point is Moment Acceleration signal in the axial direction, This represents the number of sampling points; , in, Angular frequency, The cutoff angular frequency, Let the filter order be . It is the frequency response function. The imaginary unit; , in, The cutoff frequency, The sampling frequency of the marker points, For the first Normalized cutoff frequency for each sampling point; , in, For filter molecules, These are the denominator coefficients of the filter. For the first The acceleration signal after low-pass filtering of each sampling point It is a bidirectional zero-phase filter function; , in, For the standardized first Acceleration signal at each sampling point The mean of the signal. Let be the standard deviation of the signal, where the standardized acceleration signal is used as the characterization signal.

[0033] Figure 2 This is a schematic diagram of the original acceleration signal. Figure 3 This is a schematic diagram of the filtered acceleration signal.

[0034] Based on the target Mongolian horse three-dimensional skeleton model constructed in step S11, the marker point data collected during motion capture is mapped to the three-dimensional skeleton model to determine the actual anatomical position corresponding to each marker point. Based on the motion characteristics of different skeletal parts during the gait of the Mongolian horse, key motion parts that can effectively reflect the gait cycle changes are selected as signal extraction positions.

[0035] Since the gait cycle of the Mongolian horse is mainly manifested as a cyclical process of alternating weight-bearing, propulsion, and swinging of the limbs, the hind limbs have obvious periodic movement characteristics in supporting the body weight, generating propulsion power, and completing the transition of hoof landing and lifting off the ground. Therefore, the metatarsophalangeal joint of the hind limb was selected as the key location (i.e., the marker point) for extracting the characterization signal.

[0036] Compared to the trunk or proximal joints, the metatarsophalangeal joints of the hind limbs are closer to the ground, and their movement trajectory is more easily affected by the impact of landing, the transition of support, and the swinging process, which can produce more obvious temporal change characteristics, which is beneficial for the subsequent identification of gait cycle boundaries.

[0037] Furthermore, the three-dimensional coordinate information of the corresponding marker points of the metatarsophalangeal joints of the hind limbs is extracted from the three-dimensional skeletal model to obtain the spatial displacement changes of this key moving part during continuous movement.

[0038] By continuously collecting the coordinates of marker points at multiple times, the three-dimensional motion trajectory of the metatarsophalangeal joint of the hind limb during the complete movement process can be obtained.

[0039] Since the vertical movement changes during the gait of Mongolian horses are usually strongly correlated with hoof lift-off, hoof landing, and the transition between the support and sway phases, the focus is on analyzing the displacement changes in the Z-axis direction, using the Z-axis displacement signal as the basis for subsequent kinematic feature calculations.

[0040] Subsequently, the three-dimensional coordinate information was differentiated to obtain the velocity and acceleration signals of the metatarsophalangeal joints of the hind limbs in the Z-axis direction.

[0041] The velocity signal is used to describe the rate of change of position of the metatarsophalangeal joint in the vertical direction of the hind limb.

[0042] Further derivative processing of the velocity signal was performed, and the Z-axis acceleration signal was used to reflect the degree of change in the motion state of the metatarsophalangeal joint of the hind limb.

[0043] Because acceleration signals are more sensitive to changes in motion, they usually produce obvious peak or trough changes when key motion events occur, such as when the hoof contacts the ground, leaves the ground, or switches between support states. Therefore, compared with displacement and velocity signals, acceleration signals can more clearly reflect the phase transition characteristics within the gait cycle.

[0044] This invention, through comparative analysis of motion signals in different directions, found that compared with X-axis and Y-axis signals, Z-axis acceleration signals have more obvious periodic repetition and local extremum characteristics under different gait conditions such as slow walking, fast walking, and jogging. Therefore, Z-axis acceleration signals were selected as candidate characterization signals for gait period segmentation.

[0045] However, due to the unavoidable measurement errors, soft tissue vibrations, and environmental interference during the acquisition process of motion capture equipment, and the fact that the displacement signal is converted into an acceleration signal by second-order derivative, the high-frequency noise and local abnormal changes in the original signal will be further amplified, which may cause the unprocessed acceleration signal to have non-real motion fluctuations, thus affecting the accuracy of subsequent template matching and periodic boundary positioning.

[0046] Therefore, the obtained Z-axis acceleration signal can be filtered and standardized to improve the stability of the characterization signal.

[0047] Specifically, the continuously acquired Z-axis acceleration signal is discretized.

[0048] It should be noted that the sampling points are not new spatial markers, but rather data points obtained by time discretizing the motion signals of the same hind limb metatarsophalangeal joint markers according to the sampling frequency. Each sampling point corresponds to the acceleration amplitude at a specific time position.

[0049] To reduce high-frequency noise interference while preserving the low-frequency periodic variation characteristics of the Mongolian horse's gait, a Butterworth low-pass filter was used to smooth the original acceleration signal.

[0050] Butterworth filters are characterized by a flat amplitude-frequency response within the passband and do not introduce significant ripple, making them suitable for processing biological motion signals with periodicity and low-frequency dominance.

[0051] By adjusting the cutoff frequency, the degree to which the filter suppresses high-frequency noise components can be controlled, making the processed signal more prominent in terms of the main movement rhythms within the gait cycle of the Mongolian horse.

[0052] Furthermore, based on the sampling frequency of the motion capture data and the preset cutoff frequency, the normalized cutoff frequency corresponding to the Butterworth filter is calculated: The normalized cutoff frequency is used to convert the actual cutoff frequency into a dimensionless parameter suitable for digital filter design.

[0053] Subsequently, the numerator coefficient B and denominator coefficient A of the filter were obtained according to the filter design parameters, and the original acceleration signal was processed using a bidirectional zero-phase filtering method. This invention cancels phase delay by filtering twice, in both the forward and reverse directions, which can reduce noise while maintaining the true position of gait events on the time axis and avoid period boundary shifts caused by filtering.

[0054] Finally, to further reduce the differences in signal amplitude caused by different Mongolian horse individuals, different movement speeds, and different acquisition conditions, the filtered acceleration signal was standardized.

[0055] Specifically, the Z-score normalization method is used to calculate the mean and standard deviation of the filtered signal.

[0056] Standardization can eliminate the amplitude scale effects caused by differences in body size, movement amplitude, and acquisition distance between different individuals, allowing different time-phase cycles to be matched mainly based on time variation patterns and waveform structure.

[0057] By extracting hind limb metatarsophalangeal joint motion information based on a 3D skeletal model, and performing filtering, zero-phase processing, and standardization on the Z-axis acceleration signal, a characterization signal that stably reflects the gait rhythm changes and support-swing phase transition characteristics of Mongolian horses was obtained. This characterization signal retains the local extremum features corresponding to key motion events within a single gait cycle while reducing the impact of noise and individual differences on the signal matching process, providing a reliable data foundation for subsequent template matching, gait cycle localization, and cycle boundary segmentation based on dynamic time warping algorithms.

[0058] Step S13: Use the standard gait cycle set as the template sequence and the continuous characterization signal as the continuous test sequence. Align the template sequence with the continuous test sequence through local distance calculation, cumulative cost matrix construction and minimum regularization path search. The standard gait cycle set contains multiple complete standard gait cycles.

[0059] Using a standard gait cycle set as a template sequence and continuous representation signals as continuous test sequences, the template sequence is aligned with the continuous test sequence through local distance calculation, cumulative cost matrix construction, and minimum warp path search, including: Define continuous test sequences and template sequences, including: , , in, This is a vector representation of a continuous test sequence, which is a truncated and standardized sequence. The acceleration signal at each sampling point may contain part or all of the acceleration signal. , The vector representation of the template sequence. For the first test in a continuous test sequence The representation signal corresponding to each sampling point For the template sequence of the first The representation signal corresponding to each sampling point; Calculating the local distance between the template sequence and the test sequence includes: , , , in, It is a set of temporal and representative signals of the template sequence. The time series and representation signal set of continuous test sequences. For the time of the template sequence, For the time of continuous test sequences, Local distance; Constructing a cumulative cost matrix based on the local distance includes: , in, For the cumulative cost matrix, For local distance, This is the cumulative cost matrix function; The optimal matching path between the template sequence and the test sequence is determined through minimum regularity path search, wherein boundary constraints, continuity constraints, and monotonicity constraints are satisfied during the search, including: , in, To standardize the path, For the first One point element; , in, This is the starting element.

[0060] Boundary constraints include: , Continuity constraints include: , in, For the first in the normalized path The coordinates of each element; Monotonicity constraints include: .

[0061] Figure 4 A flowchart illustrating the improved gait cycle segmentation method provided by this invention.

[0062] Figure 5 This is a schematic diagram of template samples under different motion states.

[0063] To achieve accurate identification of individual gait cycles in continuous Mongolian horse gait signals, it is necessary to establish a representative standard gait cycle template and perform time-series matching between it and the continuous characterization signal to be segmented.

[0064] Because Mongolian horses exhibit significant differences in movement speed under different movement states (such as walking, walking fast, and trotting), and because different individuals under the same movement state also show variations in stride length, movement rhythm, and local movement amplitude, a single fixed-length template is difficult to adapt to the changes in actual continuous gait signals.

[0065] The first step in this process is to obtain standardized characterization signals, construct a set of standard gait cycles from multiple complete gait cycles that have been manually labeled and filtered, and use the set of standard gait cycles as a template sequence.

[0066] The standard gait cycle set contains multiple representative complete gait cycles. Each standard gait cycle corresponds to a complete limb movement cycle and can reflect the typical time structure and waveform change pattern of the target Mongolian horse under specific movement conditions.

[0067] Specifically, the gait characterization signals of Mongolian horses obtained through continuous acquisition are used as continuous test sequences. These continuous test sequences may contain one or more complete gait cycles, and their length is usually greater than that of a single standard gait cycle.

[0068] The continuous test sequence is derived from the Z-axis acceleration signal of the hind limb metatarsophalangeal joint after filtering and standardization in step S12. That is, part or all of the standardized acceleration signal is extracted from the continuous motion process as the object to be segmented.

[0069] Since the standard gait cycle set contains multiple complete cycles, multiple template sequences can be established according to different gait types during the actual matching process, and matched with continuous test sequences respectively, so as to improve the algorithm's adaptability to different speed conditions and individual differences.

[0070] Furthermore, in order to achieve time-series matching between the template sequence and the continuous test sequence, it is necessary to first calculate the local distance between the two sequences to describe the signal differences at different time points.

[0071] Since traditional dynamic time warping algorithms typically calculate distance based solely on differences in signal amplitude, while Mongolian horse gait signals exhibit not only amplitude variations but also period length variations and local time shifts, this step comprehensively considers both the time dimension and the signal amplitude dimension to construct a two-dimensional local distance.

[0072] Specifically, the template sequence and the test sequence are represented as joint time-signal sets (the time and signal can be normalized first): Further calculate the local distance between the b-th sampling point of the template sequence and the m-th sampling point of the test sequence.

[0073] By considering both time differences and signal amplitude differences simultaneously, mismatches caused by similar local waveforms but different time positions can be avoided when matching is based solely on amplitude, thus improving the accuracy of the correspondence between the template and the continuous test sequence.

[0074] After obtaining the local distance between the template sequence and the test sequence, a cumulative cost matrix is ​​further constructed based on the idea of ​​dynamic programming to describe the minimum cumulative matching cost from the starting point of the template sequence to the current position of the test sequence.

[0075] The calculation process compares the cumulative costs of the three possible matching directions ahead of the current point and selects the minimum value as the current optimal cumulative path, so that the template sequence and the test sequence can complete the overall structural matching within the allowable time scaling.

[0076] Through the above-described process of constructing the cumulative cost matrix, the original continuous gait signal matching problem can be transformed into a path search problem in a two-dimensional matrix space.

[0077] Within this matrix space, there are multiple possible matching paths from the starting point to the ending point. The path with the minimum cumulative cost represents the optimal time correspondence between the template sequence and the continuous test sequence.

[0078] Therefore, by tracing back to the minimum position in the cumulative cost matrix, the minimum normalized path between the template sequence and the continuous test sequence can be found: The minimum regularization path records the one-to-one correspondence between the sample points of the template sequence and the sample points of the continuous test sequence. This path not only reflects the overall similarity between the two sequences, but also provides a time mapping basis for the subsequent determination of gait cycle candidate intervals and boundary localization.

[0079] To ensure that the minimum regular path conforms to the temporal continuity and motion patterns of the actual Mongolian horse movement, boundary constraints, continuity constraints, and monotonicity constraints are set during the path search process.

[0080] Boundary constraints are used to ensure that the starting and ending points of the template sequence and the test sequence can correspond effectively; that is, the normalized path must start from the beginning position of the matrix and eventually reach the ending position of the matrix. Boundary constraints can prevent the overall cycle from shifting due to the path only matching a local area.

[0081] Continuity constraints are used to limit the movement between adjacent matching points during path search, so that the path can only advance step by step along adjacent positions. This constraint ensures that a sampling point in the template sequence will not jump to a distant position in the test sequence, thus avoiding abnormal matching that does not conform to the actual movement process.

[0082] Monotonicity constraints are used to ensure that the time series matching direction always moves forward. This constraint ensures that there is no time backtracking in the matching path, so that the template sequence and the continuous test sequence maintain a time correspondence that conforms to the laws of biological motion.

[0083] By using a set of standard gait cycles as a template sequence and continuous characterization signals as test sequences, and by quantifying the local differences between the two using local distance calculation, global temporal optimization is achieved through a cumulative cost matrix. Finally, the dynamic alignment of the template sequence and the continuous test sequence is completed through a minimum regularization path that satisfies boundary constraints, continuity constraints, and monotonicity constraints.

[0084] This process can effectively adapt to the changes in cycle length, speed differences, and local waveform changes that exist in the gait of Mongolian horses, providing an accurate temporal matching basis for subsequent gait cycle localization and boundary recognition based on the improved dynamic time warping algorithm.

[0085] Step S14: Introduce a multi-template matching strategy and an improved constraint window strategy to construct an improved dynamic time warping algorithm. The improved dynamic time warping algorithm is used to optimize the alignment process between the template sequence and the continuous test sequence.

[0086] Specifically, this includes: calculating the matching cost between the template sequence and the continuous test sequence, and determining the optimal template based on the minimum matching cost. The optimal template is used for gait cycle localization, boundary correction, and determining the optimal matching interval. The constraint window range is adjusted based on the length difference between the template sequence and the continuous test sequence, where the window range is used to limit the regular path search area to optimize the alignment process.

[0087] After establishing the standard gait cycle template and initial matching of continuous test sequences with the template sequence, the gait cycle length, local motion amplitude, and temporal variation patterns of Mongolian horses vary under different motion states, individuals, and speeds. Using a single fixed template for dynamic time warping matching can easily lead to excessive stretching or compression of local areas, causing the matching path to deviate from the actual gait change process, thereby reducing the accuracy of gait cycle boundary recognition.

[0088] Therefore, this invention introduces a multi-template matching strategy based on the traditional DTW algorithm. By constructing multiple representative standard gait cycle templates, the template set can be improved to cover the actual range of gait changes in Mongolian horses, thereby enhancing the robustness of continuous gait signal matching.

[0089] Specifically, for different movement states of Mongolian horses, such as walking, trotting, and trotting, corresponding sets of standard gait cycle templates were established. Each template is derived from a manually labeled complete standard gait cycle and has undergone time length normalization processing, so that gait signals with different cycle lengths can be compared on a uniform scale.

[0090] Let the continuous test sequence be: , For the same gait state, construct a template set containing multiple standard cycles. ; , in, Indicates the number of templates. Indicates the first A standard gait cycle template.

[0091] Each template retains the gait temporal characteristics of different individuals under different periodic change conditions, so that the template set can cover the periodic fluctuations that exist in the real movement process.

[0092] During the matching process, each template in the template set is subjected to DTW calculation with the continuous test sequence to obtain the cumulative matching cost corresponding to each template.

[0093] The matching cost describes the similarity between the template sequence and the test sequence in terms of temporal variation and signal amplitude variation. The smaller the value, the closer the overall shape of the two sequences are. Therefore, by comparing the matching costs of all candidate templates, the template with the smallest matching cost is selected as the optimal template corresponding to the current continuous test sequence.

[0094] The optimal template can be represented as: , in, As the optimal template, This represents the minimum matching cost.

[0095] The optimal template not only represents the main motion characteristics of the gait state corresponding to the current test sequence, but its corresponding minimum regular path can also reflect the best time mapping relationship between the template sequence and the continuous test sequence.

[0096] Based on this optimal template, candidate gait cycle matching regions in continuous test signals can be further determined, providing a reliable basis for subsequent gait cycle start and end boundary localization and local boundary correction. Compared with the single-template matching method, the multi-template matching strategy can reduce matching deviations caused by individual template differences or local waveform changes, and improve the algorithm's adaptability to gait cycle changes in different Mongolian horse individuals and under different movement speed conditions.

[0097] Furthermore, to address the issues of reduced computational efficiency due to the excessively large search range of regular paths in the traditional DTW algorithm, and the difficulty of adapting fixed constraint windows to matching unequal gait sequences, this step introduces an improved constraint window strategy to dynamically adjust the path search region in the DTW algorithm.

[0098] This step adaptively adjusts the constraint window range based on the difference between the template sequence length and the continuous test sequence length.

[0099] Let the template sequence length be... The length of the continuous test sequence is By introducing window adjustment parameters, the allowed search area is dynamically determined based on the length relationship between the two, ensuring that regularized paths are searched only within a reasonable range, including: , in, For the window radius, This is the window adjustment coefficient, and The constraint condition is: greater than 0 and less than 1. , in, These are the coordinates of the path points.

[0100] This invention reduces invalid path calculations and improves the efficiency of the DTW algorithm by limiting the path search range; it also expands or shrinks the window range to adapt to the time scaling relationship between gait sequences of different lengths, avoiding incorrect matching due to changes in cycle length.

[0101] With the improved constraint window, the DTW algorithm is no longer limited to a linear search region of fixed width. Instead, it forms an adaptive matching region based on the actual gait sequence length variation, while satisfying boundary constraints, continuity constraints, and monotonicity constraints. This approach can improve the algorithm's adaptability to local time shifts, period length variations, and signal changes under different motion speeds while maintaining the overall sequential relationship of the time series.

[0102] Figure 6 This is a comparative diagram showing the difference before and after the introduction of the improved constraint window strategy provided by the present invention.

[0103] Step S15: Based on the optimal matching interval determined by the improved dynamic time warping algorithm and combined with local extreme value features, determine the start and end positions of a single gait cycle. The start and end positions of the gait cycle are used for gait cycle segmentation of the target Mongolian horse.

[0104] Specifically, this includes: correcting the boundary position of the optimal matching interval based on the optimal matching interval and local extreme value characteristics; and determining the start and end positions of a single gait cycle based on the corrected boundary position.

[0105] After aligning the template sequence with the continuous test sequence by improving the dynamic time warping algorithm, it is necessary to further convert the matching result into the specific periodic boundary position in the continuous gait signal.

[0106] Since the warping path obtained by the DTW algorithm essentially reflects the time mapping relationship between the template sequence and the test sequence, it can determine the corresponding region of the template period in the continuous test signal. However, since the time axis is allowed to stretch or shrink locally during the dynamic warping process, there may be some deviation in determining the period boundary by simply relying on the endpoints of the warping path.

[0107] Therefore, this step, based on the optimal matching interval output by the improved dynamic time warping algorithm, further combines the local motion event features in the characterization signal to correct the candidate boundary, thereby achieving accurate segmentation of a single gait cycle.

[0108] Specifically, the minimum warping path corresponding to the optimal template determined by the improved dynamic time warping algorithm is mapped to the continuous test sequence, and the current candidate gait cycle region is determined based on the coverage of the warping path on the test sequence.

[0109] The starting point of the regularized path is used as the starting position of the candidate gait cycle, and the ending point of the regularized path is used as the ending position of the candidate gait cycle.

[0110] This candidate region reflects the overall temporal correspondence between the template gait period and the continuous test signal, which can effectively avoid the boundary offset problem caused by the change in period length in the traditional fixed time window segmentation method.

[0111] Because Mongolian horses are affected by individual differences, changes in movement speed, and movement stability during actual movement, the signal changes within the gait cycle are not strictly repeated. The boundary position determined by the regular path may fall near the transition area of ​​movement events, and cannot completely correspond to the start and end times of the cycle in the actual biomechanical sense.

[0112] Therefore, this step further uses the Z-axis acceleration characterization signal of the hind limb metatarsophalangeal joint as the basis for boundary correction, and uses the local extreme value changes generated by this signal during hoof landing, load conversion and takeoff to perform secondary correction on the candidate period boundary.

[0113] Specifically, after obtaining the candidate gait cycle interval, local feature detection is first performed on the Z-axis acceleration signal in the interval and its vicinity to extract peak, valley or zero-crossing feature points related to gait phase transition.

[0114] Because the metatarsophalangeal joints of the hind limbs of Mongolian horses exhibit significant acceleration changes during landing and takeoff, their local extreme values ​​typically correspond to important transition events between the support phase and the swaying phase, and therefore can serve as an important reference for adjusting the gait cycle boundary.

[0115] Furthermore, wavelet transform zero-crossing detection can be used to perform multi-scale analysis on acceleration signals. By detecting abrupt changes in signal trends at different scales, more stable local extrema can be extracted to reduce the impact of noise and local fluctuations on boundary judgment.

[0116] During the boundary correction process, the candidate starting point and candidate ending point provided by the optimal matching interval are used as the initial range, and the local extreme points related to gait events in their neighboring regions are searched respectively.

[0117] If an extreme position that conforms to kinematic characteristics is detected, that position is used as the corrected periodic boundary; if there is no obvious extreme value change, the original boundary position corresponding to the optimal matching interval is retained.

[0118] By utilizing the global temporal matching capability provided by the improved DTW algorithm to constrain the period range, and by using local extreme value features to improve the boundary positioning accuracy, the final determined start and end positions are made more consistent with the biomechanical changes in the actual movement process of Mongolian horses.

[0119] After determining the boundary of a single gait cycle, the above method is used to process multiple cycles in a continuous gait signal.

[0120] First, the current optimal matching interval is determined based on the improved dynamic time warping algorithm. Then, boundary correction is performed within this interval by combining the local extremum features of the Z-axis acceleration signal. Finally, the corresponding complete gait cycle is extracted based on the corrected start and end points, and the same process is continued for the remaining continuous signals until the cycle division of the entire continuous gait sequence is completed.

[0121] Figure 7 This is a schematic diagram illustrating the determination of the start and end points of the gait cycle and the result of the cycle based on the improved DTW provided by the present invention.

[0122] In summary, this invention establishes template samples under different gait conditions, combines local distance calculation, cumulative cost matrix construction, and minimum regularization path search to achieve optimal alignment between template sequences and continuous test sequences. At the same time, it introduces a multi-template matching strategy and an improved constraint window to enhance the algorithm's adaptability to gait sequences of different lengths and local rhythm changes, and combines local extreme value features of the signal to complete periodic boundary correction.

[0123] This invention can accurately identify the periodic boundaries of continuous gait signals of Mongolian horses, providing reliable data for subsequent gait feature extraction, gait recognition, and mining of gait features related to movement performance.

[0124] Based on the same inventive concept, the present invention provides a Mongolian horse gait period segmentation device, comprising: The model building module is used to build a three-dimensional skeletal model of the target Mongolian horse; The characterization signal extraction module is used to extract and select characterization signals based on the three-dimensional skeletal model of the target Mongolian horse. The characterization signals are used to reflect the gait rhythm change characteristics and phase transition characteristics of the target Mongolian horse. The template alignment module is used to take the standard gait cycle set as the template sequence and the continuous characterization signal as the continuous test sequence, and align the template sequence with the continuous test sequence through local distance calculation, cumulative cost matrix construction and minimum warping path search. The standard gait cycle set contains multiple complete standard gait cycles. The alignment improvement module is used to introduce a multi-template matching strategy and an improved constraint window strategy to build an improved dynamic time warping algorithm. The improved dynamic time warping algorithm is used to optimize the alignment process between the template sequence and the continuous test sequence. The period segmentation module is used to determine the start and end positions of a single gait cycle based on the optimal matching interval determined by the improved dynamic time warping algorithm and combined with local extreme value features. The start and end positions of the gait cycle are used for gait cycle segmentation of the target Mongolian horse.

[0125] Based on the same inventive concept, this application also provides an electronic device, including: processor; Memory used to store processor-executable instructions; The processor is configured to execute a Mongolian horse gait period segmentation method as described above.

[0126] Based on the same inventive concept, this application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to implement the Mongolian horse gait period segmentation method provided above.

[0127] Since the electronic device described in this embodiment is an electronic device used to implement the information processing method in the embodiments of the present invention, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the information processing method described in the embodiments of the present invention. Therefore, how the electronic device implements the method in the embodiments of the present invention will not be described in detail here. Any electronic device used by those skilled in the art to implement the information processing method in the embodiments of the present invention falls within the scope of protection of the present invention.

[0128] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. A method for segmenting the gait period of Mongolian horses, characterized in that, include: Construct a three-dimensional skeletal model of the target Mongolian horse; Based on the three-dimensional skeletal model of the target Mongolian horse, characterization signals are extracted and selected, wherein the characterization signals are used to reflect the gait rhythm change characteristics and phase transition characteristics of the target Mongolian horse. The standard gait cycle set is used as the template sequence, and the continuous characterization signal is used as the continuous test sequence. The template sequence is aligned with the continuous test sequence through local distance calculation, cumulative cost matrix construction and minimum regularization path search. The standard gait cycle set contains multiple complete standard gait cycles. A multi-template matching strategy and an improved constraint window strategy are introduced to construct an improved dynamic time warping algorithm, which is used to optimize the alignment process between the template sequence and the continuous test sequence. The optimal matching interval determined by the improved dynamic time warping algorithm is used to determine the start and end positions of a single gait cycle, and the start and end positions of the gait cycle are used for gait cycle segmentation of the target Mongolian horse.

2. The method for segmenting the gait period of Mongolian horses as described in claim 1, characterized in that, Construct a 3D skeletal model of the target Mongolian horse, including: Obtain the skeletal data and body size parameters of the target Mongolian horse, and construct an initial three-dimensional skeletal model; Based on the spatial location and connection relationship of each bone in the skeletal data, the initial three-dimensional skeletal model is divided into bone segments to obtain multiple independent bone segments; The skeletal parameters of each independent bone segment are adjusted to match the initial three-dimensional skeletal model with the actual skeletal structure of the target Mongolian horse, thus obtaining a three-dimensional skeletal model. The skeletal parameters include length, angle, joint center position, mass, and center of gravity.

3. The method for segmenting the gait period of Mongolian horses as described in claim 1, characterized in that, Based on the three-dimensional skeletal model of the target Mongolian horse, representational signals are extracted and selected, including: Extracting the three-dimensional coordinate information corresponding to the marker points of the metatarsophalangeal joints of the hind limb in the three-dimensional skeletal model, including: ， in, Mark the point at Three-dimensional coordinate information at time, The marked points are respectively Always axis, shaft and Displacement of the axis; Differentiating the three-dimensional coordinate information yields... The velocity and acceleration signals in the axial direction include: ， ， in, For three-dimensional coordinate information in Moment Velocity signal in the axial direction For three-dimensional coordinate information in Moment Acceleration signal in the axial direction; The acceleration signal is low-pass filtered and normalized to obtain the characterization signal, including: ， in, For the first Each sampling point is Moment Acceleration signal in the axial direction, This represents the number of sampling points; ， in, Angular frequency, The cutoff angular frequency, Let the filter order be . It is the frequency response function. The imaginary unit; ， in, The cutoff frequency, The sampling frequency of the marker points, For the first Normalized cutoff frequency for each sampling point; ， in, For filter molecules, These are the denominator coefficients of the filter. For the first The acceleration signal after low-pass filtering of each sampling point It is a bidirectional zero-phase filter function; ， in, For the standardized first Acceleration signal at each sampling point The mean of the signal. Let be the standard deviation of the signal, where the standardized acceleration signal is used as the characterization signal.

4. The method for segmenting the gait period of Mongolian horses as described in claim 1, characterized in that, Using a standard gait cycle set as a template sequence and continuous representation signals as continuous test sequences, the template sequence is aligned with the continuous test sequences through local distance calculation, cumulative cost matrix construction, and minimum warp path search, including: Define continuous test sequences and template sequences, including: ， ， in, This is a vector representation of a continuous test sequence, which is a truncated and standardized sequence. The acceleration signal at each sampling point may contain part or all of the acceleration signal. , The vector representation of the template sequence. For the first test in a continuous test sequence The representation signal corresponding to each sampling point For the template sequence of the first The representation signal corresponding to each sampling point; Calculating the local distance between the template sequence and the test sequence includes: ， ， ， in, It is a set of temporal and representative signals of the template sequence. The time series and representation signal set of continuous test sequences. For the time of the template sequence, For the time of continuous test sequences, Local distance; Constructing a cumulative cost matrix based on the local distance includes: ， in, For the cumulative cost matrix, For local distance, This is the cumulative cost matrix function; The optimal matching path between the template sequence and the test sequence is determined through minimum regularity path search, wherein boundary constraints, continuity constraints, and monotonicity constraints are satisfied during the search, including: ， in, To standardize the path, For the first One point element; ， in, This is the starting element.

5. The method for segmenting the gait period of a Mongolian horse as described in claim 4, characterized in that, include: Boundary constraints include: ， Continuity constraints include: ， in, For the first in the normalized path The coordinates of each element; Monotonicity constraints include: 。 6. The method for segmenting the gait period of a Mongolian horse as described in claim 1, characterized in that, An improved dynamic time warping algorithm is constructed by introducing a multi-template matching strategy and an improved constraint window strategy. This improved dynamic time warping algorithm optimizes the alignment process between the template sequence and the continuous test sequence, including: The matching cost between the template sequence and the continuous test sequence is calculated, and the optimal template is determined based on the minimum matching cost. The optimal template is used for gait cycle localization, boundary correction, and determination of the optimal matching interval. The constraint window range is adjusted based on the length difference between the template sequence and the continuous test sequence, where the window range is used to limit the search area for regularized paths to optimize the alignment process.

7. The method for segmenting the gait period of Mongolian horses as described in claim 1, characterized in that, Based on the optimal matching interval determined by the improved dynamic time warping algorithm, and combined with local extremum features, the start and end positions of a single gait cycle are determined, including: The boundary position of the optimal matching interval is corrected based on the optimal matching interval and local extreme value characteristics; The start and end positions of a single gait cycle are determined based on the corrected boundary positions.

8. A Mongolian horse gait period segmentation device, characterized in that, include: The model building module is used to build a three-dimensional skeletal model of the target Mongolian horse; The characterization signal extraction module is used to extract and select characterization signals based on the three-dimensional skeletal model of the target Mongolian horse, wherein the characterization signals are used to reflect the gait rhythm change characteristics and phase transition characteristics of the target Mongolian horse. The template alignment module is used to take the standard gait cycle set as the template sequence and the continuous characterization signal as the continuous test sequence, and align the template sequence with the continuous test sequence through local distance calculation, cumulative cost matrix construction and minimum warping path search. The standard gait cycle set contains multiple complete standard gait cycles. An alignment improvement module is used to introduce a multi-template matching strategy and an improved constraint window strategy to construct an improved dynamic time warping algorithm, which is used to optimize the alignment process between the template sequence and the continuous test sequence. The cycle segmentation module is used to determine the start and end positions of a single gait cycle based on the optimal matching interval determined by the improved dynamic time warping algorithm and combined with local extreme value features. The start and end positions of the gait cycle are used for gait cycle segmentation of the target Mongolian horse.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute a Mongolian horse gait period segmentation method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform a Mongolian horse gait period segmentation method as described in any one of claims 1 to 7.

Images