Patient gait analysis method and system based on motion detection data

Abstract

The present invention provides a method and system for analyzing patient gait based on motion detection data, and relates to the technical field of gait analysis. The method comprises: obtaining motion detection data and electromyographic signal data to be analyzed of the patient, extracting motion state cycle data corresponding to multiple target gait cycles, performing gait coordination sliding window detection on multiple groups of motion state cycle data, generating coordination scores corresponding to multiple reference motion state sequences, and determining multiple gait coordination abnormality cycles; extracting abnormal motion state data and reference motion state data of each gait coordination abnormality cycle, generating motion disassembly data corresponding to multiple local motion stages; constructing a local motion state sequence for each local motion stage, performing gait local coordination detection on multiple gait coordination abnormality cycles, and obtaining the target gait abnormality detection result of the patient. The present invention realizes high-precision detection of minor abnormalities in the patient's gait.

Description

Technical Field

[0001] The present invention relates to the technical field of gait analysis, and particularly to a method and system for analyzing a patient's gait based on motion detection data. Background Art

[0002] During the rehabilitation process of stroke patients, gait analysis is one of the important means to evaluate and guide rehabilitation treatment. In the later stage of the rehabilitation phase, the gait of the hemiplegic side and the healthy side of the patient gradually tends to be consistent, and the overall gait balance may appear relatively normal. However, although the gait may seemingly have recovered on the surface, there may still be minor differences or abnormalities between the hemiplegic side and the healthy side. These minor abnormalities are generally difficult to observe when analyzing the overall gait balance, but these minor differences may have an important impact on the patient's gait stability and long-term rehabilitation effect. Identifying the details that are not obvious but potentially hide rehabilitation obstacles or future gait function problems, and early detecting these minor gait abnormalities and intervening can provide more personalized rehabilitation feedback and training programs for patients, thereby improving the gait rehabilitation effect of patients. Summary of the Invention

[0003] In view of this, the present invention proposes a method and system for analyzing a patient's gait based on motion detection data. By collecting and analyzing the relevant detection data of the patient's motion state and electromyogram state, high-precision mining of possible minor abnormalities in the patient's gait is carried out, so as to assist rehabilitation physicians in adjusting treatment strategies in a timely manner according to minor gait abnormalities, and achieving the improvement of the patient's gait rehabilitation effect.

[0004] The first aspect of the present invention provides a method for analyzing a patient's gait based on motion detection data, including:

[0005] Obtaining the motion detection data and electromyogram signal data of the patient to be analyzed, and extracting the motion state cycle data corresponding to multiple target gait cycles from the motion detection data and electromyogram signal data to be analyzed. Each group of motion state cycle data includes the healthy side state data and the affected side state data of the patient;

[0006] Extracting the reference motion state sequence and the target motion state sequence of each group of motion state cycle data, and constructing a reference state matrix including multiple reference motion state sequences and a target state matrix including multiple target motion state sequences;

[0007] Performing gait coordination sliding window detection on multiple groups of motion state cycle data, including generating a global motion feature sequence within each sliding window based on the target state matrix, performing gait coordination detection on multiple reference motion state sequences within the sliding window according to the global motion feature sequence, generating coordination scores corresponding to multiple reference motion state sequences respectively, and determining multiple gait coordination abnormal cycles according to the coordination scores;

[0008] Extract the abnormal motion state data and reference motion state data of each gait coordination abnormal cycle from multiple groups of motion state cycle data, perform motion stage segmentation on the multiple groups of abnormal motion state data and reference motion state data, and obtain the motion decomposition data corresponding to the abnormal motion state data and reference motion state data in multiple local motion stages;

[0009] Extract multiple motion state parameters of each group of motion decomposition data, construct the local motion state sequence of each local motion stage, and perform gait local coordination detection on multiple gait coordination abnormal cycles based on the multiple local motion state sequences to obtain the target gait abnormal detection result of the patient.

[0010] Preferably, perform gait coordination sliding window detection on multiple groups of motion state cycle data, including:

[0011] Perform a time-step correlation traversal operation on the reference state matrix and the target state matrix through a sliding window to generate coordination scores corresponding to multiple reference motion state sequences;

[0012] Among them, for multiple reference motion state sequences and target motion state sequences within any sliding window, generate a global motion feature sequence within the sliding window according to the multiple reference motion state sequences, and calculate the similarity distances between the multiple reference motion state sequences and the global motion feature sequence respectively;

[0013] Calculate the standard deviation of the multiple similarity distances to obtain the gait local correlation parameter, perform a stationary analysis on the multiple target motion state sequences to obtain the fluctuation parameter, process the gait local correlation parameter according to the fluctuation parameter to generate an adaptive correction parameter, and correct the similarity distance between the reference motion state sequence at the center of the sliding window and the global motion feature sequence according to the adaptive correction parameter to obtain the coordination score of the reference motion state sequence at the center of the sliding window;

[0014] Screen out multiple reference motion state sequences through the coordination score threshold, and record the target gait cycles corresponding to the multiple screened reference motion state sequences as gait coordination abnormal cycles.

[0015] Preferably, perform gait local coordination detection on multiple gait coordination abnormal cycles based on multiple local motion state sequences, including:

[0016] For the reference motion state data of the gait coordination abnormal cycle, including the healthy side state data corresponding to multiple target gait cycles, construct a local state matrix corresponding to each group of healthy side state data according to the multiple local motion state sequences;

[0017] Fuse multiple local state matrices to generate a phase change state matrix for the gait coordination abnormal cycle, and determine the corrected state matrix of each group of healthy side state data based on the phase change state matrix;

[0018] For the abnormal motion state data of the gait coordination abnormal cycle, including the affected side state data corresponding to multiple target gait cycles, construct a phase coordination state matrix corresponding to each group of affected side state data according to multiple local motion state sequences;

[0019] Based on the correlation relationship between each group of abnormal motion state data and the reference motion state data, correct multiple phase coordination state matrices respectively based on multiple corrected state matrices to obtain multiple phase coordination state correction matrices;

[0020] Perform gait local coordination detection on the gait coordination abnormal cycle according to multiple phase coordination state correction matrices, including performing local phase coordination analysis on multiple phase coordination state correction matrices to generate the coordination characteristic parameters of each local motion phase, and based on the coordination characteristic parameters and the phase change state matrix of the gait coordination abnormal cycle, perform gait local coordination detection on multiple phase coordination state correction matrices of the gait coordination abnormal cycle to obtain the local gait abnormal results of each gait coordination abnormal cycle.

[0021] Preferably, performing gait local coordination detection on the gait coordination abnormal cycle according to multiple phase coordination state correction matrices further includes:

[0022] Perform local phase coordination analysis on multiple phase coordination state correction matrices, extract multiple local phase state sequences of each local motion phase from multiple phase coordination state correction matrices, perform autocorrelation analysis on multiple local phase state sequences of each local motion phase, and calculate the coordination characteristic parameters of each local motion phase regarding multiple local phase state sequences;

[0023] Perform similarity analysis based on local motion phases on multiple phase coordination state correction matrices of the gait coordination abnormal cycle according to the phase change state matrix, including calculating the similarity parameters regarding each local motion phase in the phase coordination state correction matrix, fusing the similarity parameters of multiple local motion phases according to multiple coordination characteristic parameters to obtain the local phase similarity parameters of the phase coordination state correction matrix, and generating the similarity analysis results regarding multiple phase coordination state correction matrices of the gait coordination abnormal cycle;

[0024] According to the similarity analysis results of the collaborative state correction matrix for multiple stages, a gait local coordination feature detection vector for the gait coordination abnormal period is generated. According to the multiple local state matrices of the gait coordination abnormal period, a gait local coordination feature fluctuation vector for the gait coordination abnormal period is generated. Based on the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector, a local gait abnormal result for the gait coordination abnormal period is generated.

[0025] Preferably, generating a local gait abnormal result for the gait coordination abnormal period based on the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector includes:

[0026] Calculating the distance parameter between the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector. If the distance parameter between the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector is less than a preset distance threshold, it is recorded that there is no local gait abnormality in the gait coordination abnormal period;

[0027] Otherwise, it is recorded that there is a local gait abnormality in the gait coordination abnormal period, and the abnormal local motion stage is determined according to the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector, including determining the local stage collaboration threshold of the gait coordination abnormal period according to the gait local coordination feature fluctuation vector, determining the collaboration difference value of each local motion stage of the gait coordination abnormal period according to the local stage similarity parameter of the stage collaboration state correction matrix, judging whether the collaboration difference value of the local motion stage is greater than the local stage collaboration threshold of the gait coordination abnormal period, recording the local motion stage with the collaboration difference value greater than the local stage collaboration threshold of the gait coordination abnormal period as the abnormal local motion stage, generating a local gait abnormal result for the gait coordination abnormal period, and generating a target gait abnormal detection result of the patient according to the local gait abnormal results of multiple gait coordination abnormal periods.

[0028] Preferably, for the gait local coordination feature fluctuation vector, it further includes:

[0029] By calculating the similarity parameters between the multiple local state matrices of the gait coordination abnormal period and the stage change state matrix of the gait coordination abnormal period respectively, the gait local coordination feature fluctuation vector is constructed according to the multiple similarity parameters.

[0030] The second aspect of the present invention provides a patient gait analysis system based on motion detection data for implementing the above-mentioned patient gait analysis method based on motion detection data, including:

[0031] A data acquisition module, configured to obtain motion detection data and electromyogram signal data of a patient to be analyzed, and extract motion state cycle data corresponding to multiple target gait cycles from the motion detection data and electromyogram signal data to be analyzed. Each set of motion state cycle data includes healthy side state data and affected side state data of the patient;

[0032] A feature extraction module, configured to extract a reference motion state sequence and a target motion state sequence of each set of motion state cycle data, and construct a reference state matrix including multiple reference motion state sequences and a target state matrix including multiple target motion state sequences;

[0033] A gait coordination sliding window detection module, configured to perform gait coordination sliding window detection on multiple sets of motion state cycle data, including generating a global motion feature sequence within each sliding window based on the target state matrix, performing gait coordination detection on multiple reference motion state sequences within the sliding window according to the global motion feature sequence, generating coordination scores corresponding to the multiple reference motion state sequences respectively, and determining multiple gait coordination abnormal cycles according to the coordination scores;

[0034] A gait coordination abnormal analysis module, configured to extract abnormal motion state data and reference motion state data of each gait coordination abnormal cycle from multiple sets of motion state cycle data, perform motion stage segmentation on the multiple sets of abnormal motion state data and reference motion state data, and obtain motion decomposition data corresponding to the abnormal motion state data and reference motion state data in multiple local motion stages;

[0035] A gait local coordination detection module, configured to extract multiple motion state parameters of each set of motion decomposition data, construct a local motion state sequence of each local motion stage, perform gait local coordination detection on multiple gait coordination abnormal cycles based on the multiple local motion state sequences, and obtain a target gait abnormal detection result of the patient.

[0036] The present invention has the following beneficial effects:

[0037] The present invention performs preliminary abnormal analysis based on gait cycles on the motion detection data and electromyogram signal data of the patient, analyzes the data within multiple gait cycles based on the gait coordination sliding window detection technology, determines multiple cycles with gait coordination abnormalities in the data to be analyzed from the perspective of cycle coordination, and further performs stage decomposition on the gait cycles. For multiple abnormal cycles determined from the overall motion state of the gait cycle, analyze from multiple motion decomposition stages, combine the collaborative relevance between different local motion stages, consider the local coordination characteristics of the gait in adjacent cycles of the abnormal cycles, deeply analyze whether there is an abnormality in the local coordination of the patient's gait, and perform high-precision mining detection on possible minor abnormalities in the patient's gait, facilitating medical staff to provide personalized rehabilitation training plans for the patient and improving the rehabilitation effect of the patient. Brief Description of the Drawings

[0038] Figure 1 The figure is a schematic flowchart of a patient gait analysis method based on motion detection data provided in one embodiment of the present invention. Detailed Embodiments

[0039] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0040] As Figure 1 shown, a patient gait analysis method based on motion detection data provided in one embodiment of the present invention accurately analyzes the gait-related detection data of stroke patients in the late stage of rehabilitation, identifies gait deviations from different angles to detect subtle gait abnormalities, so as to help clinical staff accurately evaluate the gait recovery of patients and facilitate the provision of personalized rehabilitation training programs for patients. Specifically, the method includes the following steps:

[0041] Step S1: Obtain the motion detection data and electromyogram signal data of the patient to be analyzed, and extract the motion state cycle data corresponding to multiple target gait cycles from the motion detection data and electromyogram signal data to be analyzed.

[0042] Among them, the motion detection data of the patient can be collected through wearable motion detection devices such as inertial measurement units and pressure sensors to reflect the motion state of the patient at different times during walking, and the electromyogram signal data of the patient during this walking process can be collected through an electromyogram signal acquisition instrument to reflect the electromyogram activity level of the patient. After cleaning the collected original data, the data can be segmented according to the target gait cycle. The target gait cycle is used to indicate the periodically repeated motion pattern of the patient. The process from one foot touching the ground, passing through the support phase, the swing phase, and then to the other foot touching the ground is regarded as one cycle. The motion detection data and electromyogram signal data of the patient to be analyzed are segmented according to the target gait cycle, so as to obtain the motion state cycle data corresponding to each target gait cycle, which includes the healthy side state data and the affected side state data of the patient, that is, the data corresponding to the motion detection and electromyogram activity detection on the healthy side and the affected side respectively, so as to analyze the gait characteristics of the patient from the perspectives of kinematics and electromyology.

[0043] Step S2: Extract the reference motion state sequence and the target motion state sequence of each group of motion state cycle data, and construct a reference state matrix including multiple reference motion state sequences and a target state matrix including multiple target motion state sequences.

[0044] Among them, the reference motion state sequence and the target motion state sequence respectively represent the gait levels of the patient on the healthy side and the affected side within one gait cycle, which include parameters related to key changes in the motion state and parameters related to key changes in the electromyogram state within one gait cycle, such as motion state parameters like step duration, stride length, support duration, swing duration, etc., and electromyogram state parameters like electromyogram activation duration, electromyogram amplitude, etc. By splicing the reference motion state sequences and target motion state sequences corresponding to multiple target gait cycles, a reference state matrix representing the overall gait level of the healthy side during the entire movement process of the patient, and a target state matrix representing the overall gait level of the affected side are generated.

[0045] Step S3: Perform gait coordination sliding window detection on multiple groups of motion state cycle data, generate coordination scores corresponding to multiple reference motion state sequences respectively, and determine multiple gait coordination abnormal cycles according to the coordination scores.

[0046] Among them, for the reference state matrix and the target state matrix, the gait coordination sliding window detection of multiple groups of motion state cycle data is realized by means of sliding window detection. In this process, based on the time step alignment method, for multiple reference motion state sequences and target motion state sequences under the same sliding window, a global motion feature sequence within the sliding window is generated based on the target state matrix to reflect the motion pattern during the entire gait cycle. Then, gait coordination detection is performed on multiple reference motion state sequences within the sliding window according to the global motion feature sequence to evaluate the coordination between the target motion state sequence and the reference motion state sequence. Finally, the coordination scores corresponding to multiple reference motion state sequences are calculated to quantify the gait coordination degree between each reference motion state sequence and the target motion state sequence, thereby preliminarily determining multiple gait coordination abnormal cycles, that is, within some target gait cycles, the overall gait level performance of the patient shows abnormal phenomena inconsistent or uncoordinated with the healthy side, and more detailed analysis is required to determine the specific abnormal conditions.

[0047] Step S4: Extract the abnormal motion state data and reference motion state data of each gait coordination abnormal cycle from multiple groups of motion state cycle data, and perform motion stage segmentation on multiple groups of abnormal motion state data and reference motion state data to obtain the motion decomposition data corresponding to the abnormal motion state data and reference motion state data in multiple local motion stages.

[0048] Among them, after identifying the gait coordination abnormal cycle, from the multiple groups of motion state cycle data obtained from the previous segmentation, the abnormal motion state data and reference motion state data of each gait coordination abnormal cycle are extracted. The data extraction process takes the gait coordination abnormal cycle as the center, and extracts the corresponding data of the affected side and the healthy side within multiple adjacent cycles, such as the data within three or five gait cycles, to generate the abnormal motion state data and reference motion state data of the gait coordination abnormal cycle. Then, for the data used for further analysis, the motion phases are segmented. For example, a gait cycle is further decomposed into multiple local motion phases. A gait cycle as a whole can be roughly divided into a support phase and a swing phase. The support phase refers to the process from one foot touching the ground until the other foot starts to swing, and the swing phase is the process from one foot leaving the ground and going through a series of swing actions until it touches the ground again. These two major phases can be further divided into an initial segment, a middle segment, and a final segment, etc. Through this way of decomposing motion, for each local motion phase, the motion decomposition data corresponding to the phase is extracted, that is, the data related to the patient's motion state and electromyogram state within the phase, to achieve the decomposition and segmentation of the abnormal motion state data and reference motion state data of the gait coordination abnormal cycle. This is convenient for analyzing the motion characteristics of each local motion phase, and then identifying the abnormal manifestations in specific phases according to the association characteristics between different local motion phases.

[0049] Step S5: Extract multiple motion state parameters of each group of motion decomposition data, construct the local motion state sequence of each local motion phase, and perform gait local coordination detection on multiple gait coordination abnormal cycles based on the multiple local motion state sequences to obtain the target gait abnormal detection result of the patient.

[0050] Among them, for multiple groups of motion decomposition data regarding different local motion phases, more detailed motion state parameters are further extracted, such as the angle parameters of different joints in the motion detection data, the duration of the local motion phase, etc., and the electromyogram state parameters corresponding to the specific key muscle parts used to evaluate the motion state in the electromyogram signal data, etc., to detailedly characterize the patient's detailed motion state within each local motion phase. Thus, the local motion state sequence of the local motion phase is constructed according to the multiple motion state parameters of each local motion phase. Finally, through the association between different local motion states, combined with the data of the healthy side, the affected side is analyzed, and gait local coordination detection based on the coordinated association changes between different phases is performed on the multiple gait coordination abnormal cycles determined in the previous steps to identify the subtle differences between the affected side and the healthy side in each local motion phase. These small abnormal changes are crucial for the long-term rehabilitation of the patient in the later stage of rehabilitation. The finally generated target gait abnormal detection result will provide important data reference in the process of providing personalized rehabilitation training programs for the patient.

[0051] As one of the implementation processes, in the above step S3, gait coordination sliding window detection is performed on multiple groups of motion state cycle data, which specifically includes:

[0052] Through the sliding window, a time-step correlation traversal operation is performed on the reference state matrix and the target state matrix to generate coordination scores corresponding to multiple reference motion state sequences respectively.

[0053] Among them, during the traversal process, the reference motion state sequence at the center of each sliding window is analyzed to calculate the corresponding coordination score. The size of the sliding window can specifically be 3 or 5 gait cycles. For the convenience of calculation, multiple target gait cycles at the beginning and end are discarded according to the size of the sliding window, so as to calculate the coordination score of each gait cycle at the center of the sliding window with respect to the reference motion state sequence. This process is specifically generated by performing specific analysis on multiple reference motion state sequences and target motion state sequences with time-step correlation within the sliding window.

[0054] Taking multiple reference motion state sequences and target motion state sequences within any one sliding window as an example, a global motion feature sequence within the sliding window is generated based on the multiple reference motion state sequences to capture the overall dynamic changes of the multiple reference motion state sequences and reflect the trend of the entire gait cycle. In this embodiment, the mean value of the multiple reference motion state sequences within the sliding window can be taken as the global motion feature sequence within the sliding window. Then, through a distance metric algorithm such as the Euclidean distance, the similarity distances between the multiple reference motion state sequences and the global motion feature sequence are calculated respectively. Then, the standard deviation of the multiple similarity distances is calculated to obtain the gait local correlation parameter, which is used to reflect the data fluctuation characteristics of the multiple reference motion state sequences as reference data.

[0055] Stationary analysis is performed on multiple target motion state sequences to obtain fluctuation parameters. The gait local correlation parameter is processed according to the fluctuation parameters to generate an adaptive correction parameter. The similarity distance between the reference motion state sequence at the center of the sliding window and the global motion feature sequence is corrected according to the adaptive correction parameter to obtain the coordination score of the reference motion state sequence at the center of the sliding window.

[0056] Among them, the stationary analysis of multiple target motion state sequences includes constructing a covariance matrix by calculating the covariance of multiple target motion state sequences, and taking the trace of the covariance matrix as the fluctuation parameter of the multiple target motion state sequences. Then, the gait local correlation parameter is processed according to the fluctuation parameter to generate an adaptive correction parameter. Specifically, the adaptive correction parameter is the ratio of the fluctuation parameter to the gait local correlation parameter. The larger the fluctuation parameter, the more discrete the multiple target motion state sequences are, and the greater the possibility of abnormality. The smaller the gait local correlation parameter, the more referenceable the data on the healthy side is, and it will not cover up the original abnormality on the affected side due to too large fluctuations. Regarding the similarity distance between the reference motion state sequence at the center of the sliding window and the global motion feature sequence. Compared with directly measuring the coordination level between the affected side and the healthy side based on this similarity distance, the similarity distance is corrected according to the adaptive correction parameter, and the product of the two is recorded as the coordination score of the reference motion state sequence at the center of the sliding window, which can better reflect the coordination and normality of the gait. The similarity distance can use the Euclidean distance, and the smaller the value, the higher the coordination. The smaller the adaptive correction parameter, the smaller the data fluctuation in the adjacent period of the reference motion state sequence, indicating that the patient's gait tends to be stable and will not have large fluctuations due to potential abnormalities. Finally, the higher the coordination score, the greater the prediction degree. Multiple reference motion state sequences with coordination scores greater than this threshold are selected through the coordination score threshold, and the target gait cycles corresponding to the selected multiple reference motion state sequences are recorded as gait coordination abnormal cycles to reveal the difference in the overall coordination between the affected side and the healthy side considering the data fluctuations on the healthy side itself, thereby initially determining the abnormal gait cycles.

[0057] As one of the implementation processes, in step S5 above, gait local coordination detection is performed on multiple gait coordination abnormal cycles based on multiple local motion state sequences, specifically including:

[0058] For the reference motion state data of the gait coordination abnormal cycle, which includes the healthy side state data corresponding to multiple target gait cycles, the local motion state sequences corresponding to multiple groups of healthy side state data are spliced to construct a local state matrix corresponding to each group of healthy side state data. Then, multiple local state matrices are fused to generate a stage change state matrix of the gait coordination abnormal cycle, which is used to characterize the overall motion state of different local motion stages within multiple adjacent gait cycles. The stage change state matrix can be generated by taking the mean of multiple local state matrices. Then, difference analysis is performed on each local state matrix through the stage change state matrix, and a corrected state matrix of each group of healthy side state data is extracted.

[0059] Among them, the significance of the modified state matrix is ​​to characterize the difference between each group of healthy side state data and the stage change state matrix used for reference. It is worth noting that in actual data collection, it is difficult for the patient's own behavior to achieve a unified standard. Even the limbs of the healthy side will fluctuate during the actual walking process, which will cause the affected side to follow the changes. For example, the healthy side may have a slightly higher step frequency than the normal level at a certain stage, and the affected side will also increase accordingly. However, if the affected side originally has a step frequency lower than the normal level, then this situation in which the affected side follows the healthy side to respond for the sake of body coordination is likely to cover up the original abnormal situation of a step frequency slightly lower than the normal level. Therefore, by extracting the modified state matrix, the natural fluctuation phenomenon on the healthy side can be further analyzed, so as to consider its impact on the affected side based on the coordinated nature of the steps.

[0060] For the abnormal motion state data of the gait coordination abnormality cycle, which includes the affected side state data corresponding to multiple target gait cycles, the local motion state sequences corresponding to the multiple groups of affected side state data are spliced ​​to construct the stage coordination state matrix corresponding to each group of affected side state data. Then, according to the correlation between each group of abnormal motion state data and the reference motion state data, multiple stage collaborative state matrices are respectively corrected based on multiple correction state matrices, that is, through the teammate relationship between multiple groups of affected-side state data and healthy-side state data in each group of abnormal motion state data and reference motion state data, the stage collaborative state matrix corresponding to the associated affected-side state data is corrected through the correction state matrix of each group of healthy-side state data. For example, if a certain motion state parameter in the correction state matrix is ​​higher than the value in the standard stage change state matrix, then the motion state parameter in the stage collaborative state matrix corresponding to the affected-side state data needs to subtract this difference, so as to remove the data fluctuations in the affected-side state data that naturally occur during the patient's walking process as much as possible, and obtain a motion state that can better reflect the original state of the affected side. Finally, the stage collaborative state correction matrix corresponding to each group of affected-side state data is obtained through correction to generate multiple stage collaborative state correction matrices.

[0061] Finally, the gait local coordination detection is performed on the gait coordination abnormality cycle according to the multi-stage collaborative state correction matrix, and the local gait abnormality result of each gait coordination abnormality cycle is obtained.

[0062] This process includes performing local phase coordination analysis on multiple phase coordination state correction matrices to generate coordination feature parameters for each local motion phase, and performing gait local coordination detection on multiple phase coordination state correction matrices of the gait coordination abnormality cycle based on the coordination feature parameters and the phase change state matrix of the gait coordination abnormality cycle to obtain the local gait abnormality result of each gait coordination abnormality cycle.

[0063] Specifically, the process of performing local stage collaboration analysis on the multi-stage collaboration state correction matrix includes analyzing each local motion stage, extracting the corresponding local stage state sequences of each local motion stage in each stage collaboration state correction matrix, so as to obtain multiple local stage state sequences for each local motion stage. Then, autocorrelation analysis is performed on the multiple local stage state sequences of each local motion stage, and the collaboration characteristic parameters of each local motion stage with respect to the multiple local stage state sequences are calculated.

[0064] Among them, autocorrelation analysis can be used to analyze the collaboration of the performance levels of patients in multiple adjacent gait cycles under the same local motion stage. The autocorrelation value calculated by the autocorrelation analysis is used as the collaboration characteristic parameter of the local stage state sequence. For example, if the state sequence of a certain local stage shows a high degree of consistency in multiple cycles, the collaboration characteristic parameter will be relatively high, indicating that the stability of this stage in the gait cycle is better; if the autocorrelation value is low, it means that there are large fluctuations or abnormalities in this stage.

[0065] Based on the stage change state matrix, similarity analysis of the multi-stage collaboration state correction matrix of the gait coordination abnormal cycle is performed based on the local motion stage, and the similarity analysis results of the multi-stage collaboration state correction matrix of the gait coordination abnormal cycle are generated.

[0066] Among them, in the process of similarity analysis based on local motion phases, specifically, the similarity parameters for each local motion phase in the phase coordination state correction matrix are calculated first. That is, for the phase change state matrix of the gait coordination abnormal cycle and multiple phase coordination state correction matrices, first analyze each local motion phase, and calculate the similarity, such as Euclidean distance, cosine similarity, etc., between the state sequences corresponding to each local motion phase in each phase coordination state correction matrix and the phase change state matrix, so as to obtain the similarity parameters corresponding to each local motion phase. For the multiple local motion phases corresponding in the phase coordination state correction matrix, the similarity parameters corresponding to the multiple local motion phases are fused. In the fusion process, the fusion weights are determined according to the coordination feature parameters to achieve weighted fusion. The coordination feature parameters characterize the coordination characteristics under a certain local motion phase. The lower the value, the more unstable the local motion phase is in multiple adjacent gait cycles, indicating a higher possibility of abnormality, and the weight value can be relatively high, that is, the differences in this part need to be paid more attention. On the contrary, the higher the coordination feature parameter, the more stable it is, and the lower the possibility of slight abnormal fluctuations, and the weight value can be relatively low to avoid excessive attention to the differences in this phase. In this way, according to multiple coordination feature parameters, the similarity parameters of multiple local motion phases are fused to obtain the local phase similarity parameters of the gait coordination abnormal cycle with respect to each phase coordination state correction matrix, and finally the similarity analysis results of the gait coordination abnormal cycle with respect to multiple phase coordination state correction matrices are generated.

[0067] Finally, according to the similarity analysis results with respect to multiple phase coordination state correction matrices, a gait local coordination feature detection vector of the gait coordination abnormal cycle is generated, which characterizes the coordination change relationship of local phases between multiple gait cycles from the perspective of local motion phases within multiple adjacent gait cycles.

[0068] And according to the multiple local state matrices of the gait coordination abnormal cycle, a gait local coordination feature fluctuation vector of the gait coordination abnormal cycle is generated. Among them, for the construction of the gait local coordination feature fluctuation vector, by calculating the similarity parameters between the multiple local state matrices of the gait coordination abnormal cycle and the phase change state matrix of the gait coordination abnormal cycle respectively, in the same way as above, such as Euclidean distance, to calculate the similarity parameters between the multiple local state matrices and the phase change state matrix respectively. The difference is that the specific phase changes are not considered, but the overall motion state is measured. Because the local state matrix used as a reference and the phase change state matrix do not need to consider the actual possible local abnormalities, that is, they are defaulted to be in a normal state for comparison and reference, only as a reference range of fluctuations at the normal level, while for the data of the affected side, potential abnormalities need to be further explored. Therefore, the abnormal features are highlighted according to the relative importance of the fluctuations.

[0069] Subsequently, based on the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector, a local gait abnormality result of the gait coordination abnormal period is generated.

[0070] Calculate the distance parameter between the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector. If the distance parameter between the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector is less than the preset distance threshold, it is recorded that there is no local gait abnormality in the gait coordination abnormal period, that is, it is considered a normal fluctuation phenomenon, and a similar movement state fluctuation will also occur during the actual walking process of the healthy side.

[0071] Otherwise, it is recorded that there is a local gait abnormality in the gait coordination abnormal period, and the abnormal local motion stage is determined according to the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector. Specifically, the local stage coordination threshold of the gait coordination abnormal period is determined according to the gait local coordination feature fluctuation vector. For example, take the maximum value among multiple element values in the gait local coordination feature fluctuation vector as a reference, that is, the maximum fluctuation reference upper limit value, and then determine the local fluctuation upper limit value of each local motion stage. For example, if a gait cycle is divided into six local motion stages, the local fluctuation upper limit value of each local motion stage is the local stage coordination threshold, which can specifically be one-sixth of the maximum value among multiple element values in the gait local coordination feature fluctuation vector. Subsequently, the coordination difference value of the gait coordination abnormal period for each local motion stage is determined according to the local stage similarity parameter of the stage coordination state correction matrix. Specifically, during the calculation of the local stage similarity parameter, the numerical value corresponding to the similarity parameter of each local motion stage after fusing the similarity parameters of multiple local motion stages through the coordination feature parameter, that is, the numerical component used to form the local stage similarity parameter for each local motion stage.

[0072] Judge whether the coordination difference value of the local motion stage is greater than the local stage coordination threshold of the gait coordination abnormal period. If it is not greater, it indicates that the local motion stage is in a normal fluctuation state. Otherwise, the local motion stage with the coordination difference value greater than the local stage coordination threshold of the gait coordination abnormal period is recorded as an abnormal local motion stage, and a local gait abnormality result of the gait coordination abnormal period is generated to further indicate which specific local motion stage or stages are abnormal. In this way, local gait abnormality results of multiple gait coordination abnormal periods are obtained, providing an important basis for medical staff to judge the patient's gait condition and providing a direction for personalized rehabilitation training. Finally, based on the local gait abnormality results of multiple gait coordination abnormal periods, a target gait abnormality detection result of the patient is generated to achieve accurate gait analysis and abnormality detection of the patient.

[0073] One embodiment of the present invention further provides a patient gait analysis system based on motion detection data, which is specifically used to implement the above-mentioned patient gait analysis method based on motion detection data, and includes:

[0074] A data acquisition module, configured to obtain motion detection data and electromyogram signal data of a patient to be analyzed, and extract motion state cycle data corresponding to multiple target gait cycles from the motion detection data and electromyogram signal data to be analyzed. Each group of motion state cycle data includes healthy side state data and affected side state data of the patient;

[0075] A feature extraction module, configured to extract a reference motion state sequence and a target motion state sequence of each group of motion state cycle data, and construct a reference state matrix including multiple reference motion state sequences and a target state matrix including multiple target motion state sequences;

[0076] A coordinated sliding window detection module, configured to perform gait coordinated sliding window detection on multiple groups of motion state cycle data, including generating a global motion feature sequence within each sliding window based on the target state matrix, performing gait coordination detection on multiple reference motion state sequences within the sliding window according to the global motion feature sequence, generating coordination scores corresponding to the multiple reference motion state sequences respectively, and determining multiple gait coordination abnormal cycles according to the coordination scores;

[0077] Specifically, performing gait coordinated sliding window detection on multiple groups of motion state cycle data includes:

[0078] Performing a time-step association traversal operation on the reference state matrix and the target state matrix through a sliding window to generate coordination scores corresponding to multiple reference motion state sequences respectively;

[0079] Among them, for multiple reference motion state sequences and target motion state sequences within any one sliding window, a global motion feature sequence within the sliding window is generated according to the multiple reference motion state sequences, and the similarity distances between the multiple reference motion state sequences and the global motion feature sequence are calculated respectively;

[0080] Calculating the standard deviation of multiple similarity distances to obtain a gait local association parameter, performing a stationary analysis on multiple target motion state sequences to obtain a fluctuation parameter, processing the gait local association parameter according to the fluctuation parameter to generate an adaptive correction parameter, and correcting the similarity distance between the reference motion state sequence at the center of the sliding window and the global motion feature sequence according to the adaptive correction parameter to obtain the coordination score of the reference motion state sequence at the center of the sliding window;

[0081] Screening out multiple reference motion state sequences through a coordination score threshold, and recording the target gait cycles corresponding to the multiple screened reference motion state sequences as gait coordination abnormal cycles.

[0082] The gait coordination abnormality analysis module is used to extract the abnormal motion state data and reference motion state data of each gait coordination abnormal cycle from multiple groups of motion state cycle data, perform motion stage segmentation on the multiple groups of abnormal motion state data and reference motion state data, and obtain the motion disassembly data corresponding to the abnormal motion state data and reference motion state data in multiple local motion stages;

[0083] The gait local coordination detection module is used to extract multiple motion state parameters of each group of motion disassembly data, construct the local motion state sequence of each local motion stage, and perform gait local coordination detection on multiple gait coordination abnormal cycles based on the multiple local motion state sequences to obtain the target gait abnormality detection result of the patient.

[0084] The above are only the specific implementation manners of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The parts not described in detail in this specification belong to the prior art well-known to those skilled in the art.

Claims

1. A patient gait analysis method based on motion detection data, characterized in that: include: Acquire the patient's motion detection data and electromyographic signal data to be analyzed, and extract motion state cycle data corresponding to a plurality of target gait cycles from the motion detection data and electromyographic signal data to be analyzed, wherein each set of motion state cycle data includes the patient's healthy side state data and the affected side state data; Extracting a reference motion state sequence and a target motion state sequence from each set of motion state cycle data, wherein the reference motion state sequence and the target motion state sequence respectively include multiple state parameters of the patient on the healthy side and the affected side, including motion state parameters and electromyographic state parameters, and constructing a reference state matrix including multiple reference motion state sequences and a target state matrix including multiple target motion state sequences; Performing gait coordination sliding window detection on multiple groups of motion state cycle data, including generating a global motion feature sequence in each sliding window based on a target state matrix, performing gait coordination detection on multiple reference motion state sequences in the sliding window according to the global motion feature sequence, generating coordination scores corresponding to the multiple reference motion state sequences, and determining multiple gait coordination abnormality cycles according to the coordination scores; Extracting abnormal motion state data and reference motion state data of each gait coordination abnormal cycle from multiple sets of motion state cycle data, performing motion phase segmentation on the multiple sets of abnormal motion state data and reference motion state data, and obtaining motion disassembly data corresponding to the abnormal motion state data and the reference motion state data in multiple local motion phases; Extract multiple motion state parameters of each group of motion disassembly data, construct a local motion state sequence of each local motion stage, perform gait local coordination detection on multiple gait coordination abnormality cycles based on multiple local motion state sequences, and obtain the target gait abnormality detection result of the patient; Perform gait coordination sliding window detection on multiple sets of motion state cycle data, including: Through the sliding window, the reference state matrix and the target state matrix are traversed in time step association to generate coordination scores corresponding to multiple reference motion state sequences; Wherein, for a plurality of reference motion state sequences and a target motion state sequence in any sliding window, a global motion feature sequence in the sliding window is generated according to the plurality of reference motion state sequences, and similarity distances between the plurality of reference motion state sequences and the global motion feature sequence are calculated respectively; The standard deviation of multiple similarity distances is calculated to obtain the local gait correlation parameter, and the multiple target motion state sequences are subjected to stationary analysis to obtain the fluctuation parameter. The local gait correlation parameter is processed according to the fluctuation parameter to generate an adaptive correction parameter. The similarity distance between the reference motion state sequence at the center of the sliding window and the global motion feature sequence is corrected according to the adaptive correction parameter to obtain the coordination score of the reference motion state sequence at the center of the sliding window. A plurality of reference motion state sequences are screened out by using a coordination score threshold, and the target gait cycles corresponding to the screened plurality of reference motion state sequences are recorded as gait coordination abnormality cycles.

2. A patient gait analysis method based on motion detection data according to claim 1, characterized in that: The gait local coordination detection is performed on multiple gait coordination abnormality cycles based on multiple local motion state sequences, including: For the reference motion state data of the gait coordination abnormality cycle, including the healthy side state data corresponding to multiple target gait cycles, a local state matrix corresponding to each group of healthy side state data is constructed according to multiple local motion state sequences; A plurality of local state matrices are fused to generate a phase change state matrix of a gait coordination abnormality cycle, and a correction state matrix of each group of healthy side state data is determined based on the phase change state matrix; For the abnormal motion state data of the gait coordination abnormality cycle, including the affected side state data corresponding to multiple target gait cycles, a stage coordination state matrix corresponding to each group of the affected side state data is constructed according to multiple local motion state sequences; According to the correlation between each group of abnormal motion state data and the reference motion state data, the collaborative state matrices of multiple stages are respectively corrected based on multiple correction state matrices to obtain collaborative state correction matrices of multiple stages; The local gait coordination detection is performed on the gait coordination abnormality cycle according to the multi-stage coordination state correction matrix, including local stage coordination analysis on the multi-stage coordination state correction matrix to generate coordination feature parameters of each local motion stage, and based on the coordination feature parameters and the stage change state matrix of the gait coordination abnormality cycle, the local gait coordination detection is performed on the multi-stage coordination state correction matrix of the gait coordination abnormality cycle to obtain the local gait abnormality result of each gait coordination abnormality cycle.

3. A patient gait analysis method based on motion detection data according to claim 2, characterized in that: The gait local coordination detection is performed on the gait coordination abnormality cycle according to the multi-stage collaborative state correction matrix, and also includes: Performing local phase coordination analysis on multiple phase coordination state correction matrices, extracting multiple local phase state sequences of each local motion phase from the multiple phase coordination state correction matrices, performing autocorrelation analysis on the multiple local phase state sequences of each local motion phase, and calculating coordination feature parameters of the multiple local phase state sequences of each local motion phase; According to the phase change state matrix, a similarity analysis based on the local motion phase is performed on the multiple phase collaborative state correction matrices of the gait coordination abnormality cycle, including calculating the similarity parameters of each local motion phase in the phase collaborative state correction matrix, fusing the similarity parameters of the multiple local motion phases according to the multiple collaborative feature parameters, obtaining the local phase similarity parameters of the phase collaborative state correction matrix, and generating the similarity analysis results of the multiple phase collaborative state correction matrices of the gait coordination abnormality cycle; According to the similarity analysis results of the collaborative state correction matrices of multiple stages, a gait local coordination feature detection vector of the gait coordination abnormality period is generated; according to the multiple local state matrices of the gait coordination abnormality period, a gait local coordination feature fluctuation vector of the gait coordination abnormality period is generated; based on the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector, a local gait abnormality result of the gait coordination abnormality period is generated.

4. The method for analyzing patient gait based on motion detection data according to claim 3, characterized in that: A local gait abnormality result of a gait coordination abnormality period is generated based on a gait local coordination feature detection vector and a gait local coordination feature fluctuation vector, including: Calculate the distance parameter between the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector. If the distance parameter between the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector is less than a preset distance threshold, then it is recorded that there is no local gait abnormality in the gait coordination abnormality period. Otherwise, it is recorded that there is local gait abnormality in the gait coordination abnormality period, and the abnormal local motion stage is determined according to the gait local coordination feature detection vector and the gait local coordination feature fluctuation vector, including determining the local stage coordination threshold of the gait coordination abnormality period according to the gait local coordination feature fluctuation vector, determining the coordination difference value of each local motion stage of the gait coordination abnormality period according to the local stage similarity parameter of the stage coordination state correction matrix, judging whether the coordination difference value of the local motion stage is greater than the local stage coordination threshold of the gait coordination abnormality period, recording the local motion stage whose coordination difference value is greater than the local stage coordination threshold of the gait coordination abnormality period as the abnormal local motion stage, generating the local gait abnormality result of the gait coordination abnormality period, and generating the target gait abnormality detection result of the patient according to the local gait abnormality results of multiple gait coordination abnormality periods.

5. The method for analyzing patient gait based on motion detection data according to claim 3, characterized in that: For the gait local coordination feature fluctuation vector, it also includes: By calculating the similarity parameters between multiple local state matrices of the gait coordination abnormality cycle and the phase change state matrices of the gait coordination abnormality cycle, a gait local coordination characteristic fluctuation vector is constructed according to the multiple similarity parameters.

6. A patient gait analysis system based on motion detection data, characterized in that: The system is used to implement a patient gait analysis method based on motion detection data as described in any one of claims 1 to 5, comprising: A data acquisition module is used to obtain the patient's motion detection data and electromyographic signal data to be analyzed, and extract motion state cycle data corresponding to a plurality of target gait cycles from the motion detection data and electromyographic signal data to be analyzed, wherein each set of motion state cycle data includes the patient's healthy side state data and the affected side state data; A feature extraction module is used to extract a reference motion state sequence and a target motion state sequence of each set of motion state periodic data, and construct a reference state matrix containing multiple reference motion state sequences and a target state matrix containing multiple target motion state sequences; A coordinated sliding window detection module is used to perform gait coordination sliding window detection on multiple sets of motion state cycle data, including generating a global motion feature sequence in each sliding window based on a target state matrix, performing gait coordination detection on multiple reference motion state sequences in the sliding window according to the global motion feature sequence, generating coordination scores corresponding to the multiple reference motion state sequences, and determining multiple gait coordination abnormality cycles according to the coordination scores; A gait coordination anomaly analysis module is used to extract abnormal motion state data and reference motion state data of each gait coordination anomaly cycle from multiple sets of motion state cycle data, perform motion phase segmentation on multiple sets of abnormal motion state data and reference motion state data, and obtain motion disassembly data corresponding to the abnormal motion state data and reference motion state data in multiple local motion phases; The gait local coordination detection module is used to extract multiple motion state parameters from each set of motion decomposition data, construct a local motion state sequence for each local motion stage, perform gait local coordination detection on multiple gait coordination abnormality cycles based on multiple local motion state sequences, and obtain the patient's target gait abnormality detection result.

Images