Abnormal gait detection method and system based on machine learning decision classification

Abstract

The present invention provides an abnormal gait detection method and system based on machine learning decision classification, and relates to the technical field of abnormal gait detection. The present invention establishes an actual human walking model, collects acceleration, knee bending angle, and plantar pressure parameters of the human body during actual walking, performs data analysis and processing on each parameter, and generates a vertical acceleration abnormality evaluation index for reflecting the abnormal degree of up and down acceleration when a person walks, a knee bending angle abnormality index for reflecting the degree of stability of knee bending when a person walks, and a two-foot interference evaluation index for reflecting the degree of asymmetry of the movement trajectory of two feet when walking, and performs comprehensive analysis to generate an abnormal gait evaluation index reflecting the comprehensive abnormal evaluation value of human walking, thereby outputting the walking gait abnormality level, and performing further judgment based on the output result of machine learning, thereby improving the accuracy of evaluation judgment.

Description

Technical Field

[0001] The present invention relates to the technical field of abnormal gait detection, and in particular to an abnormal gait detection method and system based on machine learning decision classification. Background Art

[0002] Abnormal gait refers to a gait pattern that is significantly different from the normal gait when an individual walks. This gait abnormality may be caused by a variety of factors, including disease, injury, neurological problems, musculoskeletal diseases, aging, etc. Abnormal gait not only affects the individual's athletic ability, but may also lead to accidents such as falls, which in severe cases affect the quality of life. Abnormal gait detection is an important application in machine learning, especially in the fields of medicine, security, and human-computer interaction. The abnormal gait detection method of machine learning includes data collection, feature extraction, selection of machine learning models, model training, and abnormal gait detection.

[0003] Since the prerequisite for abnormal gait detection based on machine learning is to collect relevant abnormal gait feature values ​​and select labels for model training, its output value depends on the feature values ​​extracted from the abnormal gait, and these values ​​need to be collected manually. However, since gait abnormalities may be caused by various factors, including diseases, injuries, neurological problems, musculoskeletal diseases, aging and other factors, it is difficult to collect all sample data manually, which leads to certain inaccuracies in abnormal gait detection based on machine learning.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide an abnormal gait detection method and system based on machine learning decision classification to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An abnormal gait detection method based on machine learning decision classification, the specific steps include:

[0008] S1. Shooting a walking video of the person to be tested by a camera, analyzing the walking video, dividing each time when the person to be tested lifts his left foot as the starting point of each cycle, detecting abnormal values, locating abnormal cycles, removing frame images of abnormal cycles in the walking video, and using the video after removing the abnormal cycle frame images in the walking video as a collection reference sample, and collecting actual walking parameters of the person to be tested according to the collection reference sample, wherein the actual walking parameters include acceleration, knee bending angle, and plantar pressure;

[0009] S2, preprocessing the acceleration to generate a debiased acceleration, where the debiased acceleration is used to reflect the acceleration in the vertical direction after eliminating the offset;

[0010] S3, performing correlation analysis on the debiased acceleration to generate a vertical acceleration abnormality assessment index BYP, wherein the vertical acceleration abnormality assessment index BYP is used to reflect the abnormal degree of up and down acceleration when a person walks;

[0011] S4, performing correlation analysis on the knee bending angle to generate a knee bending angle abnormality index XWY, where the knee bending angle abnormality index XWY is used to reflect the degree of knee bending stability when a person walks;

[0012] S5, pre-processing the plantar pressure to generate a two-foot interference evaluation index r, where the two-foot interference evaluation index r is used to reflect the asymmetry of the movement trajectories of the two feet during walking;

[0013] S6. Perform correlation analysis on the foot interference assessment index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality assessment index BYP to generate an abnormal gait assessment index YP, which is used to reflect the comprehensive abnormality assessment value of human walking;

[0014] S7. Abnormal gait assessment index YP and threshold Compare and output the abnormal level of walking gait.

[0015] Furthermore, in S1, nine-axis attitude angle sensors are respectively arranged on both sides of the human waist to collect acceleration, rotation angle sensors are respectively arranged on the inner sides of the human knees to collect knee bending angles, and thin film pressure sensors are arranged on the heels of the human body to collect plantar pressure, and the acceleration, knee bending angle and plantar pressure collection frequency are all 0.01 seconds.

[0016] Furthermore, in S1, the walking video is analyzed to collect all acceleration parameters, wherein all acceleration parameters include all human body acceleration values ​​at corresponding time nodes in the walking video, and the human body acceleration values ​​are integrated into a walking acceleration set, and used Indicates that It is used to represent the acceleration value at the hth time node. C is a positive integer. The correlation analysis is performed on the set of walking accelerations to generate the mean value of walking acceleration. , the formula based on is:

[0017]

[0018] Defining detection parameters , the formula based on is:

[0019] Detection parameters The subscript h is used to index the time node. and The part is defined as an outlier, and the corresponding value of h is recorded. The abnormal loop is indexed according to the corresponding value of h, and the abnormal loop in the walking video is removed from the walking video.

[0020] The video after removing abnormal cycle frame images in the walking video is used as the collection sample, and the collection acceleration A is ,in Used to represent the acceleration value at the i-th sampling moment, where N is a positive integer.

[0021] Furthermore, the acceleration A is correlated to generate the vertical acceleration , the formula based on is:

[0022]

[0023] The vertical acceleration It is used to reflect the acceleration component in the vertical direction when the human body is actually walking. is the knee bending angle at the i-th sampling moment, is the acceleration due to gravity;

[0024] Vertical acceleration Perform correlation analysis to generate debiased acceleration , the formula based on is:

[0025]

[0026] Deflection acceleration Used to reflect the acceleration in the vertical direction after eliminating the offset.

[0027] Furthermore, the deflection acceleration Correlation processing is performed to generate the acceleration intensity index BQZ, acceleration dispersion index BWZ, and acceleration deviation index BDZ, based on the following formula:

[0028] Among them, the acceleration intensity index BQZ is used to reflect the average state statistics of acceleration, the acceleration dispersion index BWZ is used to reflect the discrete state statistics of acceleration, and the acceleration deviation index BDZ is used to reflect the degree of periodic deviation of acceleration;

[0029] The acceleration intensity index BQZ, the acceleration dispersion index BWZ, and the acceleration deviation index BDZ are subjected to correlation analysis to generate the vertical acceleration anomaly assessment index BYP, based on the following formula:

[0030] in, is the acceleration stability weight factor, The value range is ,The vertical acceleration abnormal assessment index BYP is used to reflect the abnormal degree of up and down acceleration when a person walks.

[0031] Furthermore, the knee bending angle sampled by the rotation angle sensor is used as a set Indicates that the knee angle The collection is ,in It is used to represent the knee bending angle at the i-th sampling moment. Correlation analysis was performed to generate the knee angle abnormality index XWY, based on the formula:

[0032] in, For knee angle collection Knee bend angle The minimum value of The abnormal knee bending factor, The value range is The knee angle abnormality index XWY is used to reflect the stability of the knee bending when a person walks.

[0033] Furthermore, the plantar pressure includes left foot pressure and right foot pressure. The collection is , It is used to represent the left foot pressure value at the i-th sampling moment, and the right foot pressure The collection is , Used to represent the right foot pressure value at the i-th sampling moment, where N is a positive integer.

[0034] Further, the pressure on the left foot The correlation analysis is performed with the right foot pressure FR to generate the left foot mean pressure ZJY and the right foot mean pressure YJY, based on the formula:

[0035] The left foot mean pressure ZJY is used to reflect the left foot mean pressure, and the right foot mean pressure YJY is used to reflect the right foot mean pressure;

[0036] The correlation analysis of the left foot mean pressure ZJY and the right foot mean pressure YJY is performed to generate the bilateral interference evaluation index r, based on the formula:

[0037] The foot interference assessment index r is used to reflect the degree of asymmetry of the movement trajectory of the two feet during walking.

[0038] Furthermore, correlation analysis is performed on the foot interference evaluation index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality evaluation index BYP to generate the abnormal gait evaluation index YP, based on the following formula:

[0039] The abnormal gait assessment index YP is used to reflect the comprehensive abnormal assessment value of human walking;

[0040] Threshold Set to 5.87, when , the output walking gait abnormality level is level 3, indicating that the confidence level of the human body having abnormal gait is 95%; when , the output walking gait abnormality level is level 2, indicating that the confidence level of the human body having abnormal gait is 70%; when , the output walking gait abnormality level is level one, indicating that the confidence level that the human body has an abnormal gait is 10%.

[0041] The present invention also provides an abnormal gait detection system based on machine learning decision classification, which is used to perform an abnormal gait detection method based on machine learning decision classification, including:

[0042] A parameter acquisition module, wherein the parameter acquisition module is used to capture a walking video of the person to be tested through a camera and collect the actual walking parameters of the person to be tested in real time, and to match the time nodes of the walking parameters with the time nodes of the walking video;

[0043] An acceleration analysis module, the acceleration analysis module is used to pre-process the acceleration, generate a debiased acceleration, and perform a correlation analysis on the debiased acceleration to generate a vertical acceleration anomaly assessment index BYP;

[0044] A knee angle analysis module, which is used to perform a correlation analysis on the knee angle and generate a knee angle abnormality index XWY;

[0045] A plantar pressure analysis module, which is used to pre-process the plantar pressure and generate a double-foot interference evaluation index r;

[0046] A comprehensive analysis module, which is used to perform correlation analysis on the foot interference evaluation index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality evaluation index BYP to generate an abnormal gait evaluation index YP;

[0047] A comparison module is used to compare the abnormal gait evaluation index YP with the threshold Compare and output the abnormal level of walking gait.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention establishes an actual walking model of the human body, collects acceleration, knee bending angle, and plantar pressure parameters of the human body during actual walking, performs data analysis and processing on each parameter, and generates a vertical acceleration abnormality evaluation index for reflecting the abnormal degree of up and down acceleration when a person walks, a knee bending angle abnormality index for reflecting the stability of the knee bending when a person walks, and a two-foot interference evaluation index for reflecting the degree of asymmetry of the movement trajectories of the two feet when walking, and performs comprehensive analysis to generate an abnormal gait evaluation index reflecting the comprehensive abnormal evaluation value of human walking, thereby outputting the walking gait abnormality level, and making further judgments based on the output results of machine learning, thereby improving the accuracy of evaluation and judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0051] Figure 2 It is a schematic diagram of the system module of the present invention. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0053] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0054] Example:

[0055] See also Figure 1 , the present invention provides a technical solution:

[0056] A method for detecting abnormal gait based on machine learning decision classification. Abnormal gait detection is an important application in machine learning, especially in the fields of medical care, security, and human-computer interaction. Abnormal gait usually refers to a walking pattern that is significantly different from a normal gait, which may be due to various reasons, such as physical illness, injury, aging or other factors. The abnormal gait detection method of machine learning includes data collection, feature extraction, selection of machine learning models, model training, and abnormal gait detection. The present invention is used to perform secondary verification on the model output by machine learning, thereby improving the detection accuracy. The specific steps include:

[0057] Step 1, shoot a walking video of the person to be tested by a camera, analyze the walking video, divide each time the left foot of the person to be tested is lifted as the starting point of each cycle, detect abnormal values, and locate abnormal cycles, remove frame images of abnormal cycles in the walking video, and use the video after removing the abnormal cycle frame images in the walking video as an acquisition reference sample, and collect actual walking parameters of the person to be tested according to the acquisition reference sample. The actual walking parameters include acceleration, knee bending angle, and plantar pressure. Nine-axis attitude angle sensors are respectively set on both sides of the human waist to collect acceleration, and rotation angle sensors are respectively set on the inner side of the human knee to collect knee bending angle. When sampling data, the output data of the two nine-axis attitude angle sensors and the rotation angle sensor are averaged, and a thin film pressure sensor is set on the heel of the human body to collect plantar pressure. The acquisition frequency of acceleration, knee bending angle, and plantar pressure is 0.01 second.

[0058] The walking video is analyzed to collect all acceleration parameters, which include all human body acceleration values ​​at corresponding time nodes in the walking video, and the human body acceleration values ​​are integrated into a walking acceleration set, and used Indicates that It is used to represent the acceleration value at the hth time node. C is a positive integer. The correlation analysis is performed on the set of walking accelerations to generate the mean value of walking acceleration. , the formula based on is:

[0059]

[0060] Defining detection parameters , the formula based on is:

[0061] Detection parameters The subscript h is used to index the time node. and The part is defined as an abnormal value, and the corresponding value of h is recorded. The abnormal cycle is indexed according to the corresponding value of h, and the abnormal cycle in the walking video is removed from the walking video to remove abnormal data.

[0062] The video after removing abnormal cycle frame images in the walking video is used as the collection sample, and the collection acceleration A is ,in Used to represent the acceleration value at the i-th sampling moment, where N is a positive integer.

[0063] Step 2: pre-process the acceleration to generate debiased acceleration, which is used to reflect the acceleration in the vertical direction after eliminating the offset;

[0064] Correlation processing is performed on acceleration A to generate vertical acceleration , the formula based on is:

[0065]

[0066] The vertical acceleration It is used to reflect the acceleration component in the vertical direction when the human body is actually walking. The larger the value, the greater the acceleration component, and the higher the degree of abnormal gait of the human body. is the knee bending angle at the i-th sampling moment, is the acceleration due to gravity;

[0067] In order to make the calculation results more accurate and stable, the vertical acceleration Perform correlation analysis to generate debiased acceleration , the formula based on is:

[0068]

[0069] Deflection acceleration It is used to reflect the acceleration in the vertical direction after eliminating the offset, which can make the calculation results more accurate and stable.

[0070] Step 3: Perform correlation analysis on the debiased acceleration to generate a vertical acceleration anomaly assessment index BYP, which is used to reflect the degree of abnormality of the vertical acceleration when a person walks;

[0071] Deflection acceleration Correlation processing is performed to generate the acceleration intensity index BQZ, acceleration dispersion index BWZ, and acceleration deviation index BDZ, based on the following formula:

[0072] Among them, the acceleration intensity index BQZ is used to reflect the average state statistics of acceleration, the acceleration dispersion index BWZ is used to reflect the discrete state statistics of acceleration, and the acceleration deviation index BDZ is used to reflect the degree of periodic deviation of acceleration;

[0073] The acceleration intensity index BQZ, the acceleration dispersion index BWZ, and the acceleration deviation index BDZ are subjected to correlation analysis to generate the vertical acceleration anomaly assessment index BYP, based on the following formula:

[0074] in, is the acceleration stability weight factor, The value range is The vertical acceleration abnormality assessment index BYP is used to reflect the abnormal degree of up and down acceleration when a person walks. The larger the value, the larger the up and down acceleration value, and the higher the degree of abnormal walking gait.

[0075] Step 4, performing correlation analysis on the knee bending angle to generate a knee bending angle abnormality index XWY, where the knee bending angle abnormality index XWY is used to reflect the stability of the knee bending when a person walks;

[0076] The knee bend angle sampled by the rotation angle sensor is collected Indicates that the knee angle The collection is ,in It is used to represent the knee bending angle at the i-th sampling moment. Correlation analysis was performed to generate the knee angle abnormality index XWY, based on the formula:

[0077] in, For knee angle collection Knee bend angle The minimum value of The abnormal knee bending factor, The value range is The knee angle abnormality index XWY is used to reflect the stability of the knee bending when a person walks. The larger the value, the more stable the knee bending is and the lower the abnormality is.

[0078] Step 5: Preprocess the plantar pressure to generate a two-foot interference evaluation index r, where the two-foot interference evaluation index r is used to reflect the asymmetry of the movement trajectories of the two feet during walking;

[0079] Plantar pressure includes left foot pressure and right foot pressure. The collection is , Used to indicate the left foot pressure value at the i-th sampling moment, the right foot pressure The collection is , Used to represent the right foot pressure value at the i-th sampling moment, where N is a positive integer.

[0080] Pressure on left foot The correlation analysis is performed with the right foot pressure FR to generate the left foot mean pressure ZJY and the right foot mean pressure YJY, based on the formula:

[0081] The left foot mean pressure ZJY is used to reflect the left foot mean pressure, and the right foot mean pressure YJY is used to reflect the right foot mean pressure;

[0082] The correlation analysis of the left foot mean pressure ZJY and the right foot mean pressure YJY is performed to generate the bilateral interference evaluation index r, based on the formula:

[0083] The bipedal interference assessment index r is used to reflect the degree of asymmetry of the bipedal motion trajectory when walking. The larger the value of the bipedal interference assessment index r is, the higher the degree of asymmetry of the bipedal motion trajectory when walking, and the higher the degree of abnormal gait.

[0084] Step 6: perform correlation analysis on the foot interference assessment index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality assessment index BYP to generate an abnormal gait assessment index YP, which is used to reflect the comprehensive abnormality assessment value of human walking;

[0085] The correlation analysis of the foot interference assessment index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality assessment index BYP was performed to generate the abnormal gait assessment index YP, based on the following formula:

[0086] The abnormal gait assessment index YP is used to reflect the comprehensive abnormal assessment value of human walking;

[0087] Step 7: Compare the abnormal gait assessment index YP with the threshold Compare and output the abnormal level of walking gait.

[0088] Threshold Set to 5.87, when , the output walking gait abnormality level is level 3, indicating that the confidence level of the human body having abnormal gait is 95%; when , the output walking gait abnormality level is level 2, indicating that the confidence level of the human body having abnormal gait is 70%; when , the output walking gait abnormality level is level 1, indicating that the confidence level of the human body having an abnormal gait is 10%. The confidence level is used to reflect the credibility of the walking gait. The larger the value, the more abnormal the human body's walking gait is, and the larger the abnormal gait evaluation index YP, the higher the confidence level.

[0089] See also Figure 2The present invention also provides an abnormal gait detection system based on machine learning decision classification, which is used to perform an abnormal gait detection method based on machine learning decision classification, including:

[0090] A parameter acquisition module, wherein the parameter acquisition module is used to capture a walking video of the person to be tested through a camera and collect the actual walking parameters of the person to be tested in real time, and to match the time nodes of the walking parameters with the time nodes of the walking video;

[0091] The acceleration analysis module is used to pre-process the acceleration, generate the debiased acceleration, and perform correlation analysis on the debiased acceleration to generate the vertical acceleration anomaly assessment index BYP;

[0092] A knee angle analysis module is used to perform correlation analysis on the knee angle and generate a knee angle abnormality index XWY;

[0093] The plantar pressure analysis module is used to pre-process the plantar pressure and generate a foot interference evaluation index r;

[0094] A comprehensive analysis module is used to perform correlation analysis on the foot interference evaluation index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality evaluation index BYP to generate an abnormal gait evaluation index YP;

[0095] Comparison module, which is used to compare the abnormal gait assessment index YP with the threshold Compare and output the abnormal level of walking gait.

[0096] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0097] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0098] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0099] The above are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A method for detecting abnormal gait based on machine learning decision classification, characterized in that: The specific steps include: S1. Shooting a walking video of the person to be tested by a camera, analyzing the walking video, dividing each time when the person to be tested lifts his left foot as the starting point of each cycle, detecting abnormal values, locating abnormal cycles, removing frame images of abnormal cycles in the walking video, and using the video after removing the abnormal cycle frame images in the walking video as a collection reference sample, and collecting actual walking parameters of the person to be tested according to the collection reference sample, wherein the actual walking parameters include acceleration, knee bending angle, and plantar pressure; S2, preprocessing the acceleration to generate a debiased acceleration, where the debiased acceleration is used to reflect the acceleration in the vertical direction after eliminating the offset; S3, performing correlation analysis on the debiased acceleration to generate a vertical acceleration abnormality assessment index BYP, wherein the vertical acceleration abnormality assessment index BYP is used to reflect the abnormal degree of up and down acceleration when a person walks; S4, performing correlation analysis on the knee bending angle to generate a knee bending angle abnormality index XWY, where the knee bending angle abnormality index XWY is used to reflect the degree of knee bending stability when a person walks; S5, pre-processing the plantar pressure to generate a two-foot interference evaluation index r, where the two-foot interference evaluation index r is used to reflect the asymmetry of the movement trajectories of the two feet during walking; S6. Perform correlation analysis on the foot interference assessment index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality assessment index BYP to generate an abnormal gait assessment index YP, which is used to reflect the comprehensive abnormality assessment value of human walking; S7, abnormal gait assessment index YP and threshold Perform comparison and output the abnormal level of walking gait; In S1, the walking video is analyzed to collect all acceleration parameters, which include all human body acceleration values ​​at corresponding time nodes in the walking video, and the human body acceleration values ​​are integrated into a walking acceleration set, and used Indicates that It is used to represent the acceleration value at the hth time node. C is a positive integer. The correlation analysis is performed on the set of walking accelerations to generate the mean value of walking acceleration. , the formula based on is: Defining detection parameters , the formula based on is: Detection parameters The subscript h is used to index the time node. and The part is defined as an abnormal value, and the corresponding value of h is recorded. The abnormal cycle is indexed according to the corresponding value of h, and the abnormal cycle in the walking video is removed from the walking video; The video after removing abnormal cycle frame images in the walking video is used as the collection sample, and the collection acceleration A is ,in Used to represent the acceleration value at the i-th sampling moment, N is a positive integer; Correlation processing is performed on acceleration A to generate vertical acceleration , the formula based on is: The vertical acceleration It is used to reflect the acceleration component in the vertical direction when the human body is actually walking. is the knee bending angle at the i-th sampling moment, is the acceleration due to gravity; Vertical acceleration Perform correlation analysis to generate debiased acceleration , the formula based on is: Deflection acceleration Used to reflect the acceleration in the vertical direction after eliminating the offset; Deflection acceleration Correlation processing is performed to generate the acceleration intensity index BQZ, acceleration dispersion index BWZ, and acceleration deviation index BDZ, based on the following formula: Among them, the acceleration intensity index BQZ is used to reflect the average state statistics of acceleration, the acceleration dispersion index BWZ is used to reflect the discrete state statistics of acceleration, and the acceleration deviation index BDZ is used to reflect the degree of periodic deviation of acceleration; The acceleration intensity index BQZ, the acceleration dispersion index BWZ, and the acceleration deviation index BDZ are subjected to correlation analysis to generate the vertical acceleration anomaly assessment index BYP, based on the following formula: in, is the acceleration stability weight factor, The value range is ,The abnormal assessment index of vertical acceleration BYP is used to reflect the abnormal degree of up and down acceleration when a person walks; The knee bend angle sampled by the rotation angle sensor is collected Indicates that the knee angle The collection is ,in It is used to represent the knee bending angle at the i-th sampling moment. Correlation analysis was performed to generate the knee angle abnormality index XWY, based on the formula: in, is the maximum value in the knee angle set, Knee bend angle The minimum value of The abnormal knee bending factor, The value range is , the knee angle abnormality index XWY is used to reflect the stability of knee bending when people walk; The plantar pressure includes left foot pressure and right foot pressure. The collection is , It is used to represent the left foot pressure value at the i-th sampling moment, and the right foot pressure The collection is , It is used to represent the right foot pressure value at the i-th sampling moment, where N is a positive integer; Pressure on left foot The correlation analysis is performed with the right foot pressure FR to generate the left foot mean pressure ZJY and the right foot mean pressure YJY, based on the formula: The left foot mean pressure ZJY is used to reflect the left foot mean pressure, and the right foot mean pressure YJY is used to reflect the right foot mean pressure; The correlation analysis of the left foot mean pressure ZJY and the right foot mean pressure YJY is performed to generate the bilateral interference evaluation index r, based on the formula: The foot interference assessment index r is used to reflect the degree of asymmetry of the movement trajectory of the two feet during walking; The correlation analysis of the foot interference assessment index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality assessment index BYP was performed to generate the abnormal gait assessment index YP, based on the following formula: The abnormal gait assessment index YP is used to reflect the comprehensive abnormal assessment value of human walking; Threshold Set to 5.87, when , the output walking gait abnormality level is level 3, indicating that the confidence level of the human body having abnormal gait is 95%; when , the output walking gait abnormality level is level 2, indicating that the confidence level of the human body having abnormal gait is 70%; when , the output walking gait abnormality level is level one, indicating that the confidence level that the human body has an abnormal gait is 10%.

2. The abnormal gait detection method based on machine learning decision classification according to claim 1, characterized in that: In S1, nine-axis attitude angle sensors are respectively arranged on both sides of the human waist to collect acceleration, rotation angle sensors are respectively arranged on the inner sides of the human knees to collect knee bending angles, and thin film pressure sensors are arranged on the heels of the human body to collect plantar pressure. The acceleration, knee bending angle and plantar pressure collection frequencies are all 0.01 seconds.

3. An abnormal gait detection system based on machine learning decision classification, used to execute the abnormal gait detection method based on machine learning decision classification according to claim 1, characterized in that: include: A parameter acquisition module, wherein the parameter acquisition module is used to capture a walking video of the person to be tested through a camera and collect the actual walking parameters of the person to be tested in real time, and to match the time nodes of the walking parameters with the time nodes of the walking video; An acceleration analysis module, the acceleration analysis module is used to pre-process the acceleration, generate a debiased acceleration, and perform a correlation analysis on the debiased acceleration to generate a vertical acceleration anomaly assessment index BYP; A knee angle analysis module, which is used to perform a correlation analysis on the knee angle and generate a knee angle abnormality index XWY; A plantar pressure analysis module, which is used to pre-process the plantar pressure and generate a double-foot interference evaluation index r; A comprehensive analysis module, which is used to perform correlation analysis on the foot interference evaluation index r, the knee bending angle abnormality index XWY, and the vertical acceleration abnormality evaluation index BYP to generate an abnormal gait evaluation index YP; A comparison module is used to compare the abnormal gait evaluation index YP with the threshold Compare and output the abnormal level of walking gait.

Images