Parkinson's disease patient action analysis system and method based on wearable equipment

Abstract

The invention belongs to the field of action analysis, and particularly relates to a Parkinson's disease patient action analysis system and method based on wearable equipment. The Parkinson's disease patient action analysis system based on the wearable device comprises an acquisition module, a processing module, a calculation module and an evaluation module, the acquisition module comprises a wearable bracelet, the wearable bracelet is provided with a three-axis acceleration sensor, and the three-axis acceleration sensor is used for acquiring linear acceleration of the hand of the patient in three orthogonal directions as first information; the processing module is used for acquiring the first information, calculating the movement speed of the hand of the patient in the horizontal direction as first data according to the acceleration in the horizontal direction and the acquisition time in the first information, and calculating the movement speed of the hand of the patient in the vertical direction as second data according to the acceleration in the vertical direction and the acquisition time in the first information. In conclusion, the scheme solves the problems that the monitoring data of the Parkinson's disease patient is not comprehensive and the accuracy is limited.

Description

Technical Field

[0001] This solution belongs to the field of motion analysis, and specifically relates to a Parkinson's patient motion analysis system and method based on wearable devices. Background Art

[0002] Parkinson's Disease (PD) is a common chronic neurodegenerative disease. The main pathological feature is the degeneration of dopaminergic neurons in the substantia nigra, resulting in motor symptoms such as bradykinesia, resting tremor, and muscle rigidity in patients. The traditional assessment of the motor ability of Parkinson's patients (i.e., Parkinson's patients) mainly relies on clinical scales, such as the Unified Parkinson's Disease Rating Scale (UPDRS) and its revised version MDS-UPDRS. These scales evaluate the hand motor ability by the observation and scoring of medical staff on patients. However, there may be significant scoring differences among different medical staff, resulting in limited repeatability and accuracy of the assessment results.

[0003] Therefore, a Chinese patent discloses a Parkinson's tremor assessment device based on a wearable inertial sensor, which designs various motion paradigms for detecting tremor indicators, and collects the three-axis acceleration signals of the test object under different motion paradigms; preprocesses the three-axis acceleration signals to obtain three-dimensional effective tremor signals; performs integration and dimensionality reduction processing on the three-dimensional effective tremor signals to obtain one-dimensional effective tremor signals; and calculates the index result information of each of the evaluation indicators according to the one-dimensional effective tremor signals to evaluate the Parkinson's tremor of the test object.

[0004] However, when a Parkinson's tremor assessment device based on a wearable inertial sensor collects three-axis acceleration signals, in order to accurately collect tremor signals, patients need to cooperate according to preset motion paradigms (such as specific hand movements, static postures, etc.), making it difficult to comprehensively collect data in daily life, resulting in incomplete and inaccurate monitoring results. Summary of the Invention

[0005] The purpose of this solution is to provide a Parkinson's patient motion analysis system and method based on wearable devices to solve the problem of incomplete and inaccurate monitoring data of Parkinson's patients.

[0006] To achieve the above purpose, this solution provides a Parkinson's patient motion analysis system based on wearable devices, including:

[0007] A collection module, including a wearable bracelet, on which a three-axis acceleration sensor is provided. The three-axis acceleration sensor is used to collect the linear accelerations in three orthogonal directions of the patient's hand as the first information;

[0008] A processing module that obtains first information, calculates the movement speed of the patient's hand in the horizontal direction as first data based on the acceleration and acquisition time in the horizontal direction in the first information, and calculates the movement speed of the patient's hand in the vertical direction as second data based on the acceleration and acquisition time in the vertical direction in the first information;

[0009] A calculation module, including a human upper limb model, for calculating the movement trajectory of the patient's hand according to the first data and the second data, and for combining the hand movement trajectory with the human upper limb model to calculate a first upper limb model; the calculation module also includes deep learning models for various human upper limb movement trainings, and the calculation module is used to put the first upper limb model into the deep learning model to obtain human actions; the calculation module is also used to calculate the tremor frequency according to the change frequencies of the first data and the second data, and associate the tremor frequency with the human actions;

[0010] An evaluation module, including a scale established based on human actions, tremor frequency, and degree of illness, for obtaining the tremor frequency and associated human actions, and matching the degree of illness according to the tremor frequency and associated human actions.

[0011] And, a method for analyzing the actions of Parkinson's patients based on a wearable device using a Parkinson's patient action analysis system based on a wearable device.

[0012] The principle and technical effect of this solution are as follows: This solution can collect data in real time during the patient's daily life without the patient's cooperation according to a preset action paradigm. It can not only obtain more comprehensive movement data but also enable the patient not to deliberately cooperate with specific actions, improving the comfort when measuring tremor data in this solution.

[0013] At the same time, this solution combines acceleration signals, can capture the patient's movement characteristics more comprehensively, and improve the accuracy of monitoring. At the same time, it can also refine or expand the depth and breadth of judgment and recognition in this solution by expanding the learning training data of the deep learning model and the scale established based on human actions, tremor frequency, and degree of illness, thereby improving the accuracy of action recognition.

[0014] In summary, this solution solves the problem of incomplete and limited accuracy of monitoring data for Parkinson's patients.

[0015] Furthermore, the processing module obtains the acquisition time of the first information, obtains the temporal characteristics between the first data and the second data and human actions, including but not limited to the mean, variance, skewness, kurtosis of the acceleration signal, and the cross-correlation coefficient of the three-axis accelerations; combines a preset temporal model to perform real-time monitoring and analysis of the movement state of the patient's hand.

[0016] By extracting time-series features such as the mean, variance, skewness, kurtosis of the acceleration signal, and the cross-correlation coefficients of the three-axis accelerations, this solution can more accurately identify the differences between tremor signals and voluntary movement signals. Moreover, tremor signals usually have a specific frequency range (such as 4 - 6 Hz) and regular time-series patterns, while voluntary movement signals are more complex and irregular. Through time-series feature analysis, these two types of signals can be effectively distinguished, reducing misjudgment.

[0017] Furthermore, the processing module obtains the acquisition time corresponding to the human movement, associates the acquisition time with the tremor frequency, takes the time with adjacent acquisition times and the same human movement as the duration, obtains the tremor frequency associated with the duration, extracts the time-series change features of the tremor frequency, and based on the time-series change features, combines with a preset frequency threshold to judge the severity of the tremor, and performs correlation analysis on the tremor frequency and the time-series features of the human movement to achieve a quantitative assessment of the movement disorder in Parkinson's disease patients.

[0018] By extracting the time-series change features of the tremor frequency (such as frequency fluctuations, duration, etc.) and combining with a preset frequency threshold, the severity of the tremor can be judged more accurately, enabling this solution to monitor the changes in the tremor frequency in real time, capture the dynamic features of the tremor, and thus provide richer data for the quantitative assessment of the condition.

[0019] Furthermore, the acquisition module further includes a camera for acquiring image information around the hand. The processing module is used to obtain the image information, identify and extract the hand contour in the image information, identify the object contour in contact with the hand contour, and then obtain the tremor frequency when the object contour in contact with the hand contour. The calculation module is used to obtain the complexity of the hand movement according to the change speed of the hand contour, and match the degree of illness according to the complexity of the hand movement and the tremor frequency.

[0020] Furthermore, the camera is a depth camera. The processing module is further used to identify and label the human contour features in the image information as other body features, establish a relative model of the other body features in combination with the hand position, and fuse the relative model into the first upper limb model according to the relative position between the relative model and the hand.

[0021] By collecting the image information around the hand through a depth camera, the system can identify the hand contour and the contour of the object in contact with the hand. This multi-modal data fusion (depth information + acceleration information) can not only monitor hand movements more accurately, but also obtain the tremor frequency more accurately. It can also quantify the complexity of hand movements through the change speed of the hand contour, enabling this solution to more comprehensively evaluate the motor function of Parkinson's disease patients, thereby improving the accuracy of this solution. Moreover, the depth information provided by the depth camera can reduce misjudgments caused by background interference or poor lighting conditions. Combining multi-modal data, this solution can more reliably identify tremors and movement patterns, improving the accuracy of diagnosis.

[0022] Furthermore, the processing module is also used to obtain the first upper limb model and the corresponding tremor frequency, and establish a tremor prediction model based on the first upper limb model and the corresponding tremor frequency; the calculation module puts the obtained first upper limb model into the tremor prediction model to predict the tremor frequency as the first frequency, and then obtains the tremor frequency associated with the first upper limb model as the second frequency. If the difference between the second frequency and the first frequency is greater than the preset difference, the extraction time of the human action timing feature is extended; if the difference between the first frequency and the second frequency obtained after the extension time is greater than the preset difference, the difference is used as the first difference value, and an abnormal signal is generated and sent to the evaluation module; the evaluation module receives and plays or displays the abnormal signal.

[0023] Furthermore, when playing or displaying the abnormal signal, the evaluation module is also used to play or display a preset paradigm behavior to the patient and generate a signal for increased acquisition and send it to the acquisition module; after receiving the signal for increased acquisition, the acquisition module increases the acquisition frequency of the camera; the calculation module compares the obtained first upper limb model with the preset paradigm behavior. If the comparison result is greater than the preset difference, an adjustment prompt is generated and sent to the evaluation module, and the evaluation module is used to play or display the adjustment prompt; if the comparison result is not greater than the preset difference, it is judged whether the difference between the first frequency and the second frequency and the first difference value is greater than the preset difference. If the difference between the difference value and the first difference value is not greater than the preset difference, a condition change prompt is generated, and the first frequency and the change prompt are sent to the evaluation module, and the evaluation module receives and plays or displays the first frequency and the change prompt; if the difference between the difference value and the first difference value is greater than the preset difference, the first upper limb model and the first frequency are used to train the tremor prediction model.

[0024] Furthermore, when generating a condition change prompt and sending it to the evaluation module, the calculation module is also used to associate and store the first frequency with the first upper limb model as a change behavior. The calculation module regularly statistically outputs the change behavior according to the distribution of the first upper limb model in the change behavior, and generates a change report according to the statistical distribution result and sends it to the evaluation module, and the evaluation module receives and plays or displays the change report.

[0025] The processing module establishes a tremor prediction model by obtaining the first upper limb model and the corresponding tremor frequency, enabling this solution to predict the future tremor frequency (the first frequency) based on the current motion state, providing prospective information for the change of the condition, and then comparing the prospective information with the actually measured condition to determine the abnormal state (getting better or worse) of the patient. When the patient shows abnormalities, the completion of the paradigm action by the patient and the tremor situation under the paradigm action are collected to verify the abnormal situation of the patient again. In this way, this solution can better adapt to the change of the patient's condition, improve the accuracy of early warning, and at the same time improve the judgment accuracy of this solution, making the data provided by this solution more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic structural diagram of a Parkinson's patient motion analysis system based on a wearable device in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The following will clearly and completely describe the concept of the present invention and the technical effects produced in combination with the embodiments to fully understand the purpose, features, and effects of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative efforts shall fall within the scope of protection of the present invention:

[0028] As Figure 1 shown, a Parkinson's patient motion analysis system based on a wearable device includes:

[0029] An acquisition module, including a wearable bracelet, on which a three-axis acceleration sensor is provided. The three-axis acceleration sensor is used to acquire the linear accelerations of the patient's hand in three orthogonal directions as the first information;

[0030] A processing module, which acquires the first information, calculates the motion speed of the patient's hand in the horizontal direction as the first data according to the acceleration in the horizontal direction and the acquisition time in the first information, and calculates the motion speed of the patient's hand in the vertical direction as the second data according to the acceleration in the vertical direction and the acquisition time in the first information;

[0031] A calculation module, including a human upper limb model, is used to calculate the motion trajectory of the patient's hand according to the first data and the second data, and is used to calculate the first upper limb model by combining the hand motion trajectory with the human upper limb model; the calculation module also includes a deep learning model for training various human upper limb actions. The calculation module is used to put the first upper limb model into the deep learning model to obtain human actions; the calculation module is also used to calculate the tremor frequency according to the change frequency of the first data and the second data, and associate the tremor frequency with the human actions;

[0032] An evaluation module, including a scale established based on human movements, tremor frequency, and disease severity, is used to obtain the tremor frequency and associated human movements, and match the disease severity according to the tremor frequency and associated human movements.

[0033] Among them, the processing module obtains the acquisition time of the first information, and obtains the temporal characteristics between the first data and the second data and human movements, including but not limited to the mean, variance, skewness, kurtosis of the acceleration signal, and the cross-correlation coefficient of the three-axis accelerations; combines a preset temporal model (the temporal model is used to analyze and process temporal information such as the collected acceleration signal and human movement data to achieve real-time monitoring and analysis of the patient's hand movement state. According to existing research and technical applications, the temporal model can include the following common types: dynamic time warping, long short-term memory network, temporal convolutional network, autoregressive moving average model, autoregressive integrated moving average model), and conducts real-time monitoring and analysis of the patient's hand movement state.

[0034] Among them, the processing module obtains the acquisition time corresponding to the human movement, associates the acquisition time with the tremor frequency, takes the time with adjacent acquisition times and the same human movement as the duration, obtains the tremor frequency associated with the duration, extracts the temporal change characteristics of the tremor frequency, and according to the temporal change characteristics, combines a preset frequency threshold (generally 3 Hz, and specifically can be set by medical staff according to the patient's situation) to judge the severity of the tremor, and conducts correlation analysis on the temporal characteristics of the tremor frequency and the human movement.

[0035] Specifically, a triaxial acceleration sensor is used to collect the linear acceleration signals of the patient's hand in three orthogonal directions (X, Y, Z), the continuous acceleration signals are segmented into multiple time windows, each window contains data of a certain time length (such as 1 second), and the acceleration signals are filtered to remove noise and high-frequency interference. Then, the mean value of the acceleration signal in each time window is calculated using formula (1), and formula (1) is as follows:

[0036]

[0037] Among them, N is the number of data points in the window, is the acceleration mean value, a i is the acceleration value of the i-th data point. The meaning of formula (1) is to add up the acceleration values of all data points in the window and then divide by the total number of data points to obtain the average level of the acceleration signal in this time window. Then, the variance of the acceleration signal in each time window is calculated through formula (2) and the acceleration mean value The formula (2) for calculating the variance of the acceleration signal in each time window is as follows:

[0038]

[0039] Where Var(a) is the variance of the acceleration signal. When calculating the variance, first find the difference between each data point and the mean, then square these differences, add all the squared values ​​and divide by the total number of data points. The larger the variance, the more drastic the fluctuation of the acceleration signal in the time window.

[0040] When the patient's hand movement speed has an average acceleration When it is greater than the preset average level, the stability of the patient's hand movement is obtained through the signal variance Var(a) of the hand acceleration. When the value of Var(a) is larger, it means that the patient's hand movement is more unstable. At this time, the symmetry of the patient's hand movement is obtained through the acceleration variance Var(a) and formula (3). Formula (3) is as follows:

[0041]

[0042] When the symmetry of the patient's hand movement is greater than the preset value, it means that the patient is tremoring. At this time, the acquisition time of the acceleration signal is taken as the tremor time. The kurtosis Kurt(a) of the acceleration during the patient's hand movement is obtained according to the acceleration in the time window corresponding to the tremor time and formula (4). Formula (4) is as follows:

[0043]

[0044] Formula (4) can measure the sharpness of the acceleration signal distribution, that is, it can reflect the suddenness of the patient's hand movement. The processing module converts the kurtosis Kurt (a), symmetry and and The change cycle is sent to the evaluation module, which puts these values ​​into the scale, conducts a multi-dimensional evaluation of the patient's current condition, and outputs accurate comprehensive evaluation results.

[0045] In this embodiment, the linear acceleration signal of the patient's hand is collected by a three-axis acceleration sensor, and it is divided into multiple time windows for filtering processing, and then the mean, variance, skewness and kurtosis and other characteristics in each window are calculated. These features describe the characteristics of hand movement from different angles, and can monitor and analyze the movement state in real time, and detect abnormal patterns such as tremors in time. By setting a threshold, this solution can detect abnormal states of hand movement and issue an early warning. At the same time, these features are input into the evaluation module, combined with the scale for multi-dimensional evaluation, to provide comprehensive decision support for patients and their doctors, improve diagnostic accuracy, and reduce misjudgment. In addition, the solution supports personalized treatment recommendations, helps to observe the long-term trend of the disease, and thus better manage the condition of Parkinson's patients.

[0046] Among them, the acquisition module further includes a camera for acquiring image information around the hand. The processing module is used to obtain the image information, identify and extract the hand contour in the image information, identify the object contour in contact with the hand contour, and obtain the tremor frequency when the object contour in contact with the hand contour is touched. The calculation module is used to obtain the complexity of the hand movement according to the change speed of the hand contour, and match the degree of illness according to the complexity of the hand movement and the tremor frequency.

[0047] Among them, the camera is a depth camera. The processing module is further used to identify and label the human body contour features in the image information as other body features, combine the other body features with the hand position to establish a relative model of the other body features, and fuse the relative model into the first upper limb model according to the relative position between the relative model and the hand.

[0048] Specifically, when the hand contour of the patient contacts the object contour, the monitoring window is opened. The acquisition module acquires the image information of the patient's hand contour and sends it to the processing module. The processing module compares the image information acquired twice adjacent to each other to calculate the change of the patient's hand contour, obtains the complexity of the patient's hand movement through the change of the patient's hand contour, and simultaneously obtains the average value of the hand acceleration of the patient. The kurtosis Kurt(a), and the average value of the hand acceleration of the patient's hand movement and the kurtosis Kurt(a) are combined with the change of the hand contour to obtain the influence of the complexity of the patient's hand movement on the hand tremor.

[0049] In this embodiment, the depth camera is used to collect the image information of the patient's hand contour and the image information when it contacts the object in real time. When it is detected that the hand contacts the object, the monitoring window is immediately opened to accurately capture the instantaneous change of the hand movement. The processing module compares and analyzes the continuously collected images, calculates the dynamic change of the hand contour, so as to quantify the complexity of the hand movement. At the same time, combined with the average value of the hand acceleration and the kurtosis obtained by the three-axis acceleration sensor, the stability and suddenness of the hand movement are comprehensively evaluated. By combining the acceleration characteristics of the hand movement with the change of the contour, the influence of the complexity of the hand movement on the hand tremor is deeply analyzed, providing a more comprehensive and accurate quantitative index for the condition monitoring of Parkinson's disease patients, helping doctors to more accurately evaluate the disease progression, formulate personalized treatment plans, and improve the quality of life of patients.

[0050] Among them, the processing module is further configured to obtain the first upper limb model and the corresponding tremor frequency, and establish a tremor prediction model based on the first upper limb model and the corresponding tremor frequency; the calculation module puts the obtained first upper limb model into the tremor prediction model to predict the tremor frequency as the first frequency, and then obtains the tremor frequency associated with the first upper limb model as the second frequency. If the difference between the second frequency and the first frequency is greater than a preset difference (generally 95%, which can be specifically adjusted by medical staff according to the actual situation of the patient), the extraction time of the human motion timing feature is extended; if the difference between the first frequency and the second frequency obtained after the extension time is greater than the preset difference, an abnormal signal is generated and sent to the evaluation module; the evaluation module receives and plays or displays the abnormal signal.

[0051] Among them, the evaluation module is further configured to, when playing or displaying the abnormal signal, play or display a preset paradigm behavior to the patient (such as: static tremor action paradigm, postural tremor action paradigm, kinetic tremor action paradigm, etc. When selecting, a paradigm behavior with a high matching similarity can be selected according to the first upper limb model of the patient), and generate a signal for increasing acquisition and send it to the acquisition module; after receiving the signal for increasing acquisition, the acquisition module increases the acquisition frequency of the camera; the calculation module compares the obtained first upper limb model with the preset paradigm behavior. If the comparison result is greater than the preset difference, an adjustment prompt is generated and sent to the evaluation module, and the evaluation module is used to play or display the adjustment prompt; if the comparison result is not greater than the preset difference (generally 95%, which can be specifically adjusted by medical staff according to the actual situation of the patient), it is judged whether the difference between the difference between the first frequency and the second frequency and the first difference value is greater than the preset difference. If the difference between the difference value and the first difference value is not greater than the preset difference, a condition change prompt is generated, and the first frequency and the change prompt are sent to the evaluation module, and the evaluation module receives and plays or displays the first frequency and the change prompt; if the difference between the difference value and the first difference value is greater than the preset difference, the first upper limb model and the first frequency are used to train the tremor prediction model.

[0052] Among them, when generating a condition change prompt and sending it to the evaluation module, the calculation module also associates and stores the first frequency with the first upper limb model as a change behavior. The calculation module regularly statistically outputs the change behavior according to the distribution of the first upper limb model in the change behavior, and generates a change report according to the statistical distribution result and sends it to the evaluation module, and the evaluation module receives and plays or displays the change report.

[0053] Specifically in implementation, Mr. Zhang is taken as a Parkinson's patient, and Mr. Zhang uses a wearable bracelet. The wearable bracelet is provided with a three-axis acceleration sensor and a depth camera. The three-axis acceleration sensor on the wearable bracelet works continuously and real-time collects the linear accelerations of Mr. Zhang's hand in three orthogonal directions.

[0054] At a certain moment, the sensor recorded different acceleration values of Mr. Zhang's hand in the horizontal and vertical directions, and this data was transmitted to the processing module as the first information. The depth camera equipped on the bracelet was also synchronously collecting the image information around Mr. Zhang's hand. When Mr. Zhang picked up the water cup to drink water, the camera captured the picture of the hand contacting the water cup and the posture changes of the hand during the movement process.

[0055] After the processing module obtains the first information, based on the acceleration in the horizontal direction and the acquisition time, it calculates the movement speed of Mr. Zhang's hand in the horizontal direction as the first data; similarly, according to the acceleration in the vertical direction and the acquisition time, it obtains the movement speed in the vertical direction, that is, the second data. The calculation module uses the first data and the second data, combined with the human upper limb model, to calculate the movement trajectory of Mr. Zhang's hand and further obtains the first upper limb model.

[0056] The calculation module calculates the tremor frequency of Mr. Zhang's hand according to the change frequency of the first data and the second data, and correlates it with the human body movements identified by the deep learning model. For example, when Mr. Zhang is in a static state, the tremor frequency at this time is calculated to be 5Hz.

[0057] The processing module obtains the acquisition time corresponding to the human body movement, and correlates the acquisition time with the tremor frequency. If Mr. Zhang remains static for a period of time and his hand has tremors, this period of time is used as the duration, and the temporal change characteristics of the tremor frequency during this duration are extracted. Combining the preset frequency threshold (assuming that the medical staff set it to 4Hz according to Mr. Zhang's situation), it is judged that the tremor is in a mild severity at this time.

[0058] The processing module processes the image information collected by the depth camera, identifies and extracts the hand contour, and identifies the object contour in contact with the hand contour. When it is recognized that Mr. Zhang's hand contacts the water cup, the tremor frequency at this time is obtained, and at the same time, the calculation module judges that the complexity of the hand movement is medium according to the change speed of the hand contour. Then, according to the complexity of the hand movement and the tremor frequency, (combining the judgment result of the frequency threshold), it is matched that Mr. Zhang's current disease state is in the medium stage.

[0059] The processing module also identifies and labels the human contour features in the image information as other body features, such as arms, shoulders, etc., and combines these other body features with the hand position to establish a relative model, and then integrates it into the first upper limb model to make the first upper limb model more accurate and comprehensive.

[0060] The processing module establishes a tremor prediction model according to the first upper limb model and the corresponding tremor frequency. The calculation module puts the obtained first upper limb model into the tremor prediction model to predict the tremor frequency as the first frequency; at the same time, it obtains the actual tremor frequency associated with the first upper limb model as the second frequency.

[0061] Assume that the preset difference is 90% (adjusted by medical staff according to Mr. Zhang's condition). When the calculated difference between the second frequency and the first frequency is greater than this preset difference, the system extends the extraction time of the human motion timing characteristics. If the difference is still greater than the preset difference after the extension, the calculation module generates an abnormal signal and sends it to the evaluation module.

[0062] After receiving the abnormal signal, the evaluation module displays the abnormal signal and shows Mr. Zhang the preset paradigm behavior, such as the static tremor action paradigm. At the same time, the evaluation module generates a signal for increasing collection and sends it to the collection module. After receiving the signal, the collection module increases the collection frequency of the depth camera.

[0063] The calculation module compares the obtained first upper limb model with the preset paradigm behavior. If the comparison result is greater than the preset difference (assumed to be 90%), the calculation module generates an adjustment prompt and sends it to the evaluation module, and the evaluation module displays the adjustment prompt to remind Mr. Zhang to adjust his actions.

[0064] If the comparison result is not greater than the preset difference, the calculation module further determines whether the difference between the difference between the first frequency and the second frequency and the first difference value is greater than the preset difference. If not, it generates a condition change prompt and sends the first frequency and the change prompt to the evaluation module, and the evaluation module displays the first frequency and the change prompt; if it is greater, the first upper limb model and the first frequency are used to train the tremor prediction model to optimize the prediction accuracy of the model.

[0065] When the calculation module generates a condition change prompt and sends it to the evaluation module, it associates and stores the first frequency with the first upper limb model as a change behavior. The calculation module regularly (such as every day) performs statistical output according to the distribution of the first upper limb model in the change behavior, and generates a change report according to the statistical distribution result and sends it to the evaluation module. The evaluation module displays the change report, and the doctor can timely understand Mr. Zhang's condition change according to the report and adjust the treatment plan.

[0066] The above are only embodiments of the present invention, and common knowledge such as specific structures and characteristics known in the solution is not described in detail here. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can still be made, and these should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicality of the patent. The protection scope required by this application should be subject to the content of its claims, and the specific implementation manners and the like recorded in the specification can be used to interpret the content of the claims.

Claims

1. Parkinson's patient motion analysis system based on a wearable device, characterized in that, Comprising: A collection module, including a wearable bracelet, on which a triaxial acceleration sensor is provided, and the triaxial acceleration sensor is used to collect the linear accelerations in three orthogonal directions of the patient's hand as the first information; A processing module, which obtains the first information, calculates the movement speed of the patient's hand in the horizontal direction as the first data according to the acceleration in the horizontal direction and the collection time in the first information, and calculates the movement speed of the patient's hand in the vertical direction as the second data according to the acceleration in the vertical direction and the collection time in the first information; A calculation module, including a human upper limb model, which is used to calculate the movement trajectory of the patient's hand according to the first data and the second data, and is used to calculate the first upper limb model by combining the hand movement trajectory with the human upper limb model; in the calculation module, there are also deep learning models for various human upper limb movement trainings, and the calculation module is used to put the first upper limb model into the deep learning model to obtain human actions; the calculation module is also used to calculate the tremor frequency according to the change frequencies of the first data and the second data, and associate the tremor frequency with the human actions; An evaluation module, including a scale established based on human actions, tremor frequency and disease severity, which is used to obtain the tremor frequency and associated human actions, and match the disease severity according to the tremor frequency and associated human actions.

2. The Parkinson's patient motion analysis system based on a wearable device according to claim 1, wherein: The processing module obtains the collection time of the first information, and obtains the temporal characteristics between the first data and the second data and the human actions, including but not limited to the mean value, variance, skewness, kurtosis of the acceleration signal, and the cross-correlation coefficient of the triaxial acceleration; Combined with a preset temporal model, the movement state of the patient's hand is monitored and analyzed in real time.

3. The Parkinson's patient movement analysis system based on a wearable device according to claim 2, characterized in that: The processing module obtains the collection time corresponding to the human action, associates the collection time with the tremor frequency, takes the time with adjacent collection times and the same human action as the duration, obtains the tremor frequency associated with the duration, extracts the temporal change characteristics of the tremor frequency, and according to the temporal change characteristics, combined with a preset frequency threshold, judges the severity of the tremor, and conducts an association analysis on the temporal characteristics of the tremor frequency and the human actions.

4. The Parkinson's patient motion analysis system based on a wearable device according to claim 3, wherein: The collection module further includes a camera for collecting image information around the hand, the processing module is used to obtain the image information, identify and extract the hand contour in the image information, identify the object contour in contact with the hand contour, and then obtain the tremor frequency when the object contour in contact with the hand contour is in contact, and the calculation module is used to obtain the complexity of the hand action according to the change speed of the hand contour, and match the disease severity according to the complexity of the hand action and the tremor frequency.

5. The Parkinson's patient motion analysis system based on a wearable device according to claim 4, wherein: The camera is a depth camera, and the processing module is further used to identify and label the human contour features in the image information as other body features, establish a relative model of the other body features in combination with the hand position, and fuse the relative model into the first upper limb model according to the relative position between the relative model and the hand.

6. The Parkinson's patient motion analysis system based on a wearable device according to claim 5, characterized in that: The processing module is further configured to obtain the first upper limb model and the corresponding tremor frequency, and establish a tremor prediction model based on the first upper limb model and the corresponding tremor frequency; the calculation module puts the obtained first upper limb model into the tremor prediction model to predict the tremor frequency as the first frequency, and then obtains the tremor frequency associated with the first upper limb model as the second frequency. If the difference between the second frequency and the first frequency is greater than a preset difference, the extraction time of the human motion timing feature is extended; If the difference between the first frequency and the second frequency after the extension time is greater than the preset difference, an abnormal signal is generated and sent to the evaluation module; the evaluation module receives and plays or displays the abnormal signal.

7. The Parkinson's patient motion analysis system based on a wearable device according to claim 6, characterized in that: The evaluation module is further configured to play or display a preset paradigm behavior to the patient and generate an increased acquisition signal and send it to the acquisition module when playing or displaying the abnormal signal; after receiving the increased acquisition signal, the acquisition module increases the acquisition frequency of the camera; The calculation module compares the obtained first upper limb model with the preset paradigm behavior. If the comparison result is greater than the preset difference, an adjustment prompt is generated and sent to the evaluation module, and the evaluation module is configured to play or display the adjustment prompt; if the comparison result is not greater than the preset difference, it is determined whether the difference between the difference between the first frequency and the second frequency and the first difference value is greater than the preset difference. If the difference between the difference value and the first difference value is not greater than the preset difference, a disease condition change prompt is generated, and the first frequency and the change prompt are sent to the evaluation module, and the evaluation module receives and plays or displays the first frequency and the change prompt; if the difference between the difference value and the first difference value is greater than the preset difference, the first upper limb model and the first frequency are used to train the tremor prediction model.

8. The Parkinson's patient motion analysis system based on a wearable device according to claim 7, wherein: The calculation module is further configured to, when generating a disease condition change prompt and sending it to the evaluation module, associate and store the first frequency with the first upper limb model as a change behavior. The calculation module regularly outputs the change behavior according to the distribution statistics of the first upper limb model in the change behavior, and generates a change report according to the statistical distribution result and sends it to the evaluation module, and the evaluation module receives and plays or displays the change report.

9. A method for analyzing the movements of Parkinson's patients based on wearable devices, characterized in that, The Parkinson's patient motion analysis system based on a wearable device according to claims 1-8 is used.

Images