A method of gait assessment for parkinson's patients and related devices

Abstract

The application belongs to the technical field of gait assessment, and discloses a Parkinson's disease patient gait assessment method and related equipment, the method comprising: acquiring radar point cloud and micro-Doppler time-frequency spectrum of a user walking in a test area by using a millimeter wave radar, and preprocessing the radar point cloud; extracting human point cloud clusters from each frame of preprocessed radar point cloud by a clustering method, and identifying the mass center position trajectory of the left and right feet in combination with Kalman filtering and the Hungarian algorithm; calculating the speed of the left and right feet according to the mass center position trajectory of the left and right feet, and detecting the step event of the left and right feet in combination with the micro-Doppler time-frequency spectrum; calculating gait assessment parameters according to the time sequence of the step event and the mass center point position of the human point cloud cluster; comparing the gait assessment parameters with prestored reference data to generate a Parkinson's disease condition assessment result of the user; thereby providing continuous and convenient rehabilitation monitoring for Parkinson's disease patients, and significantly improving the accuracy and practicality of Parkinson's disease condition assessment.

Description

Technical Field

[0001] This application relates to the field of gait assessment technology, and more specifically, to a gait assessment method and related equipment for Parkinson's patients. Background Technology

[0002] Parkinson's disease is a common neurodegenerative disease. Its motor symptoms, such as tremor and bradykinesia, are subtle and variable, making it difficult to fully and accurately represent the true state of the disease during clinical consultations. Gait assessment is an extremely important component of the Parkinson's disease assessment system, providing crucial reference for judging the patient's rehabilitation status.

[0003] Traditional gait analysis systems primarily rely on 3D motion capture systems, force measurement platforms, and electromyography (EMG) equipment in hospital laboratories. These devices are not only expensive and complex to operate, but also require professional personnel for operation and maintenance, making them difficult to widely adopt in everyday home environments. This limits the ability of Parkinson's patients to conduct continuous and convenient home rehabilitation monitoring.

[0004] In recent years, the development of wearable sensing technology has provided new solutions for gait monitoring. However, contact sensors, such as the gait analysis device based on an inertial sensor disclosed in Chinese patent CN120345891A, whose contact sensors are placed on the soles of the feet, can affect the patient's natural gait to some extent, thus affecting the accuracy of gait assessment. This contact monitoring method may cause changes in the patient's gait while wearing the sensor, failing to truly reflect their daily gait characteristics, thereby reducing the reliability of the assessment results. Therefore, existing technologies for gait assessment in Parkinson's patients still face problems such as expensive equipment, complex operation, difficulty in home application, and the impact of contact sensors on assessment accuracy.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] The purpose of this application is to provide a gait assessment method and related equipment for Parkinson's disease patients, aiming to solve the problems of expensive equipment, complicated operation, difficulty in home application, and the impact of contact sensors on the accuracy of assessment in the existing technology for gait assessment of Parkinson's disease patients.

[0007] In one aspect, this application provides a gait assessment method for Parkinson's patients, the method comprising the following steps:

[0008] A1. Use millimeter-wave radar to acquire radar echo signals when the user walks in the test area, and generate radar point cloud and micro-Doppler time spectrum;

[0009] A2. Perform coordinate transformation and normalization on the radar point cloud to obtain the preprocessed radar point cloud;

[0010] A3. Using clustering methods, human point cloud clusters are extracted from each preprocessed radar point cloud frame, and combined with Kalman filtering and Hungarian algorithm, the centroid position trajectory of the left and right feet is identified.

[0011] A4. Calculate the velocities of the left and right feet based on the trajectories of their center of mass positions, and detect the step events of the left and right feet by combining the micro-Doppler time spectrum.

[0012] A5. Calculate gait evaluation parameters based on the time sequence of the step events and the centroid position of the human point cloud cluster; the gait evaluation parameters include at least one of step frequency, step frequency mutation parameter, stride length, walking speed, and gait symmetry index;

[0013] A6. Compare the calculated gait assessment parameters with the pre-stored reference data to generate the user's Parkinson's disease assessment results.

[0014] Secondly, this application provides a gait assessment system for Parkinson's patients, which includes millimeter-wave radar and a processing device;

[0015] The millimeter-wave radar is used to transmit radar beams to the test area and receive radar echo signals, which are then uploaded to the processing device.

[0016] The processing device is used to acquire radar echo signals from the user walking within the test area using the millimeter-wave radar, generating radar point clouds and micro-Doppler time-frequency spectra; performing coordinate transformation and normalization on the radar point clouds to obtain pre-processed radar point clouds; extracting human point cloud clusters from each frame of the pre-processed radar point cloud using a clustering method, and identifying the centroid positions and trajectories of the left and right feet using Kalman filtering and the Hungarian algorithm; calculating the speeds of the left and right feet based on their centroid positions and trajectories, and detecting gait events based on the micro-Doppler time-frequency spectra; calculating gait evaluation parameters based on the time series of the gait events and the centroid positions of the human point cloud clusters; the gait evaluation parameters include at least one of gait frequency, gait frequency mutation parameters, stride length, walking speed, and gait symmetry index; comparing the calculated gait evaluation parameters with pre-stored reference data to generate the user's Parkinson's disease assessment results.

[0017] Thirdly, this application provides an electronic device including a processor and a memory, the memory storing a computer program executable by the processor, wherein when the processor executes the computer program, it performs the steps in the Parkinson's disease patient gait assessment method described above.

[0018] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the Parkinson's disease patient gait assessment method described above.

[0019] Beneficial Effects: This application provides a gait assessment method and related equipment for Parkinson's disease patients. By employing non-contact millimeter-wave radar technology, it offers a low-cost, easy-to-operate, and home-environment-suitable gait assessment method. This provides continuous and convenient rehabilitation monitoring for Parkinson's patients, significantly improving the accuracy and practicality of Parkinson's disease assessment. It effectively solves the problems of expensive equipment, complex operation, difficulty in home application, and the impact of contact sensors on assessment accuracy in existing gait assessment methods for Parkinson's patients. Attached Figure Description

[0020] Figure 1 A flowchart of a gait assessment method for Parkinson's patients provided in this application.

[0021] Figure 2 A schematic diagram of a gait assessment system for Parkinson's patients provided in this application.

[0022] Figure 3 A schematic diagram of the structure of the electronic device provided in this application.

[0023] Labeling explanations: 1. Millimeter-wave radar; 2. Processing device; 301. Processor; 302. Memory; 303. Communication bus. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0025] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0026] Please refer to Figure 1A gait assessment method for Parkinson's patients, as described in some embodiments of this application, includes the following steps:

[0027] A1. Use millimeter-wave radar to acquire radar echo signals when the user walks in the test area, and generate radar point cloud and micro-Doppler time spectrum;

[0028] A2. Perform coordinate transformation and normalization on the radar point cloud to obtain the preprocessed radar point cloud;

[0029] A3. Using clustering methods, human point cloud clusters are extracted from each preprocessed radar point cloud frame, and combined with Kalman filtering and Hungarian algorithm, the centroid position trajectory of the left and right feet is identified.

[0030] A4. Calculate the velocities of the left and right feet based on the trajectories of their center of mass positions, and detect step events of the left and right feet by combining the micro-Doppler time spectrum.

[0031] A5. Calculate gait evaluation parameters based on the time series of step events and the centroid position of the human point cloud cluster; the gait evaluation parameters include at least one of step frequency, step frequency mutation parameter, stride length, walking speed, and gait symmetry index;

[0032] A6. Compare the calculated gait assessment parameters with the pre-stored reference data to generate the user's Parkinson's disease assessment results.

[0033] This application overcomes the interference of traditional contact sensors on gait naturalness by using non-contact millimeter-wave radar technology, while avoiding the expense and complexity of large medical equipment. This enables gait assessment of Parkinson's patients to be carried out in the daily home environment, thereby providing patients with continuous and convenient rehabilitation monitoring.

[0034] Among them, millimeter-wave radar is a radar system that uses electromagnetic waves in the millimeter-wave band for detection. It has high resolution, the ability to penetrate non-metallic materials, and is not easily affected by ambient light, making it suitable for non-contact monitoring of human movement.

[0035] Radar echo signals are signals reflected back from a target (such as a human body) after the electromagnetic waves emitted by millimeter-wave radar encounter the target. They carry information such as the target's distance, speed, and angle.

[0036] Radar point cloud is a three-dimensional set of points generated by processing radar echo signals. Each point represents a reflection point on the target surface and contains the spatial coordinates and velocity information of that point.

[0037] Micro-Doppler time spectrum is a spectrum obtained by time-frequency analysis of radar echo signals. It can reflect the frequency changes caused by minute movements of the target (such as limb swings) and is often used to identify and analyze fine movements in human gait.

[0038] Human body point cloud clusters are sets of points belonging to the human body identified by clustering algorithms in radar point clouds, representing the user's body outline at a certain moment.

[0039] The centroid position trajectory refers to the continuous path of the geometric center of the human body point cloud cluster or foot point cloud cluster as time changes, and is used to describe the movement trend of the human body or foot.

[0040] Gait events refer to key events in the gait cycle, such as heel strike and toe lift-off, and are important time points for gait analysis.

[0041] Gait assessment parameters are indicators used to quantify and describe gait characteristics, such as cadence, stride length (distance of each step), walking speed, and gait symmetry. These parameters are of great significance for assessing the motor function of Parkinson's patients.

[0042] This application provides a gait assessment method for Parkinson's disease patients, the core of which is to utilize the non-contact sensing capability of millimeter-wave radar, combined with advanced signal processing and data analysis algorithms, to achieve accurate assessment of the gait of Parkinson's disease patients.

[0043] In step A1, millimeter-wave radar is used to acquire radar echo signals as the user walks within the test area, generating radar point clouds and micro-Doppler time spectra. For example, the millimeter-wave radar can be placed on one side of the test area. When the user walks within this area, the radar continuously emits millimeter-wave beams and receives the reflected signals. These raw radar echo signals are then sent to a signal processing module. Through a series of digital signal processing techniques, radar point cloud data can be extracted. This point cloud data contains real-time three-dimensional position information of various parts of the user's body. Simultaneously, time-frequency analysis, such as short-time Fourier transform, is performed on the radar echo signals to generate micro-Doppler time spectra. This spectrogram can capture the Doppler frequency changes caused by minute movements of the user's limbs (such as foot swings), providing detailed motion features for subsequent step event detection.

[0044] In step A2, the radar point cloud undergoes coordinate transformation and normalization to obtain a preprocessed radar point cloud. For example, the original radar point cloud's coordinate system typically uses the radar itself as the origin. To facilitate subsequent human recognition and gait parameter calculation, the coordinates of the point cloud data can be uniformly transformed to a ground coordinate system, making the ground the XY plane and the height the Z-axis. Furthermore, the point cloud data can be normalized; for example, the intensity or distance values ​​of the point cloud can be adjusted to a uniform range to eliminate differences caused by different radars or environmental factors, ensuring data consistency and comparability.

[0045] In step A3, human point cloud clusters are extracted from each preprocessed radar point cloud frame using clustering methods. Kalman filtering and the Hungarian algorithm are then combined to identify the centroid positions of the left and right feet. Specifically, clustering algorithms such as DBSCAN and K-means can be used to process each preprocessed radar point cloud frame, clustering points belonging to the human body into human point cloud clusters. For example, densely connected points can be identified as a cluster based on the density and spatial distribution characteristics of the point cloud. Once the human point cloud cluster is identified, its centroid position can be calculated, representing the overall position of the human body. To further identify the left and right feet, local sub-clustering can be performed within the human point cloud clusters, selecting sub-clusters that match foot characteristics as candidate foot point cloud clusters. Then, based on the relative positional relationship between these candidate foot point cloud clusters and the centroid positions of the human point cloud clusters, they are distinguished into left foot point cloud clusters and right foot point cloud clusters. Finally, Kalman filtering is used to smooth and predict the centroid positions of the left and right feet to reduce the impact of measurement noise. The Hungarian algorithm is then used to correlate the data, thereby identifying the centroid positions and trajectories of the left and right feet and ensuring the continuity and accuracy of the trajectory.

[0046] In step A4, the velocities of the left and right feet are calculated based on their centroid positions and trajectories, and then combined with the micro-Doppler time spectrum to detect step events. For example, the movement velocities of the left and right feet at various times can be calculated by performing a difference operation on their centroid positions. To more accurately detect step events, the velocity components of the left and right feet's movement velocities in the user's forward direction can be further calculated to obtain the forward speed of the left and right feet. When the forward speed of the left or right foot approaches zero (e.g., less than a preset threshold), a landing event can be preliminarily determined, and the time of the event is recorded. To verify whether the landing event is a genuine step event, the micro-Doppler time spectrum can be used for auxiliary judgment. By calculating the correlation between the peak time series of the reference micro-Doppler time spectrum and the peak time series of the control velocity data, and comparing it with a preset threshold, it can be determined whether the landing event is a step event.

[0047] In step A5, gait evaluation parameters are calculated based on the time series of step events and the centroid positions of the human point cloud clusters. These gait evaluation parameters include at least one of gait frequency, gait frequency mutation parameter, stride length, walking speed, and gait symmetry index. For example, gait frequency can be obtained by calculating the average step period from the time series of step events and then taking its reciprocal. Walking speed can be obtained by calculating the average speed of the centroids of the human point cloud clusters in each frame. Stride length can be obtained by dividing walking speed by gait frequency. The gait frequency mutation parameter can be obtained by analyzing the coefficient of variation of the step period in the time series of step events; when the coefficient of variation exceeds a preset threshold, a gait frequency mutation event is determined to have occurred. The gait symmetry index can be obtained by calculating the average stride length difference between the left and right feet.

[0048] In step A6, the calculated gait assessment parameters are compared with pre-stored reference data to generate the user's Parkinson's disease assessment result. For example, the calculated gait frequency, stride length, walking speed, gait symmetry index, and other parameters can be compared with reference data from healthy individuals or Parkinson's patients with different disease severities. This reference data can be normal ranges, abnormal thresholds, or typical parameter values ​​for different disease stages, obtained in advance through statistical analysis of a large amount of clinical data. By comparing these parameters, it is possible to assess whether the user's current gait characteristics deviate from the normal range and the degree of deviation, thereby generating the user's Parkinson's disease assessment result. For example, it can determine the severity of the user's Parkinson's disease or provide specific gait abnormality alerts.

[0049] The gait assessment method for Parkinson's disease patients presented in this application effectively solves the problem of interference with the patient's natural gait caused by traditional contact sensors through non-contact millimeter-wave radar technology, thereby improving the accuracy and reliability of the assessment. Compared with traditional large medical devices, this method has the advantages of low cost, simple operation, and easy home deployment, enabling Parkinson's disease patients to conduct continuous and convenient gait monitoring at home.

[0050] Specifically, the core innovation of this application lies in:

[0051] First, by utilizing millimeter-wave radar to acquire radar point clouds and micro-Doppler time-spectrum data, non-contact, high-precision perception of the user's gait is achieved. The radar point cloud provides precise three-dimensional positional information for various parts of the user's body, while the micro-Doppler time-spectrum captures the fine features of minute limb movements. The combination of these two provides a rich and comprehensive data foundation for subsequent gait analysis. Compared to traditional gait analysis devices based on inertial sensors, this avoids the potential impact of sensor wearing on the patient's gait, ensuring the accuracy of the assessment results.

[0052] Secondly, this application achieves accurate identification of the centroid positions and trajectories of the left and right feet and reliable detection of gait events through refined data processing and algorithms. Human body point cloud clusters and foot point cloud clusters are extracted using clustering methods, and Kalman filtering and the Hungarian algorithm are combined to smooth and correlate the foot trajectories, effectively overcoming noise and uncertainty in radar data. Furthermore, combining left and right foot velocities with micro-Doppler time-spectrum data to verify gait events further improves the accuracy of gait event detection, laying the foundation for accurate calculation of subsequent gait parameters. This approach offers higher robustness and accuracy compared to methods that rely solely on a single data source (such as a velocity threshold) to determine gait events.

[0053] Finally, this application can calculate multiple key gait assessment parameters and compare them with pre-stored reference data to generate a user's Parkinson's disease assessment results. These parameters include gait frequency, gait frequency mutation parameters, stride length, walking speed, gait symmetry indices, etc., comprehensively reflecting the gait characteristics of Parkinson's patients. By comparing with reference data, the severity of the patient's condition can be objectively and quantitatively assessed, providing important reference for clinical diagnosis and rehabilitation treatment. This comprehensive assessment method, compared with assessment methods that only focus on a single or a few gait parameters, can more comprehensively and accurately reflect the patient's true condition.

[0054] In summary, the gait assessment method for Parkinson's patients proposed in this application has made significant progress in terms of non-contact, high precision, and ease of home application by introducing millimeter-wave radar technology and advanced data processing algorithms, providing effective technical support for the long-term monitoring and rehabilitation management of Parkinson's patients.

[0055] Preferably, step A1 includes:

[0056] A101. Use millimeter-wave radar to acquire radar echo signals when a user walks toward or away from the millimeter-wave radar in the test area.

[0057] A102. Generate multi-frame radar point clouds based on radar echo signals;

[0058] A103. Perform a short-time Fourier transform on the radar echo signal to obtain the micro-Doppler time spectrum.

[0059] Step A101 aims to ensure the effectiveness and consistency of radar signal acquisition. By guiding the user to walk towards or away from the millimeter-wave radar within the test area (i.e., walking in the direction of the radar beam or in the direction of the radar beam), the Doppler effect related to gait can be captured to the maximum extent, thereby obtaining high-quality radar echo signals.

[0060] Further, in step A102, the acquired radar echo signal is processed to generate a multi-frame radar point cloud. A radar point cloud is a dataset composed of a series of discrete points, each representing a reflector detected by the radar beam in space. These points typically contain information such as range, angle, and velocity. By processing continuous radar echo signals, multi-frame point cloud data that varies over time can be generated, thereby reflecting the user's dynamic position and shape within the test area.

[0061] Furthermore, in step A103, a short-time Fourier transform (STFT) is performed on the radar echo signal to obtain the micro-Doppler time spectrum. The micro-Doppler effect refers to the minute changes in the frequency of the radar echo signal caused by the internal motion or vibration of the target. For human gait, the swinging of limbs such as the feet and legs produces unique micro-Doppler characteristics. The short-time Fourier transform is a time-frequency analysis method that can convert a time-domain signal into a time-frequency domain representation, thus clearly showing the changes in the micro-Doppler information contained in the radar echo signal over time, forming the micro-Doppler time spectrum.

[0062] The solution in this application refines step A1 into A101, A102, and A103, achieving systematic acquisition and preliminary processing of the raw radar echo signal. First, by defining the user's walking direction (facing or away from the millimeter-wave radar), it ensures that the radar echo signal can effectively capture key information related to gait, providing high-quality raw data for subsequent analysis and facilitating the differentiation of left and right foot point cloud clusters among candidate foot point clouds. Second, converting the radar echo signal into multi-frame radar point clouds allows for a visual representation of the user's spatial position and movement trajectory within the test area, laying the foundation for subsequent human body recognition and foot centroid trajectory extraction. Finally, by performing a short-time Fourier transform on the radar echo signal to generate a micro-Doppler time spectrum, fine features of limb movement can be captured from the frequency dimension. This is crucial for subsequent gait event detection, as the micro-Doppler time spectrum reflects the unique frequency variation patterns when the foot contacts the ground.

[0063] The above technical solution clarifies and refines the steps for millimeter-wave radar data acquisition and preliminary processing. This refinement not only improves the standardization and effectiveness of data acquisition, ensuring that the acquired radar echo signals fully reflect the user's gait characteristics, but also provides two complementary and crucial data sources for subsequent gait parameter calculations by generating radar point clouds and micro-Doppler time spectra separately. Radar point clouds provide spatial location information, while micro-Doppler time spectra provide detailed motion frequency information; the combination of the two significantly improves the accuracy and reliability of gait assessment.

[0064] In some implementations, step A3 includes:

[0065] A301. Using a clustering method, human point cloud clusters are extracted from each preprocessed radar point cloud frame, and the centroid position of each human point cloud cluster is obtained.

[0066] A302. Perform local sub-clustering in each human body point cloud cluster, select sub-clusters that meet the preset height range and point cloud quantity threshold as candidate foot point cloud clusters, and obtain the centroid position of the candidate foot point cloud clusters.

[0067] A303. Based on the relative positional relationship between the centroids of the candidate foot point cloud clusters and the human body point cloud clusters, the candidate foot point cloud clusters are divided into left foot point cloud clusters and right foot point cloud clusters.

[0068] A304. Based on the centroid positions of the left foot point cloud cluster and the right foot point cloud cluster in each frame, Kalman filtering and Hungarian algorithm are used to identify the centroid positions and trajectories of the left and right feet.

[0069] In step A301, the clustering method can employ algorithms such as DBSCAN, K-means, or Mean-shift. The purpose is to group points belonging to the same target in the radar point cloud, thereby separating the human body from background noise or other clutter. The centroid of the human body point cloud cluster is typically obtained by calculating the average coordinates of all points within the cluster; this centroid can serve as an approximate representation of the human body's position in the current frame.

[0070] Specifically, in step A302, since the feet are typically located at a lower position in the human body point cloud cluster and possess a certain point cloud quantity characteristic, potential foot regions can be identified by setting a height range (e.g., 0.05 meters to 0.3 meters above the ground) and a point cloud quantity threshold. That is, for the identified sub-clusters, only those sub-clusters whose height is within the preset height range and whose number of points is not less than the point cloud quantity threshold are considered candidate foot point cloud clusters. Through local sub-clustering, the foot point cloud can be separated from the legs or other parts of the body more precisely, and the centroid positions of these candidate foot point cloud clusters can be calculated.

[0071] In step A303, the centroid of the human body point cloud cluster can be used as a reference to determine whether the candidate foot point cloud cluster is located to its left or right. When the user walks towards or away from the millimeter-wave radar, the left and right feet can be effectively distinguished by comparing the lateral coordinates of the candidate foot point cloud cluster with the lateral coordinates of the centroid of the human body point cloud cluster.

[0072] Specifically, in step A304, Kalman filtering is used to smooth and predict the centroid positions of the left and right feet to reduce the impact of radar measurement noise and handle situations with missing data or temporary occlusion, thereby obtaining a more stable and continuous trajectory. The Hungarian algorithm is used to correlate data between consecutive frames, ensuring that the left and right foot point cloud clusters identified in the current frame can correctly match the left and right foot trajectories of the previous frame, avoiding confusion of left and right foot identities due to instantaneous changes in the point cloud clusters.

[0073] The proposed solution refines step A3 into sub-steps A301 to A304, achieving accurate identification of the left and right foot centroid positions in gait assessment for Parkinson's patients. First, preliminary clustering of the radar point cloud (A301) effectively separates the human target from the complex environment. Based on this, local sub-clustering combined with height and point cloud quantity thresholds (A302) accurately extracts the foot region from the human point cloud, avoiding misidentification of other body parts as feet. Further, by analyzing the relative position between the foot point cloud clusters and the human centroid point (A303), the left and right feet can be reliably distinguished, which is crucial for subsequent gait symmetry analysis. Finally, Kalman filtering is used to smooth and predict the centroid position, effectively suppressing radar measurement noise and ensuring trajectory continuity; while the Hungarian algorithm ensures correct data association during multi-target (left and right feet) tracking, avoiding trajectory overlap or identity confusion, thus jointly guaranteeing the accuracy and robustness of left and right foot centroid position trajectory identification.

[0074] The above technical solution significantly improves the accuracy and stability of extracting human point cloud clusters from radar point cloud data and identifying the centroid positions and trajectories of the left and right feet. This detailed foot point cloud recognition and trajectory tracking method effectively overcomes the problems of traditional methods, such as susceptibility to noise interference in complex environments, difficulty in distinguishing between left and right feet, and trajectory jumps, laying a solid foundation for the accurate calculation of subsequent gait assessment parameters. Therefore, this application can provide more reliable and detailed Parkinson's disease assessment results, helping doctors to accurately diagnose patients' conditions and formulate treatment plans.

[0075] In practical applications, simply combining speed and micro-Doppler time spectrum may not be sufficient to accurately distinguish between true gait events and non-gait events, such as brief pauses, body swaying, or foot slips, thus affecting the accuracy and reliability of gait assessment. Failure to address this issue may lead to misjudgment of gait events, consequently affecting the accuracy of subsequent gait assessment parameter calculations and ultimately the accuracy of Parkinson's disease assessment results. Therefore, this application further proposes a refined and optimized scheme for the above-mentioned steps in detecting gait events to improve the accuracy and reliability of gait event detection. Specifically, step A4 includes:

[0076] A401. Calculate the moving speed of the left and right feet based on the trajectory of the center of mass of the left and right feet;

[0077] A402. Calculate the velocity components of the left and right feet in the user's forward direction to obtain the forward velocity of the left and right feet;

[0078] A403. When the forward speed of the left or right foot is zero, a landing event is determined to have occurred, and the time of the landing event is recorded;

[0079] A404. Use micro-Doppler time spectrum to verify whether the landing event is a step event;

[0080] A405. If the landing event is a step event, then add the corresponding landing event occurrence time to the time series of the step event.

[0081] In step A401, the trajectories of the centroid positions of the left and right feet can be obtained from step A3 above. Based on these trajectory data, the instantaneous movement velocities of the left and right feet at various moments can be calculated using methods such as differential or filtering. Specifically, the movement velocity can be understood as the displacement of the centroid point per unit time in three-dimensional space.

[0082] Furthermore, in step A402, to more accurately reflect the actual state of the user's gait, the movement speed of the left and right feet needs to be decomposed into the user's forward direction. The user's forward direction can be determined based on the overall movement direction of the human body point cloud cluster or the direction of a preset test area. By calculating the projection of the velocity vector onto the forward direction, the forward speed of the left and right feet can be obtained.

[0083] In step A403, when the forward speed of the left or right foot approaches zero, it usually means that the foot is in contact with the ground and is in a relatively stationary state, i.e., a landing event has occurred. To avoid noise interference, "zero speed" can be defined as a speed less than a preset minimum threshold. Once a landing event is determined to have occurred, the time of its occurrence will be accurately recorded.

[0084] As a preferred embodiment, in step A404, to ensure the accuracy of the landing event, this application introduces a micro-Doppler time-spectrum verification mechanism. The micro-Doppler time-spectrum can reflect the fine characteristics of human movement, especially the minute vibrations and velocity changes when the foot contacts the ground. By analyzing the micro-Doppler time-spectrum characteristics at the moment of the landing event, it can be determined whether the event conforms to the typical characteristics of a gait event, thereby eliminating misjudgments caused by other non-gait movements.

[0085] Therefore, in step A405, only landing events confirmed as step events by micro-Doppler time-spectrum verification will have their occurrence time added to the step event time series. Furthermore, for the convenience of subsequent calculation of gait symmetry indices, a marker indicating whether the step event was generated by the left or right foot can also be recorded simultaneously.

[0086] This application's solution effectively addresses the problem of insufficient accuracy in traditional step event detection methods by refining the step event detection process into multiple logical steps and introducing a micro-Doppler time-spectrum verification mechanism. Specifically, firstly, by calculating the velocity and extracting the forward direction component from the trajectories of the left and right foot centroids (steps A401 and A402), a preliminary ground contact event can be identified (step A403). This method, based on a zero velocity, can relatively intuitively capture the moment the foot touches the ground. However, velocity information alone may not completely eliminate misjudgments caused by non-gait movements (such as swaying in place or foot slippage). To address this limitation, this application further utilizes micro-Doppler time-spectrum to verify the preliminary ground contact event (step A404). The micro-Doppler time-spectrum can provide fine features of foot movement, such as the impact and slight slippage when the foot touches the ground, which exhibit specific patterns in the time-spectrum. By comparing the micro-Doppler time spectrum corresponding to a landing event with preset step event characteristics, true step events and non-step events can be effectively distinguished. Therefore, only landing events verified by the micro-Doppler time spectrum are confirmed as step events and recorded (step A405), thus greatly improving the accuracy and reliability of step event detection and avoiding deviations in subsequent gait evaluation parameters due to misjudgment.

[0087] Through the above technical solution, this application provides a more accurate and reliable method for gait event detection. Compared to schemes that rely solely on speed or simply combine micro-Doppler time-spectrum analysis, this application effectively eliminates the interference of non-gait motion on gait event determination by introducing forward velocity component analysis and micro-Doppler time-spectrum verification mechanisms, significantly reducing the false positive rate. Therefore, the detected gait event time series are more accurate, laying a solid foundation for the accurate calculation of subsequent gait assessment parameters such as gait frequency, stride length, walking speed, and gait symmetry indices, ultimately improving the accuracy and clinical reference value of Parkinson's disease assessment results.

[0088] Preferably, step A404 may include:

[0089] Extract the micro-Doppler time spectrum within a preset time window centered on the time of the landing event, and use it as the reference micro-Doppler time spectrum;

[0090] Extract the peak time series of the reference micro-Doppler time spectrum, and denote it as the first peak time series;

[0091] Extract the foot's movement speed within the time window corresponding to the landing event, and use it as the reference speed data;

[0092] Extract the peak time series of the control velocity data and denote it as the second peak time series;

[0093] Calculate the correlation between the first peak time series and the second peak time series;

[0094] If the correlation is greater than the preset correlation threshold, the landing event is determined to be a step event.

[0095] Specifically, the preset time window length can be set according to the actual application scenario and radar data sampling rate, for example, it can be set to 0.5 seconds to 2 seconds. The purpose is to capture sufficiently rich micro-Doppler features and foot velocity changes before and after the landing event. The peak time series of the micro-Doppler time spectrum refers to the sequence of time points within the time window where the energy or amplitude in the micro-Doppler time spectrum reaches a local maximum. These peaks typically correspond to specific stages of foot movement, such as toe lift-off or heel strike. The control velocity data refers to the foot movement velocity data obtained by differentiating the trajectory of the foot's center of mass within the same time window. The peak time series of the control velocity data refers to the sequence of time points where the foot movement velocity reaches a local maximum within the time window; these peaks reflect the velocity changes of the foot during the swing or support phase. Correlation can be calculated using various statistical methods such as the Pearson correlation coefficient and cross-correlation function, with the aim of quantifying the consistency or synchronicity between the two time series. The preset correlation threshold is a criterion used to determine whether a landing event is a step event. This threshold can be adjusted based on experimental data and expert experience to balance the sensitivity and specificity of the detection.

[0096] The proposed solution uses correlation analysis between the peak time series of the micro-Doppler time spectrum and the peak time series of foot movement velocity to more accurately verify whether a ground contact event is a stepping event. This is because when a genuine stepping event occurs, there is a high degree of synchronicity and intrinsic correlation between the foot's motion state (including velocity changes) and the resulting micro-Doppler effect. For example, the foot reaches its peak velocity during the swing, and the micro-Doppler time spectrum also shows a characteristic peak at the corresponding time point. By calculating the correlation between these two peak time series, this intrinsic correlation can be effectively captured. A high correlation indicates a high degree of consistency between foot movement and micro-Doppler characteristics, thereby enhancing the confidence in determining a stepping event and avoiding misclassification of non-gait-related ground contact as a stepping event.

[0097] Through the above technical solution, this application can significantly improve the accuracy and robustness of gait event detection. By combining micro-Doppler time-spectrum data and foot movement speed data for cross-validation, false positives caused by other interfering factors (such as body swaying, ground friction, etc.) can be effectively eliminated, ensuring that the detected gait events are real and closely related to the gait cycle. This is crucial for the accurate calculation of subsequent gait assessment parameters (such as cadence, stride length, etc.), thus providing more reliable and accurate data support for the assessment of Parkinson's disease patients.

[0098] In step A5, at least one of the following needs to be calculated: gait frequency, gait frequency mutation parameter, stride length, walking speed, and gait symmetry index. These gait evaluation parameters can be calculated using the following methods.

[0099] The steps for calculating the step frequency may include:

[0100] The average step period is calculated based on the time series of step events, and the reciprocal of the average step period is calculated as the step frequency.

[0101] By analyzing the time series of step events, the time intervals between adjacent step events can be determined, and the average of these time intervals can be calculated to obtain the average step cycle. Step frequency can be understood as the number of steps that occur per unit time, and it is calculated as the reciprocal of the average step cycle. Its purpose is to quantify the user's walking rhythm.

[0102] The steps for calculating the step frequency mutation parameters may include:

[0103] Extract continuous segments of a preset duration (which can be set according to actual needs) from the time series of step events and use them as a reference time series;

[0104] Multiple sets of window data are extracted from the reference time series using a window with a preset number of steps (which can be set according to actual needs, such as 10 steps);

[0105] Calculate the step period standard deviation for each group of window data, and divide the step period standard deviation by the step frequency to obtain the coefficient of variation for each group of window data;

[0106] When the coefficient of variation exceeds the preset coefficient threshold (which can be set according to actual needs), a step frequency mutation event is determined to have occurred, and the number of occurrences of the step frequency mutation event is counted as the step frequency mutation parameter.

[0107] The gait frequency mutation parameter is used to assess the stability of a user's gait. By sliding a window across a reference time series and calculating the standard deviation of the step period and the coefficient of variation, abnormal fluctuations in gait rhythm can be detected. When the coefficient of variation (the ratio of the standard deviation of the step period to the gait frequency) exceeds a preset threshold, it indicates a significant decrease in gait stability, i.e., a gait frequency mutation event has occurred. Counting the number of these events (within a preset time period) quantifies the degree of gait instability.

[0108] The steps for calculating walking speed may include:

[0109] Based on the position of the centroid of the human point cloud cluster in each frame, the average velocity of the centroid of the human point cloud cluster is calculated and used as the walking speed.

[0110] Specifically, the centroid positions of the human body point cloud clusters are extracted in step A3, representing the overall position of the user's body. By tracking the displacement of these centroids between consecutive frames and combining this with the frame rate (the reciprocal of the frame rate is the inter-frame time interval), the user's average movement speed within the test area, i.e., walking speed, can be calculated.

[0111] The step of calculating the step size may include:

[0112] Divide walking speed by step frequency to get stride length.

[0113] In practical applications, stride length refers to the average distance a user travels with each step. Given walking speed and cadence, stride length can be directly calculated by dividing walking speed by cadence, reflecting the amplitude of the user's gait.

[0114] The steps for calculating the gait symmetry index may include:

[0115] Based on the centroid position of each human point cloud cluster and the time sequence of step events, calculate the single step length corresponding to each step event; for example, obtain the moving distance of the centroid position of the human point cloud cluster within the time interval of adjacent step events, and divide the moving distance by the time interval of adjacent step events to obtain the single step length corresponding to the corresponding step event.

[0116] Calculate the average step length of the left foot and the average step length of the right foot based on the single step length corresponding to each step event; for example, based on the pre-recorded markers of whether each step event was generated by the left foot or the right foot, calculate the average single step length corresponding to the left foot to obtain the average step length of the left foot, and the average single step length corresponding to the right foot to obtain the average step length of the right foot.

[0117] The gait symmetry index is calculated using the following formula:

[0118] ;

[0119] in, As an index of gait symmetry, The average stride length of the left foot. This represents the average stride length of the right foot.

[0120] The Gait Symmetry Index (SLA) is used to assess the coordination of left and right gait. It is calculated by separately calculating the average stride length of the left and right feet. and By substituting this value into the formula above, the difference in stride length between the left and right feet can be quantified. Calculating this indicator helps identify gait asymmetry, a common problem in Parkinson's patients.

[0121] Through the above technical solution, this application provides a comprehensive and accurate method for calculating gait assessment parameters. These parameters not only include basic gait frequency, stride length, and walking speed, but also introduce gait frequency mutation parameters, which are important indicators of gait stability, as well as gait symmetry indicators reflecting left-right foot coordination. These detailed and quantitative parameters can provide richer and more objective data support for gait assessment of Parkinson's patients, thereby helping doctors to more accurately determine the severity and progression of the patient's condition, and providing a scientific basis for subsequent treatment and rehabilitation.

[0122] In some implementations, step A6 includes:

[0123] The gait evaluation parameters are weighted and combined to obtain a comprehensive score.

[0124] By comparing the comprehensive score with pre-stored reference data for Parkinson's patients with different disease severity, the severity of the user's Parkinson's disease is determined as the Parkinson's disease assessment result; the reference data is the reference range for the comprehensive score.

[0125] Specifically, after calculating multiple gait assessment parameters, such as gait frequency, gait frequency mutation parameters, stride length, walking speed, and gait symmetry indices, these parameters need to be weighted and synthesized. Weighted synthesis refers to assigning a weight coefficient to each gait assessment parameter based on its relative importance in reflecting the severity of Parkinson's disease. For example, some parameters may be more closely associated with specific symptoms of Parkinson's disease (such as bradykinesia and gait instability) and therefore can be given higher weights. By multiplying the standardized value of each parameter by its corresponding weight and summing all the weighted parameter values, a comprehensive score can be obtained. This comprehensive score can more comprehensively reflect the patient's overall gait condition.

[0126] Furthermore, the resulting composite score is compared with pre-stored reference data. This reference data is not a threshold for a single gait parameter, but rather a reference range for the composite score, corresponding to different severity levels of Parkinson's disease. For example, one composite score range might be pre-defined for "mild Parkinson's disease," another for "moderate Parkinson's disease," and yet another for "severe Parkinson's disease." By determining whether a user's composite score falls within a specific reference range, the severity of their Parkinson's disease can be identified and used as the final assessment result. The reference data is typically established through statistical analysis of gait data from a large number of Parkinson's patients with known disease severity.

[0127] This application's solution effectively addresses the limitations of traditional methods that rely on single or independent parameter assessments by introducing a weighted comprehensive processing of gait evaluation parameters. Specifically, since Parkinson's disease has multifaceted effects on gait, a single parameter often fails to fully characterize the condition. By weighted integration of multiple parameters such as gait frequency, stride length, and symmetry, the abnormal characteristics of the patient's gait can be captured more comprehensively, and corresponding weights can be assigned according to their different contributions to the disease assessment, thereby generating a comprehensive score that more accurately reflects the patient's overall condition. Furthermore, comparing this comprehensive score with reference ranges for different disease severities, rather than simply comparing it with a single threshold, allows for a more refined differentiation of the patient's disease stage, avoiding assessment biases caused by interactions between parameters or individual differences.

[0128] Through the above technical solution, this application provides a more comprehensive, accurate, and refined method for assessing Parkinson's disease. Compared with traditional methods that rely solely on single or simple comparisons of multiple gait parameters, this application generates a more representative comprehensive score by weighting and synthesizing each gait assessment parameter, thereby significantly improving the reliability and clinical applicability of the assessment results. Furthermore, by comparing the comprehensive score with reference ranges for different disease severity, it is possible to more effectively identify and differentiate the disease stages of Parkinson's patients, providing a more precise basis for clinical diagnosis, treatment plan development, and efficacy evaluation, thus contributing to the early detection and dynamic management of Parkinson's disease.

[0129] refer to Figure 2 This application provides a gait assessment system for Parkinson's patients, which includes a millimeter-wave radar 1 and a processing device 2;

[0130] Millimeter-wave radar 1 is used to transmit radar beams to the test area and receive radar echo signals, which are then uploaded to the processing device 2.

[0131] Processing device 2 is used to acquire radar echo signals from the user walking within the test area using millimeter-wave radar 1, generating radar point clouds and micro-Doppler time spectra; performing coordinate transformation and normalization on the radar point clouds to obtain preprocessed radar point clouds; extracting human point cloud clusters from each frame of preprocessed radar point clouds using clustering methods, and identifying the centroid positions and trajectories of the left and right feet using Kalman filtering and Hungarian algorithm; calculating the speeds of the left and right feet based on their centroid positions and trajectories, and detecting gait events based on the micro-Doppler time spectra; calculating gait evaluation parameters based on the time sequence of gait events and the centroid positions of the human point cloud clusters; gait evaluation parameters include at least one of gait frequency, gait frequency mutation parameters, stride length, walking speed, and gait symmetry index; comparing the calculated gait evaluation parameters with pre-stored reference data to generate the user's Parkinson's disease assessment results (the specific process is described in steps A1-A6 above).

[0132] The millimeter-wave radar 1 can be a standalone radar sensor module, operating at frequencies such as 60GHz-70GHz, with a maximum detection range of 5m-10m, providing high-resolution distance, velocity, and angle information. In practical applications, the millimeter-wave radar 1 can be fixedly mounted on a wall or bracket in the test area to ensure stable monitoring of the user's gait. For example, the millimeter-wave radar 1 can be configured as a single-transmit multiple-receive (MIMO) array to improve spatial resolution and target recognition capabilities. Furthermore, the millimeter-wave radar 1 can also be integrated into a smart home device, such as a smart speaker or smart display, for easier deployment and use.

[0133] Processing unit 2 is used to receive radar echo signals uploaded by millimeter-wave radar 1 and perform a series of complex signal processing and data analysis tasks. Processing unit 2 can be a high-performance central processing unit (CPU), such as an embedded processor, a dedicated digital signal processor (DSP), or a programmable gate array (FPGA), to meet the needs of real-time data processing. Processing unit 2 can be configured as a local computing unit, directly integrated into the evaluation device, or connected to a cloud server via a network for data processing and storage.

[0134] Please refer to Figure 3This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device includes a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other via a communication bus 303 and / or other forms of connection mechanisms (not shown). The memory 302 stores a computer program executable by the processor 301. When the electronic device is running, the processor 301 executes the computer program to perform the Parkinson's disease patient gait assessment method in any optional implementation of the above embodiments, to achieve the following functions: acquiring radar echo signals when the user walks within a test area using millimeter-wave radar, generating radar point clouds and micro-Doppler time spectra; and processing radar points... The cloud is subjected to coordinate transformation and normalization to obtain a preprocessed radar point cloud. Human point cloud clusters are extracted from each frame of the preprocessed radar point cloud using a clustering method, and the centroid positions and trajectories of the left and right feet are identified using Kalman filtering and the Hungarian algorithm. The velocities of the left and right feet are calculated based on their centroid positions and trajectories, and gait events are detected using micro-Doppler time-spectrum analysis. Gait evaluation parameters are calculated based on the time series of gait events and the centroid positions of the human point cloud clusters. These gait evaluation parameters include at least one of the following: gait frequency, gait frequency mutation parameter, stride length, walking speed, and gait symmetry index. The calculated gait evaluation parameters are compared with pre-stored reference data to generate the user's Parkinson's disease assessment results.

[0135] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs a gait assessment method for Parkinson's patients according to any optional implementation of the above embodiments, to achieve the following functions: acquiring radar echo signals of a user walking in a test area using millimeter-wave radar, generating radar point clouds and micro-Doppler time spectra; performing coordinate transformation and normalization on the radar point clouds to obtain preprocessed radar point clouds; extracting human point cloud clusters from each frame of preprocessed radar point clouds using a clustering method, and identifying the centroid positions of the left and right feet using Kalman filtering and the Hungarian algorithm; calculating the speeds of the left and right feet based on their centroid positions, and detecting gait events based on the micro-Doppler time spectra; calculating gait assessment parameters based on the time sequence of gait events and the centroid positions of the human point cloud clusters; the gait assessment parameters include at least one of gait frequency, gait frequency mutation parameters, stride length, walking speed, and gait symmetry index; comparing the calculated gait assessment parameters with pre-stored reference data to generate a Parkinson's disease assessment result for the user. The computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0136] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method of gait assessment in Parkinson's patients, characterized in that, The method comprises the following steps: A1. acquiring radar echo signals of a user walking in a test area by using a millimeter wave radar, and generating a radar point cloud and a micro-Doppler time-frequency spectrum; A2. performing coordinate conversion and normalization processing on the radar point cloud to obtain a preprocessed radar point cloud; A3. extracting a human point cloud cluster from each frame of the preprocessed radar point cloud by using a clustering method, and identifying the centroid position trajectories of the left and right feet by combining Kalman filtering and the Hungarian algorithm; A4. calculating the speeds of the left and right feet according to the centroid position trajectories of the left and right feet, and detecting the step events of the left and right feet by combining the micro-Doppler time-frequency spectrum; A5. calculating gait evaluation parameters according to the time sequence of the step events and the centroid point positions of the human point cloud cluster; the gait evaluation parameters comprise at least one of a step frequency, a step frequency mutation parameter, a step length, a walking speed, and a gait symmetry index; A6. comparing the calculated gait evaluation parameters with pre-stored reference data to generate a Parkinson's disease evaluation result of the user; Step A4 comprises: A401. calculating the moving speeds of the left and right feet according to the centroid position trajectories of the left and right feet; A402. calculating the speed components of the moving speeds of the left and right feet in the forward direction of the user to obtain the forward speeds of the left and right feet; A403. when the forward speed of the left foot or the right foot is zero, determining that a landing event occurs, and recording the landing event occurrence time; A404. verifying whether the landing event is a step event by using the micro-Doppler time-frequency spectrum; A405. if the landing event is a step event, adding the corresponding landing event occurrence time to the time sequence of the step events; The step of calculating the step frequency comprises: calculating an average step period according to the time sequence of the step events, and calculating the reciprocal of the average step period as the step frequency; The step of calculating the step frequency mutation parameter comprises: extracting a continuous segment of a preset time length from the time sequence of the step events as a reference time sequence; extracting multiple groups of window data in a preset step number window in the reference time sequence by sliding window; calculating the step period standard deviation of each group of window data, and dividing the step period standard deviation by the step frequency to obtain the coefficient of variation corresponding to each group of window data; when the coefficient of variation exceeds a preset coefficient threshold, determining that a step frequency mutation event occurs, and counting the occurrence number of the step frequency mutation event as the step frequency mutation parameter; The step of calculating the walking speed comprises: calculating the centroid point average speed of the human point cloud cluster as the walking speed according to the centroid point positions of the human point cloud cluster in each frame; The step of calculating the step length comprises: dividing the walking speed by the step frequency to obtain the step length; The step of calculating the gait symmetry index comprises: calculating the single-step length corresponding to each step event according to the centroid point positions of the human point cloud cluster in each frame and the time sequence of the step events; calculating the left foot average step length and the right foot average step length according to the single-step length corresponding to each step event; calculating the gait symmetry index according to the following formula: ； wherein, is the gait symmetry index, is the left foot average stride length, is the right foot average stride length.

2. The method for gait assessment of Parkinson's patients according to claim 1, characterized in that, Step A1 comprises: A101. acquiring radar echo signals of a user walking towards or away from the millimeter wave radar in a test area by using a millimeter wave radar; A102. generating a multi-frame radar point cloud according to the radar echo signal; A103. performing a short-time Fourier transform on the radar echo signal to obtain the micro-Doppler time-frequency spectrum.

3. The method of gait assessment of a Parkinson's patient according to claim 1, characterized in that, Step A3 comprises: A301. extracting a human point cloud cluster from each preprocessed frame of the radar point cloud by a clustering method, and obtaining the centroid point position of each human point cloud cluster; A302. performing local sub-clustering in each human point cloud cluster to screen out a sub-cluster meeting a preset height range and a point cloud quantity threshold as a candidate foot point cloud cluster, and obtaining the centroid point position of the candidate foot point cloud cluster; A303. according to the relative position relationship between the centroid point position of the candidate foot point cloud cluster and the human point cloud cluster, distinguishing the candidate foot point cloud cluster into a left foot point cloud cluster and a right foot point cloud cluster; A304. according to the centroid point position of each frame of the left foot point cloud cluster and each frame of the right foot point cloud cluster, identifying the centroid position trajectory of the left and right feet by using Kalman filtering and the Hungarian algorithm.

4. The method of gait assessment of a Parkinson's patient according to claim 1, characterized in that, Step A404 comprises: extracting the micro-Doppler time-frequency spectrum in a time window of a preset length centered on the landing event occurrence time as a reference micro-Doppler time-frequency spectrum; extracting a peak time sequence of the reference micro-Doppler time-frequency spectrum, denoted as a first peak time sequence; extracting the moving speed of the foot corresponding to the landing event in the time window as the control speed data; extracting a peak time sequence of the control speed data, denoted as a second peak time sequence; calculating the correlation degree of the first peak time sequence and the second peak time sequence; if the correlation degree is greater than a preset correlation degree threshold, determining that the landing event is a step event.

5. The method of gait assessment of a Parkinson's patient according to claim 1, characterized in that, Step A6 comprises: performing weighted synthesis on each gait evaluation parameter to obtain a comprehensive score; comparing the comprehensive score with reference data of different severity levels of Parkinson's patients pre-stored, determining the severity level of Parkinson's disease of the user as a Parkinson's disease severity assessment result; the reference data is a reference range of the comprehensive score.

6. A system for gait assessment of Parkinson's patients, characterized by The system comprises a millimeter wave radar and a processing device; the millimeter wave radar is used to emit a radar beam to a test area and receive a radar echo signal uploaded to the processing device; the processing device is used to obtain the radar echo signal of the user walking in the test area by using the millimeter wave radar, generate a radar point cloud and a micro-Doppler time-frequency spectrum; performing coordinate conversion and normalization processing on the radar point cloud to obtain a preprocessed radar point cloud; extracting a human point cloud cluster from each preprocessed frame of the radar point cloud by a clustering method, and identifying the centroid position trajectory of the left and right feet by combining Kalman filtering and the Hungarian algorithm; calculating the speed of the left and right feet according to the centroid position trajectory of the left and right feet, and detecting the step event of the left and right feet in combination with the micro-Doppler time-frequency spectrum; calculating gait evaluation parameters according to the time sequence of the step event and the centroid point position of the human point cloud cluster; the gait evaluation parameters comprise at least one of step frequency, step frequency mutation parameter, step length, walking speed, and gait symmetry index; comparing the calculated gait evaluation parameters with pre-stored reference data to generate a Parkinson's disease severity assessment result of the user; The processing device, when detecting a step event of the left foot and the right foot according to the velocity of the left foot and the right foot calculated from the centroid position trajectory of the left foot and the right foot and in combination with the micro-Doppler time-frequency spectrum, performs the following steps: A401. calculating the moving velocity of the left foot and the right foot according to the centroid position trajectory of the left foot and the right foot; A402. calculating the velocity component of the moving velocity of the left foot and the right foot in the forward direction of the user to obtain the forward velocity of the left foot and the right foot; A403. determining that a landing event occurs when the forward velocity of the left foot or the right foot is zero, and recording the landing event occurrence time; A404. verifying whether the landing event is a step event by using the micro-Doppler time-frequency spectrum; A405. if the landing event is a step event, adding the corresponding landing event occurrence time to the time sequence of step events; The step of calculating the step frequency comprises: calculating the average step period according to the time sequence of step events, and calculating the reciprocal of the average step period as the step frequency; The step of calculating the step frequency mutation parameter comprises: extracting a continuous segment of a preset time length from the time sequence of step events as a reference time sequence; extracting multiple groups of window data in a preset step number window in the reference time sequence by sliding window; calculating the step period standard deviation of each group of window data, and dividing the step period standard deviation by the step frequency to obtain the coefficient of variation corresponding to each group of window data; when the coefficient of variation exceeds a preset coefficient threshold, determining that a step frequency mutation event occurs, and counting the occurrence number of step frequency mutation events as the step frequency mutation parameter; The step of calculating the walking speed comprises: calculating the average speed of the centroid point of the human body point cloud cluster as the walking speed according to the centroid point position of each frame of the human body point cloud cluster; The step of calculating the step length comprises: dividing the walking speed by the step frequency to obtain the step length; The step of calculating the gait symmetry index comprises: calculating the single-step length corresponding to each step event according to the centroid point position of each frame of the human body point cloud cluster and the time sequence of step events; calculating the left foot average step length and the right foot average step length according to the single-step length corresponding to each step event; calculating the gait symmetry index according to the following formula: ； wherein, is the gait symmetry index, is the left foot average stride length, is the right foot average stride length.

7. An electronic device, comprising: The computer program is executed by the processor to run the steps of the gait evaluation method for Parkinson's patients according to any one of claims 1-5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to run the steps of the gait evaluation method for Parkinson's patients according to any one of claims 1-5.

Images