Real-time safety monitoring system based on motion state analysis

Abstract

The present application relates to the technical field of sports safety monitoring, and discloses a real-time safety monitoring system based on sports state analysis.The system comprises a sports state acquisition module, a historical sports data index module, a sports mode classification module, a sports ability evaluation module, a sports ability sequence generation module and a sports trend analysis module.The sports state acquisition module acquires real-time sports data and environmental state information of a target object.The historical sports data index module indexes historical sports data according to an identity.The sports mode classification module classifies the sports mode of the historical sports data, extracts historical sports parameters and behavior data sets, and calculates a historical sports ability score set through data deviation.The sports ability sequence generation module optimizes dimension reduction and generates a score sequence according to time.The sports trend analysis module analyzes sports trends according to the score sequence, obtains a sports ability improvement rate and a decline rate, and is suitable for multi-scene safety monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sports safety monitoring, in particular to a real-time safety monitoring system based on sports state analysis. BACKGROUND

[0002] In the current safety monitoring field, the demand for monitoring the movement state of target objects is increasing, and is widely used in sports training, special operation protection, elderly health care and other scenes. However, the existing monitoring system has many limitations in actual application, and it is difficult to meet the monitoring demand of real-time and accuracy.

[0003] Traditional sports state monitoring means mostly rely on single sensor data collection, and can only obtain basic movement information of the target object, such as position, speed, etc., and cannot fully capture the detailed features in the movement process, such as action amplitude, posture change and other key data. This single-dimensional data collection method leads to one-sidedness in the system's judgment of the target object's movement state, making it difficult to accurately identify potential safety risks. For example, in the high-altitude operation monitoring scene, relying only on position information cannot determine whether the operator has a dangerous state such as imbalance or abnormal action, and often can only conduct post-incident analysis after the accident occurs, and cannot achieve pre-warning.

[0004] The existing system lacks effective use of historical movement data. Most monitoring systems only use historical data as storage records, and do not establish a correlation analysis mechanism between historical data and real-time data, so they cannot determine whether the current movement state of the target object is normal by analyzing the historical movement rules. For example, in the monitoring of athletes' training, coaches need to understand the long-term movement ability trend of athletes to develop a scientific training plan, but the existing system cannot integrate and analyze the data such as speed, endurance and explosive power of athletes at different periods, and cannot intuitively present the improvement or decline of movement ability, resulting in a lack of data support for the development of training plan, and it is difficult to achieve the ideal training effect.

[0005] The existing system has the problem of low efficiency in data processing. Sports data usually has the characteristics of mass and multi-dimension, and traditional data processing methods cannot quickly optimize and reduce the dimension of these data, resulting in a long data processing period and failing to meet the real-time monitoring demand. In emergency situations, such as a sudden fall at home of an elderly person, if the system cannot process the movement data in time and issue a warning, it will delay the rescue opportunity and cause serious consequences. Moreover, the existing system lacks a unified standard and scientific calculation method for evaluating movement ability, and the evaluation result is subjective, which cannot accurately reflect the actual movement ability level of the target object, further reducing the practicality and reliability of the monitoring system. SUMMARY

[0006] The present application aims to provide a real-time safety monitoring system based on motion state analysis to solve the problems presented in the background.

[0007] To achieve the above-mentioned purpose, the present application provides a real-time safety monitoring system based on motion state analysis, which comprises:

[0008] A motion state acquisition module, which is used to acquire real-time motion data of a target object and collect environmental state information of the target object;

[0009] A historical motion data indexing module, which is used to index historical motion data of the target object in a historical time period according to an identity of the target object;

[0010] A motion pattern classification module, which is used to classify motion pattern information in the historical motion data, obtain a plurality of classified motion patterns, and extract a historical motion parameter set and a plurality of historical behavior data sets in the plurality of classified motion patterns;

[0011] A motion ability evaluation module, which is used to calculate a plurality of historical motion ability score sets according to a deviation degree of the historical motion parameter set and a standard motion parameter set;

[0012] A motion ability sequence generation module, which is used to perform optimized dimension reduction processing on the plurality of historical motion ability score sets, obtain a plurality of dimension-reduced motion ability score sets, and arrange them in time sequence to generate a plurality of historical motion ability score sequences;

[0013] A motion trend analysis module, which is used to perform motion trend analysis according to the plurality of historical motion ability score sequences to obtain a motion ability improvement rate and a motion ability decline rate.

[0014] Preferably, the motion state acquisition module acquires real-time motion data of a target object and collects environmental state information of the target object, which comprises:

[0015] Acquiring real-time motion trajectory data, motion speed data and motion acceleration data of the target object through a sensor device;

[0016] Collecting temperature data, humidity data and terrain feature data around the target object through an environmental monitoring device as environmental state information.

[0017] Preferably, the historical motion data indexing module indexes historical motion data of the target object in a historical time period according to an identity of the target object, which comprises:

[0018] According to the identity of the target object, all movement records of the target object in a historical time period are queried in a movement behavior database;

[0019] A movement mode type, a movement timestamp, and a movement environment parameter in each movement record are extracted to constitute historical movement data.

[0020] Preferably, the movement mode classification module classifies movement mode information in the historical movement data to obtain a plurality of classified movement modes, including:

[0021] The movement modes in the historical movement data are divided by using a clustering analysis method to obtain a plurality of movement mode categories;

[0022] Each movement mode category is assigned a unique identifier, and a typical movement parameter under each category is extracted as a classified movement mode.

[0023] Preferably, the movement ability evaluation module calculates a plurality of historical movement ability score sets according to a deviation degree of the historical movement parameter set and a standard movement parameter set, including:

[0024] A historical movement parameter set under each classified movement mode is obtained and compared with a pre-stored standard movement parameter set;

[0025] A movement ability score is assigned according to the deviation, wherein the smaller the deviation, the higher the movement ability score;

[0026] The movement ability score is weighted and corrected in combination with a completion time and a completion quality in the historical behavior data to generate a historical movement ability score set.

[0027] Preferably, the movement ability sequence generation module optimizes and reduces dimensions of the plurality of historical movement ability score sets to obtain a plurality of reduced-dimension movement ability score sets, including:

[0028] A reference score is selected in each historical movement ability score set;

[0029] A difference between other scores and the reference score is calculated, and a probability weight is assigned according to a difference size, wherein the larger the difference, the smaller the weight;

[0030] The score set is sampled and reduced in dimension based on the probability weight to retain key score data, forming a reduced-dimension movement ability score set.

[0031] Preferably, the movement trend analysis module analyzes a movement trend according to the plurality of historical movement ability score sequences to obtain a movement ability improvement rate and a movement ability decline rate, including:

[0032] Time series analysis is performed on each historical motion ability score sequence to extract the change trend characteristics in the sequence;

[0033] The change direction and rate of motion ability are identified by a machine learning classifier;

[0034] The average motion ability improvement rate and the average motion ability decline rate are calculated by integrating the analysis results of multiple sequences.

[0035] Preferably, the system further comprises an environment matching correction module, which is configured to match the environment state information with the plurality of classified motion patterns to obtain a matching motion pattern, and correct the motion ability improvement rate and the motion ability decline rate according to the difference between the environment state information and the standard environment state information of the matching motion pattern, to obtain a corrected motion ability improvement rate and a corrected motion ability decline rate;

[0036] The matching of the environment state information with the plurality of classified motion patterns to obtain a matching motion pattern comprises:

[0037] The similarity between the environment state information and the standard environment state information of each classified motion pattern is calculated;

[0038] The classified motion pattern with the highest similarity is selected as the matching motion pattern;

[0039] According to the specific difference value between the environment state information and the standard environment state information of the matching motion pattern, an environment correction coefficient is generated.

[0040] Preferably, the system further comprises a safety risk decision module, which is configured to make a safety risk decision according to the corrected motion ability improvement rate and the corrected motion ability decline rate, and generate a safety risk early warning scheme, comprising:

[0041] The corrected motion ability improvement rate and the corrected motion ability decline rate are input into a risk decision model;

[0042] According to the model output, the risk level of the current motion state is determined;

[0043] In combination with the abnormal records in the historical behavior data, a safety risk early warning scheme containing the risk level and the countermeasures is generated.

[0044] Preferably, the system further comprises a warning execution module, which is configured to execute corresponding warning operations according to the safety risk early warning scheme, comprising:

[0045] According to the risk level in the safety risk early warning scheme, a corresponding warning mode is selected;

[0046] sending the early warning information through an acousto-optic device, a display device, or a communication device;

[0047] record the early warning execution result and update to the motion behavior database.

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

[0049] The motion state acquisition module of the system can not only obtain real-time motion data of the target object, but also synchronously acquire environmental state information of the target object. This multi-dimensional data acquisition method breaks through the limitation of traditional single sensor data acquisition, enabling the system to comprehensively grasp the motion scene and motion details of the target object. For example, in outdoor work monitoring, the system can not only obtain motion data such as the action posture and movement trajectory of the worker, but also acquire environmental information such as temperature, humidity, light intensity, and whether there are obstacles, etc. By combining motion data and environmental data, the system can more accurately determine whether the motion state of the target object is affected by environmental factors, and further more accurately identify potential safety risks, avoiding misjudgment or missed judgment due to neglect of environmental factors, and making the monitoring process more comprehensive and detailed.

[0050] The cooperation of the historical motion data index module and the motion pattern classification module realizes the deep mining and utilization of historical motion data. The historical motion data index module accurately indexes the historical motion data of the target object according to its identity, providing a data basis for subsequent analysis; the motion pattern classification module classifies the motion pattern information in the historical motion data, obtains multiple classified motion patterns, and extracts the corresponding historical motion parameter set and historical behavior data set. This processing method can convert scattered historical data into regular motion patterns, enabling the system to quickly determine whether the current motion state of the target object conforms to its regular motion rules by comparing the current motion data with the historical classified motion patterns. For example, in the elderly health care scene, the system can classify the common motion patterns of the elderly such as daily walking, sitting, resting, etc. through historical data. When the elderly appear motion states that exceed the regular patterns, such as long-term static or sudden intense exercise, the system can detect the abnormality in time and provide effective reminders for caregivers.

[0051] The motion ability evaluation module obtains a plurality of historical motion ability score sets by calculating the deviation degree of a historical motion parameter set from a standard motion parameter set. This evaluation method based on data deviation eliminates the subjectivity of traditional evaluation methods, making the motion ability evaluation result more objective and scientific. Corresponding standard motion parameter sets can be set in different scenarios, such as for athletes training, the standard parameters can refer to the average level of professional athletes in this project; for ordinary old people, the standard parameters can refer to the general motion ability level of healthy old people of the same age. Through this personalized evaluation method, the actual motion ability of different target objects can be accurately reflected, providing a more practical basis for subsequent motion guidance or safety protection.

[0052] The motion ability sequence generation module optimizes and reduces the dimension of the plurality of historical motion ability score sets to obtain a plurality of reduced dimension motion ability score sets, and arranges them in time sequence to generate a plurality of historical motion ability score sequences. This processing process effectively solves the problem of low data processing efficiency of traditional systems. By optimizing and reducing the dimension, while retaining key data information, the data processing amount is greatly reduced, the data processing period is shortened, and the system can quickly respond to real-time monitoring requirements. For example, in the real-time monitoring of athletes in large-scale sports events, the system can quickly process a large amount of motion ability data to generate clear time sequences, allowing the coaching team to monitor the changes in the athletes' motion ability in real time and adjust the training arrangements in a timely manner.

[0053] The motion trend analysis module analyzes the motion trend according to the plurality of historical motion ability score sequences to obtain a motion ability improvement rate and a motion ability decline rate. By intuitively presenting the change trend of motion ability, this module can help relevant personnel to discover abnormal changes in the motion ability of the target object in advance. In special operation scenarios, if the system analysis finds that the motion ability decline rate of the operator is significantly increased, the management personnel can be prompted to arrange the operator to rest or have a physical examination, avoiding safety accidents caused by the decline of motion ability; in the training scenario of athletes, the motion ability improvement rate can intuitively reflect the training effect, helping the coaching team to adjust the training plan in a timely manner and ensuring the correctness of the training direction. At the same time, this trend analysis can also provide a reference for long-term motion planning, whether it is personal health management or team training arrangement, a more reasonable plan can be made based on clear trend changes, so that the application value of the monitoring system is further extended. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The timing diagram of the real-time safety monitoring system based on motion state analysis of the present application;

[0055] Figure 2 The flowchart for simplifying the process of the motion state collection and indexing system;

[0056] Figure 3 Flow chart for the motion pattern categorization module;

[0057] Figure 4 Flow chart for the motion capacity sequence generation and trend analysis system;

[0058] Figure 5 Flow chart for the environment matching correction module. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0060] Referring to Figure 1 The present application provides a real-time safety monitoring system based on motion state analysis, which comprises a motion state acquisition module, a historical motion data index module, a motion pattern categorization module, a motion capacity evaluation module, a motion capacity sequence generation module, a motion trend analysis module, an environment matching correction module, a safety risk decision module and a warning execution module. The specific implementation is as follows:

[0061] The motion state acquisition module is responsible for acquiring real-time motion data and environment state information of a target object, and data acquisition is realized through sensor devices and environment monitoring devices. The historical motion data index module retrieves relevant motion records in a historical time period from a motion behavior database according to the identity of the target object. The motion pattern categorization module divides the motion patterns in the historical motion data by using clustering analysis method, and extracts the categorized motion patterns and the corresponding historical motion parameters and behavior data set. The motion capacity evaluation module calculates a historical motion capacity score set by comparing the deviation degree of the historical motion parameter set and the standard motion parameter set. The motion capacity sequence generation module optimizes and reduces the dimension of the historical motion capacity score set, and generates a historical motion capacity score sequence arranged in time sequence. The motion trend analysis module performs trend analysis based on the historical motion capacity score sequence, and obtains the motion capacity improvement rate and the decline rate. The environment matching correction module matches the environment state information with the categorized motion patterns, generates an environment correction coefficient and corrects the motion capacity improvement rate and the decline rate. The safety risk decision module makes a risk decision according to the corrected motion capacity change rate, and generates a safety risk warning scheme. The warning execution module executes corresponding warning operations according to the warning scheme, and outputs warning information through various devices.

[0062] Embodiment 1: Referring to Figure 2, the core function of the motion state acquisition module is to comprehensively capture the dynamic information of the target object in the motion process and the static and dynamic characteristics of the environment in which it is located. The module relies on an integrated sensor network, which is usually composed of miniature sensor units attached to key parts of the target object's body, and environmental monitoring nodes deployed at fixed positions in the monitoring area. Taking a runner who is running outdoors as an example, the process of acquiring his motion state data is as follows: the inertial measurement unit (IMU) attached to the body continuously records the original data of its three-axis acceleration and three-axis angular velocity at a sampling frequency of 100 times per second. These high-frequency data are transmitted in real time to a local edge computing node, which has an embedded algorithm that first performs sensor data fusion and coordinate system unification, then eliminates signal jitter caused by step impact and sensor noise through a Kalman filter, and finally outputs smooth and accurate real-time motion trajectory data, motion speed data, and motion acceleration data. The motion trajectory is generated by integrating the absolute position information provided by the global positioning system (GPS) receiver and the relative displacement information calculated by the IMU, so that the continuity of the trajectory can be maintained in both open areas and signal-shielded areas. The motion speed is calculated by complementing the GPS speed information and the speed obtained by integrating the acceleration, and the motion acceleration is directly derived from the preprocessed data of the IMU.

[0063] On the route the runner is taking, pre-deployed wireless weather stations continuously collect and broadcast temperature and humidity data in the area. The runner's equipment also integrates a small environmental sensor to collect the temperature and humidity of his immediate micro-environment. The collection of terrain feature data is more complex: a lightweight LiDAR mounted on the runner's backpack emits laser pulses to the surrounding environment at a frequency of several rotations per second, and by measuring the time difference of the laser return, a high-precision point cloud model of the surrounding terrain is constructed. These raw point cloud data are sent in real time to the edge computing node, which has a real-time processing algorithm built-in that quickly extracts key terrain features from the point cloud, such as calculating the inclination angle of the current slope, identifying the road surface material as soil, gravel or asphalt and evaluating its roughness level, and detecting whether there are significant obstacles such as rocks, tree roots, etc. in the path ahead and their approximate size. All these processed environmental state information: temperature, humidity, terrain slope, road roughness, obstacle information are stamped with precise timestamps and associated with the same time motion data to form a multi-modal, time-aligned data package.

[0064] All the real-time collected data are transmitted to the nearest gateway device through Bluetooth Low Energy (BLE) or LoRa wireless technology, and then uploaded to the cloud-based central data repository through the cellular network by the gateway, completing the collection process of real-time motion data and environmental state information.

[0065] The task of the historical motion data indexing module is to accurately and efficiently retrieve relevant historical records from the vast historical data store based on the identity of the currently monitored target object. The identity is designed as a unique, tamper-proof cryptographic hash value generated from the user's registered biometric feature (e.g. heart rate feature fused fingerprint) to ensure data privacy and security. When the system starts to analyze the current target (i.e. the aforementioned runner), the historical motion data indexing module receives the identity hash value of the user passed from the upper layer application.

[0066] The module then initiates a structured query request to the system's motion behavior database. The database is a distributed time-series database optimized for handling time-series data. The core condition of the query statement is to filter out all motion records of the specific identity within a configurable historical time period (e.g. the past 365 days). Each motion record is a document containing multiple dimensions of fields in the database. Upon completion of the query, the module does not simply fetch all the raw data, but performs a refined extraction process.

[0067] For each historical motion record that meets the condition, the module extracts three key dimensions of information: motion pattern type, motion timestamp, and motion environment parameters. The motion pattern type is annotated by the classification model when the data is ingested, e.g. "flat slow running", "mountain climbing", "intermittent sprint", "technical downhill", etc. The motion timestamp records the precise start time and end time of the motion activity. The motion environment parameters are a nested structured data body containing the average temperature, average humidity, and the aggregated terrain features of the motion route such as average slope, maximum slope, and primary road surface type during the historical motion.

[0068] These extracted historical data items are reorganized into a new, lightweight historical motion dataset, sorted by motion timestamp from early to late. The dataset provides the subsequent analysis modules with detailed historical context about the target object's various motion activities under different environmental conditions over a period of time, which is the cornerstone of trend analysis, ability assessment, and risk prediction.

[0069] Embodiment 2: refer to Figure 3The motion pattern categorization module receives the processed historical motion dataset from the historical motion data indexing module. This dataset contains records of multiple motion activities of the target object in the past time period, each record contains a timestamp, a motion trajectory, a speed sequence, an acceleration sequence, and environmental parameters. The task of this module is to discover and define recurring motion patterns with common characteristics from these seemingly chaotic data. This process is not simply a name-based classification, but an unsupervised discovery based on the characteristics of the data itself. The module uses clustering analysis methods to achieve this goal. The module performs feature engineering on each historical motion record. These features are not raw data points, but statistical quantities and indicators that can summarize the core characteristics of a motion activity. For example, from the speed sequence, extract the average speed, maximum speed, and speed standard deviation; from the acceleration sequence, extract the average acceleration, maximum acceleration, and acceleration change frequency; from the motion trajectory, extract the total displacement, net displacement, and path tortuosity; from the environmental parameters, extract the average slope and elevation gain, etc. All these features are combined into a high-dimensional feature vector, representing a complete motion record.

[0070] A clustering algorithm (such as the K-means++ algorithm) is applied to all these feature vectors. The goal of the algorithm is to automatically group these vectors so that vectors within the same group (i.e., the same cluster) are very similar to each other, while vectors between different groups are significantly different. Similarity is measured by calculating the Euclidean distance between vectors. The algorithm needs to pre-set a rough range of cluster numbers, but the final number of specific categories is determined by the data distribution itself. After multiple iterations of calculation, the algorithm converges, and each historical motion record is assigned a cluster label, meaning it is categorized into a certain motion pattern category.

[0071] The module assigns a unique identifier to each identified motion pattern category, such as "pattern_001", "pattern_002", etc. More importantly, the module extracts the typical motion parameters of the pattern from all motion records belonging to the same category. These parameters are not simply average values, but are selected by analyzing the distribution of data within the cluster to choose the most representative values. For example, for the "slow and steady run on a gentle slope" category, its typical motion parameters may include an average speed range, a characteristic step frequency, a typical ground contact time, and a common slope interval. These categorized motion patterns and their typical parameters are stored in a pattern knowledge base, which continuously evolves and refines as new data is continuously incorporated.

[0072] On this basis, the movement ability evaluation module accurately quantifies the performance of the target object in each historical movement. The work of this module is divided into two closely linked steps. First, the module obtains the classified movement mode to which each historical movement record belongs, and retrieves the standard movement parameter set corresponding to this mode from the mode knowledge base. This standard parameter set represents an idealized or expected optimal performance level under this movement mode. For example, for "Mode_001 (high-intensity interval training)", the standard parameters may include the preset target sprint speed, target recovery period heart rate, ideal movement amplitude, etc.

[0073] The module compares the actual observed movement parameter set in the single historical movement record with this standard set item by item. This comparison is not a binary pass or fail judgment, but a calculation of the deviation degree of each parameter. The deviation calculation considers the nature and unit of the parameter and performs standardization processing, so that parameters of different dimensions can be compared comprehensively. For example, the speed deviation may be calculated in percentage form, while the movement amplitude deviation may be calculated in the form of angle difference. All single deviations are integrated into a total deviation index through a weighted function, and the weights can be adjusted according to the key performance indicators of different movement modes. The total deviation index is mapped to a preset score interval, such as 0 to 100 points. The mapping function is designed so that the smaller the deviation, the higher the movement ability score, reflecting that the performance is closer to the ideal standard. A near-perfect performance may score 95 points or more, while a performance with significant deviation may score below 70 points.

[0074] The parameter comparison score alone may not fully reflect the quality of movement performance, so the module introduces a second step: weighted correction. The basis for correction comes from the completion time and completion quality information recorded in the historical behavior data set. The completion time refers to the time spent to complete a specific training unit or distance; the completion quality may include subjective feeling score, technical action evaluator score or other objective indicators such as touch ground balance. These factors are converted into correction coefficients. For example, in the case of the same parameter deviation, completing the training in a shorter time will result in a positive correction coefficient, thereby increasing the final score; while a record indicates that the movement technique action is distorted, a negative correction coefficient will be introduced, reducing the final score. After such weighted correction, the generated historical movement ability score more comprehensively and fairly reflects the comprehensive ability exhibited in a single movement. All historical movement records are processed through this process, and the final output is a set of historical movement ability scores in chronological order, containing multiple score values, each score accurately corresponding to a past movement event and its context.

[0075] Example 3: see Figure 4The athletic ability sequence generation module receives multiple sets of historical athletic ability scores from the athletic ability assessment module. Each set corresponds to a categorized athletic pattern, containing the ability scores of all historical athletic records of the target object under that pattern. These sets may have an excessive number of data points, uneven density, or contain redundant information, making direct use for trend analysis inefficient and susceptible to noise interference. Therefore, the core task of the module is to optimize and reduce the dimensionality of each score set, generating a more concise data sequence that is more suitable for time series modeling while preserving its core variation characteristics.

[0076] The dimensionality reduction process begins by selecting a baseline score from each set of historical motion capability scores. The selection strategy for the benchmark score is not fixed but dynamically determined based on the statistical characteristics of the dataset. One strategy is to select the median score of the entire dataset, which effectively resists the influence of extreme outliers; another strategy is to select the score corresponding to a specific time point (such as the start of a training cycle) as the benchmark. Benchmark Score The determination of this marks the starting point for data standardization and weight allocation.

[0077] The module calculates every other score in the set. Compared with the benchmark score absolute difference This difference The deviation of an individual score from a benchmark point is quantified. Based on this difference, the module assigns a value to each score. Assign a probability weight Weight allocation follows a basic principle: difference. The larger the value, the more abnormal or volatile the data point may be, and the smaller its contribution to characterizing the long-term core trend may be. Therefore, its probability weight is retained. The smaller the value, the better. One way to calculate the weights to achieve this principle is as follows:

[0078]

[0079] in: Indicates the first The probability weight of each athletic ability score being sampled and retained, with a value range between 0 and 1. Indicates the first Individual score and benchmark score The absolute difference. It is a positive adjustment factor used to control the sensitivity of the weights to the magnitude of the difference; The larger the value, the faster the weight decays as the difference increases.

[0080] Based on the calculated set of probability weights The module probabilistically samples from the original set of historical athletic performance scores. This is a weighted random sampling process, where higher scores (i.e. scores closer to the benchmark, with less fluctuation) have a higher probability of being retained in the new set, while lower scores (i.e. scores further from the benchmark, with more fluctuation) are more likely to be discarded. Rather than simply removing low or high scores, this process filters out data points that are likely to represent random noise or short-term anomalies, while retaining the core data that best represents a stable performance level. After sampling is complete, a new, significantly reduced set of dimensionally reduced performance scores is formed. Finally, all of these reduced sets are arranged in their original chronological order, resulting in a number of clear historical performance score sequences. Each sequence corresponds to an athletic pattern, and reflects how the purified, short-term disturbance-removed performance scores evolve over time in that pattern.

[0081] The athletic trend analysis module then performs in-depth analysis on these refined historical performance score sequences to extract their underlying change patterns. The module performs time series analysis on each input historical performance score sequence. The data points in the sequence are arranged in chronological order, and the analysis aims to identify and quantify the long-term trend component, possible periodic changes (such as seasonal fluctuations), and random fluctuations contained therein. The module uses classical time series analysis methods such as the autoregressive integrated moving average model to fit the data. Through model fitting, a series of key trend features can be extracted, such as the overall slope of the sequence (indicating whether it is rising or falling in the long term), the curvature of the trend line (indicating whether the change is accelerating or decelerating), and the volatility of the sequence, etc. These features numerically describe the evolution behavior of the performance scores in that athletic pattern.

[0082] These extracted trend features are sent to a machine learning classifier for further interpretation. The classifier has learned the mapping relationship between many known trend patterns and the corresponding feature combinations during the training phase. For the current input trend feature vector, the classifier identifies the most likely change direction category it belongs to, and simultaneously estimates its change rate. The rate is a continuous numerical value representing the magnitude of the average change in performance scores per unit time.

[0083] The last step of the motion trend analysis module is to make a comprehensive integration. It will aggregate the change direction recognition results and the change rate estimates from all the analyzed motion pattern sequences. For all the sequences identified as "upward" direction, the average of their change rates is calculated, which is output as the overall motion ability upward rate. Similarly, for all the sequences identified as "downward" direction, the average of their change rates is calculated, which is output as the motion ability downward rate. These two rate indicators are finally presented in the form of scalar values, which comprehensively reflect the overall ability change trend and intensity of the target object, providing a core trend basis for subsequent risk assessment and decision-making.

[0084] Example 4: see Figure 5 When the environment matching correction module is started, it receives two inputs. One is the real-time environmental state information from the motion state acquisition module, and the other is the multiple classified motion patterns and their respective corresponding standard environmental state information generated by the motion pattern classification module. The real-time environmental state information is a multi-dimensional vector, specifically containing the current collected temperature data, humidity data, and key parameters obtained through terrain feature analysis, such as average slope, road unevenness index, and obstacle density. Each classified motion pattern is associated with a standard environmental state information, which is not a fixed value, but a numerical range defining the ideal or typical environmental conditions under which the pattern usually occurs. For example, the standard environment for the "gentle slope running" pattern may be defined as a slope between 3% and 8%, a temperature between 10 and 25 degrees Celsius, and a road surface of compacted soil.

[0085] The execution flow of the module begins with the matching process, which first compares the current real-time environmental state information with the standard environmental state information of each classified motion pattern in the knowledge base, calculating their similarity. This similarity calculation is not a simple scalar comparison, but a multi-dimensional space distance measurement based on all environmental parameters. The module compares each parameter value of the current environment with its position within the standard parameter range of a certain classified motion pattern. For example, if the current temperature is 22 degrees Celsius and the standard temperature range of "Pattern_A" is 18 to 26 degrees Celsius, the similarity in this dimension is high; if the current slope reaches 12%, and the standard slope range of "Pattern_A" is 2% to 5%, the similarity in this dimension is very low. The system uses a weighted cosine similarity algorithm to comprehensively evaluate the matching degree in all dimensions, and each environmental parameter (such as temperature, humidity, slope) is given different weights according to its known importance to the motion performance, and the slope parameter is usually given a higher weight. After calculation, the module outputs a comprehensive similarity score between 0 and 1 for each classified motion pattern, and the closer the score to 1, the more the current environment conforms to the typical conditions of the motion pattern.

[0086] The module selects the one with the highest similarity score to the current real-time environment among all categorized movement patterns, and determines it as the matching movement pattern. This selection means that the system judges that the type of movement pattern that the target object is most likely to perform in the current environmental conditions, or is most affected by the current environment, is the matching pattern. For example, the system can judge that the current steep and rugged mountain environment has the highest similarity to the "high-intensity mountain cross-country" pattern, and thus selects it as the matching movement pattern.

[0087] After determining the matching movement pattern, the module enters the correction coefficient generation stage. This stage aims to quantify the specific differences between the current environment and the ideal environment, and convert these differences into a correction multiplier that can be applied to the trend data. The module will compare the specific difference values of each parameter between the current environmental state information and the standard environmental state information of the matching movement pattern (usually taking the median value of the standard range as the representative value). For example, the difference between the current temperature and the standard temperature, the difference between the current humidity and the standard humidity, the difference between the current slope and the standard slope, etc. Each parameter difference value will be mapped to an independent environmental impact factor according to a pre-set lookup table based on knowledge of exercise physiology and biomechanics. This lookup table defines the expected impact direction and magnitude of different environmental parameters deviating from the ideal value on exercise ability. For example, a slope significantly greater than the ideal value will usually correspond to a negative impact factor, indicating that the environment is more difficult; while a temperature that is moderate and within the ideal range will correspond to a positive or neutral impact factor.

[0088] Table 1: Mapping table of environmental parameter differences and impact factors

[0089] Ambient parameter Current measurement value Matching pattern criterion value (median value) Absolute difference Calculated influence factor Temperature 22°C 20°C 2°C 0.98 Humidity 85% 60% 25% 0.92 Average slope 12% 4% 8% 0.82 Road unevenness High Medium - 0.88 Obstacle density Medium Low - 0.95

[0090] All these independent impact factors are then integrated into a single environmental correction coefficient. The integration method usually uses weighted geometric mean, because the influence of environmental factors is often interrelated and superimposed in a multiplicative manner, rather than independent addition. The weights are again determined by the importance of each parameter. The environmental correction coefficient calculated in this way is a value between 0 and 1 or slightly greater than 1. This coefficient encapsulates the expected overall impact of the current environmental conditions on exercise ability relative to the ideal conditions. A coefficient less than 1 indicates that the environmental conditions are more severe than typical conditions, and are expected to have a negative impact on exercise ability; a coefficient equal to 1 indicates that the environment is no different from typical conditions; and a coefficient greater than 1 indicates that the environmental conditions are better than typical conditions, which may be beneficial to exercise performance.

[0091] The module applies this environmental correction factor to correct the motion capacity increase rate and the motion capacity decrease rate previously calculated by the motion trend analysis module. The correction logic is that an unfavorable environment (correction factor < 1) will weaken the calculated capacity increase rate and amplify the calculated capacity decrease rate. Conversely, a favorable environment (correction factor > 1) will amplify the increase rate and weaken the decrease rate. The specific correction operation is a multiplication relationship: corrected motion capacity increase rate = original motion capacity increase rate x environmental correction factor; corrected motion capacity decrease rate = original motion capacity decrease rate / environmental correction factor. Through this series of steps, the motion capacity trend indicators originally calculated purely based on historical data without considering the immediate environmental background are effectively calibrated to the current specific environmental context, thereby significantly improving their representation ability for the current real risk situation. These corrected indicators provide more situational and more accurate basis for subsequent safety decisions.

[0092] In embodiment 5, the safety risk decision module receives the corrected motion capacity increase rate and the corrected motion capacity decrease rate from the environmental matching correction module. These two rate indicators quantify the net trend of the motion capacity of the target object over time under the current specific environmental conditions. The core task of the module is to interpret the risk implications implied by these two values and make a graded decision. This process is completed through a pre-built risk decision model. This model is not a simple threshold judgment, but a multivariate, nonlinear calculation graph. This model is trained with a large amount of historical data during the development stage, which records the correlation between different motion capacity change trends and the final occurrence of safety events (such as physical exhaustion, loss of control, accidental falls, etc.). The model can learn complex patterns, for example, even if the absolute value of the motion capacity decrease rate is not large, but if it occurs in a certain motion pattern and the environmental conditions are poor, the risk may be significantly amplified.

[0093] The module inputs the corrected increase rate and decrease rate into the model. The model processes internally, considering the ratio of the two rates, their absolute values, the direction of change, and their recent stability. After a series of nonlinear transformations and weighted integration, the model outputs a continuous risk probability value, which is between zero and one. The higher the value, the greater the likelihood that the current state predicts an imminent safety event. According to this probability value, the module maps it to a pre-set risk level. The risk level is usually divided into three levels. The low risk level corresponds to a probability value below the first threshold, indicating that the current motion state is stable, the capacity trend is flat or good, and the likelihood of an accident is very low. The medium risk level corresponds to a probability value between the first and second thresholds, indicating that there are signs of capacity decline or unfavorable environmental factors that need to be vigilant. The high risk level corresponds to a probability value exceeding the second threshold, indicating that the capacity is rapidly declining or the environment is extremely harsh, and an accident may be imminent.

[0094] After determining the base risk level, the module does not immediately generate the warning plan, but performs an important context-enhancing operation: it retrieves and fuses abnormality records from historical behavior data. These records are special events marked in past exercises, such as a brief loss of balance, an abnormal spike in heart rate, or an instance of deviating from the planned trajectory. If the current risk level is medium or high, and the historical records show that the target object has experienced relevant abnormal behavior in a similar exercise mode or environment, the final risk level may be upgraded by one level, or the response measures of the warning plan may become more specific and intense. For example, the system may upgrade from a general "pay attention to the environment" prompt to a specific warning "be alert to the left road obstacle, there is a record of stumbling in history". After all the judgments are completed, the module generates the final safety risk warning plan. The plan is a structured data object that explicitly includes the determined risk level (low, medium, high) and the corresponding list of specific response measures recommendations. These measures are pre-defined according to a large amount of domain knowledge, for example, low risk may correspond to "continuous monitoring", medium risk may correspond to "issue a prompt sound and suggest adjusting the breathing rhythm", and high risk may correspond to "issue a strong warning and suggest stopping exercise immediately or seeking help".

[0095] The warning execution module is the final interface between the system and the physical world, which receives the safety risk warning plan generated by the safety risk decision module and is responsible for converting it into specific, perceptible actions. The module first parses the warning plan and extracts the risk level code. According to the preset mapping rules, the module selects the corresponding warning mode combination for the risk level. The warning modes are mainly divided into three categories: acoustic warning, optical warning and communication warning. Acoustic and optical devices are used to create direct on-site warnings. For low risk, the module may only activate a green LED indicator light, providing a reassuring visual feedback. For medium risk, the module may control the buzzer to emit intermittent, gentle prompt sounds, while switching the LED indicator light to yellow flashing. For high risk, the module will start the highest level of acoustic and optical alarm, including red LED light high-frequency flashing and buzzer emitting continuous, urgent sound, aiming to forcibly attract the attention of the target object at the first time.

[0096] Display devices are used to provide more detailed text or graphical information, and the module will send the risk level and specific response measure text information to devices such as smart watches or head-up displays that the target object may wear. For example, the screen may display the detailed message "high risk: significant decrease in balance ability detected, suggest reducing speed immediately and finding support", providing clear operation guidance for the user.

[0097] The communication device is used to realize remote alarm and help-seeking. Through the integrated wireless communication module (such as a cellular mobile network or a satellite communication module), the module packs the key content of the warning information, including the target object identity, real-time geographic location, risk level and brief context, into a data packet, and sends it to the terminal device of the preset monitoring center or emergency contact person. This enables remote caregivers to know the on-site situation in a timely manner and initiate external rescue procedures when necessary.

[0098] After all the warning instructions are executed, the module generates an execution log, which records the time of the warning trigger, the specific warning method used, the sending status of the message (such as whether it successfully reached the monitoring center), and any feedback information during the execution process. This log is transmitted back and updated to the motion behavior database in real time, leaving a complete audit trail for this event and enriching historical data for optimizing future risk decision models.

[0099] It should be noted that the relational terms herein, such as first and second, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements recited, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0100] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A real-time safety monitoring system based on motion state analysis, characterized in that, The system comprises: a motion state acquisition module, configured to acquire real-time motion data of a target object and collect environmental state information in which the target object is located; a historical motion data indexing module, configured to index historical motion data of the target object in a historical time period according to an identity of the target object; a motion pattern classification module, configured to classify motion pattern information in the historical motion data, obtain a plurality of classified motion patterns, extract a set of historical motion parameters and a plurality of sets of historical behavior data under the plurality of classified motion patterns, and calculate a plurality of sets of historical motion ability scores according to a deviation degree of the set of historical motion parameters from a set of standard motion parameters; a motion ability evaluation module, configured to calculate a plurality of sets of historical motion ability scores according to a deviation degree of the set of historical motion parameters from a set of standard motion parameters; a motion ability sequence generation module, configured to perform optimized dimension reduction processing on the plurality of sets of historical motion ability scores, obtain a plurality of sets of reduced dimension motion ability scores, and generate a plurality of historical motion ability score sequences in time sequence; a motion trend analysis module, configured to perform motion trend analysis according to the plurality of historical motion ability score sequences, and obtain a motion ability improvement rate and a motion ability decline rate; the motion pattern classification module classifies motion pattern information in the historical motion data, and obtains a plurality of classified motion patterns, including: dividing motion patterns in the historical motion data by using a clustering analysis method to obtain a plurality of motion pattern categories; assigning a unique identifier to each motion pattern category, and extracting typical motion parameters under each category as a classified motion pattern; the motion ability evaluation module calculates a plurality of sets of historical motion ability scores according to a deviation degree of the set of historical motion parameters from a set of standard motion parameters, including: acquiring a set of historical motion parameters under each classified motion pattern, and comparing the set of historical motion parameters with a pre-stored set of standard motion parameters; assigning a motion ability score according to the deviation, wherein the smaller the deviation, the higher the motion ability score; combining completion time and completion quality in historical behavior data to weight and correct the motion ability score, and generating a set of historical motion ability scores.

2. The real-time safety monitoring system based on motion state analysis as claimed in claim 1, wherein, The motion state acquisition module acquires real-time motion data of a target object and collects environmental state information in which the target object is located, including: acquiring real-time motion trajectory data, motion speed data and motion acceleration data of the target object by using a sensor device; acquiring temperature data, humidity data and terrain feature data around the target object as environmental state information by using an environmental monitoring device.

3. The real-time safety monitoring system based on motion state analysis as claimed in claim 1, wherein, The historical motion data indexing module indexes historical motion data of the target object in a historical time period according to an identity of the target object, including: querying all motion records of the target object in the historical time period in a motion behavior database according to the identity of the target object. Extract the exercise mode type, exercise timestamp and exercise environment parameters in each exercise record to form historical exercise data.

4. The real-time safety monitoring system based on motion state analysis of claim 1, wherein, The exercise ability sequence generation module optimizes and reduces the dimensionality of the plurality of historical exercise ability score sets to obtain a plurality of reduced dimensionality exercise ability score sets, including: Selecting a reference score in each historical exercise ability score set; Calculating the difference between other scores and the reference score, and assigning a probability weight according to the difference size, the larger the difference, the smaller the weight; Based on the probability weight, the score set is sampled and reduced in dimension, and the key score data is retained to form a reduced dimensionality exercise ability score set.

5. The real-time safety monitoring system based on motion state analysis of claim 1, wherein, The exercise trend analysis module analyzes the exercise trend according to the plurality of historical exercise ability score sequences to obtain an exercise ability improvement rate and an exercise ability decline rate, including: Performing time series analysis on each historical exercise ability score sequence to extract the change trend characteristics in the sequence; Identify the change direction and rate of exercise ability through a machine learning classifier; Integrate the analysis results of multiple sequences to calculate the average exercise ability improvement rate and the average exercise ability decline rate.

6. The real-time safety monitoring system based on motion state analysis of claim 1, wherein, The system further comprises an environment matching correction module, which is configured to match the environment state information with the plurality of classified exercise modes to obtain a matching exercise mode, and correct the exercise ability improvement rate and the exercise ability decline rate according to the difference between the environment state information and the standard environment state information of the matching exercise mode, to obtain a corrected exercise ability improvement rate and a corrected exercise ability decline rate. The environment state information is matched with the plurality of classified exercise modes to obtain a matching exercise mode, including: Calculate the similarity between the environment state information and the standard environment state information of each classified exercise mode; Select the classified exercise mode with the highest similarity as the matching exercise mode; According to the specific difference value between the environment state information and the standard environment state information of the matching exercise mode, generate an environment correction coefficient.

7. The real-time safety monitoring system based on motion state analysis as claimed in claim 6, wherein, The system further comprises a safety risk decision module, which is configured to make a safety risk decision according to the corrected exercise ability improvement rate and the corrected exercise ability decline rate, and generate a safety risk warning scheme, including: Input the corrected exercise ability improvement rate and the corrected exercise ability decline rate into a risk decision model; Determine the risk level of the current exercise state according to the model output; Combine the abnormal records in the historical behavior data to generate a safety risk warning scheme containing the risk level and the countermeasures.

8. The real-time safety monitoring system based on motion state analysis as claimed in claim 7, wherein, The system further comprises a warning execution module, which is configured to execute corresponding warning operations according to the safety risk warning scheme, including: According to the risk level in the safety risk warning scheme, select the corresponding warning mode; Send warning information through sound and light devices, display devices or communication devices; Record the warning execution result and update it to the exercise behavior database.

Images