A sports fitness training cloud data collection system and method

Abstract

The application relates to the technical field of sports fitness cloud data, and discloses a sports fitness training cloud data collection system and method. The method drives the synchronous capture of electrocardio signals, inertial signals and pressure distribution signals through the starting of adaptive collection instructions, and the local real-time processing of the signals obtains a preliminary processing signal set with a time identifier, which is wirelessly sent to a remote end. The remote computing center analyzes the signals to construct a personal training file, identifies a targeted body area activated in specific training, comprehensively evaluates the mechanical stimulation and metabolic reaction of the area to quantize the area load value. The load evolution path is derived by comparing the historical and current area load values of the user, the constituent elements and execution parameters of subsequent training are automatically arranged according to the path and the preset load reference boundary, and a customized training plan is formed. The application realizes the synchronous and accurate collection of multi-modal signals and the intelligent training planning based on the dynamic evolution of individual load.

Description

Technical Field

[0001] This invention relates to the field of sports and fitness cloud data technology, specifically to a sports and fitness training cloud data acquisition system and method. Background Technology

[0002] In the current field of sports and fitness training monitoring, data acquisition typically relies on single-function, standalone devices. Electrocardiogram (ECG) monitoring, motion capture, and stress measurement are performed by different sensors or systems, each with its own independent clock and sampling mechanism. This approach results in time-base differences between electrophysiological, kinematic, and mechanical signals at the source of acquisition. Subsequent multimodal data analysis requires complex post-hoc synchronization corrections, and it is difficult to guarantee a strict instantaneous correspondence between signals, thus affecting the accuracy of the assessment.

[0003] Existing technologies struggle to provide precise and individualized objective assessments of training load. Due to fragmented data sources, systems cannot accurately correlate real-time cardiac metabolic responses with the mechanical stimulation experienced by specific body parts at specific times. Most solutions only offer general statistics such as overall exercise volume or average heart rate, failing to analyze and quantify the complex loads borne by the targeted muscle groups or joints actually activated during specific training movements. Training plan generation and adjustments are often based on generic models or the coach's experience, lacking dynamic analytical models that reflect the user's individual historical load data and their physical adaptability and changing trends to support decision-making.

[0004] A technical solution is needed that enables the simultaneous acquisition and fusion processing of multimodal physiological and motor signals. This solution should be able to identify and quantify personalized loads for specific body regions from the fused data, and automatically generate structured, customized training plans based on the historical evolution of the load. Summary of the Invention

[0005] The purpose of this invention is to provide a cloud data acquisition system and method for sports and fitness training to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a method for collecting cloud data on sports and fitness training, the method comprising:

[0007] Initiate a data acquisition command that is compatible with the training type, and drive the sensor network deployed on the user's body surface;

[0008] The sensor network synchronously captures electrical signals reflecting cardiac activity, inertial signals reflecting spatial changes in the limbs and trunk, and distribution signals reflecting contact pressure.

[0009] The captured electrical signals, inertial signals, and distributed signals are processed locally in real time to obtain a preliminary processed signal set carrying time stamps;

[0010] The initial processed signal set is transmitted completely to the remote computing center via a wireless communication network;

[0011] At the remote computing center, the received preliminary processing signal set is parsed and reassembled to construct a structured personal training profile;

[0012] Based on the kinematic information contained in an individual's training profile, target body regions that are activated during specific training types are identified;

[0013] For each targeted body region, the mechanical stimulation and metabolic response it experiences during the training cycle are comprehensively evaluated, and the region load value is quantified.

[0014] By retrieving stored user training files and comparing historical and current regional load values, the load evolution path can be deduced.

[0015] Based on the load evolution path and the preset load reference boundary, the components and execution parameters of subsequent training are automatically arranged to form a customized training plan.

[0016] Preferably, the activation and training type-adaptive acquisition command drives the sensor network deployed on the user's body surface, specifically including:

[0017] The system receives the type of sports and fitness training selected by the user through the terminal interface. The training type includes strength training, endurance training, or flexibility training.

[0018] Based on the selected training type, retrieve the corresponding sensor enablement list and parameter initialization settings from the device configuration library;

[0019] According to the sensor activation list, wake-up and self-test commands are sent to the electrocardiogram recorder, blood oxygen monitoring probe, inertial measurement unit array, and thin-film pressure sensing grid;

[0020] After confirming that all sensors have returned a ready signal, a synchronization acquisition start pulse is sent to ensure that the ECG recorder starts recording ECG waveforms, the blood oxygen monitoring probe starts acquiring blood oxygen saturation, the inertial measurement unit array starts tracking the angular velocity of the head, torso, and limbs, and the thin-film pressure sensing grid starts monitoring the pressure distribution on the hands and soles of the feet.

[0021] Preferably, the local real-time processing of the captured electrical signals, inertial signals, and distributed signals to obtain a preliminary processed signal set carrying a time stamp specifically includes:

[0022] In the built-in processor of the electrocardiogram recorder, a filtering program is applied to the raw electrocardiogram waveform to eliminate power frequency interference and electromyographic noise, and to extract a clear heart rate cycle signal.

[0023] In the local processing unit of the inertial measurement unit array, attitude calculation is performed on the raw angular velocity and acceleration data, and the real-time three-dimensional orientation data of each body segment relative to the global coordinate system is obtained through the quaternion update algorithm.

[0024] In the edge nodes of the thin-film pressure sensing grid, the original pressure distribution data is aggregated and characterized to identify the pressure center trajectory, peak pressure point, and pressure change curve over time.

[0025] A time identifier code generated by a high-precision clock source is uniformly added to the processed heart rate cycle signal, real-time three-dimensional orientation data, and pressure characteristic data.

[0026] The various types of data, after being appended with time identifiers, are packaged and integrated to form the preliminary processing signal set.

[0027] Preferably, the step of transmitting the preliminary processed signal set completely to the remote computing center via a wireless communication network specifically includes:

[0028] Check the current status of the wireless communication network and prioritize the 5G network channel as the data transmission link;

[0029] The preliminary processed signal set is encapsulated, and the source device identifier, training session identifier, and data packet sequence number are added to the header of the data packet;

[0030] Encapsulated data packets are continuously sent to the designated remote computing center Internet Protocol address via a selected 5G network channel in a streaming manner.

[0031] During transmission, the integrity and timing of data packets are monitored in real time. If packet loss or out-of-order delivery is detected, a data packet retransmission mechanism is triggered to ensure that the initial processing signal set is delivered completely.

[0032] Preferably, the step of parsing and reassembling the received preliminary processing signal set at the remote computing center to construct a structured personal training profile specifically includes:

[0033] The remote computing center's data receiving service sorts and reassembles the delivered preliminary processing signal set according to the data packet sequence number to restore its original timing.

[0034] After analyzing the reconstructed data, the heart rate cycle signal, real-time three-dimensional orientation data, and pressure characteristic data are separated into different processing pipelines based on the source device identifier;

[0035] Guided by time identification codes, data in different processing pipelines are strictly aligned in the time domain to ensure that physiological, posture and stress data at the same moment correspond precisely.

[0036] The aligned data is organized according to a preset structured format, which includes a data block header, a timeline, physiological data segments, posture data segments, and stress data segments, thereby generating the personal training profile.

[0037] Preferably, the step of identifying target body regions activated in specific training types based on kinematic information contained in an individual's training profile specifically includes:

[0038] Real-time three-dimensional orientation data of each body segment throughout the training process are extracted from the posture data segment of the individual training file.

[0039] Calculate the motion trajectory, average amplitude, and frequency of each body segment in three-dimensional space;

[0040] The calculated motion trajectory, average motion amplitude, and motion frequency are compared with a predefined training type action feature template library.

[0041] Based on the comparison results, body segments whose motion features match the selected training type's motion feature template with a set threshold are selected.

[0042] These body segments are defined as the target body regions activated during this training, and the activation intensity index of each target body region is recorded.

[0043] Preferably, for each targeted body region, a comprehensive assessment of the mechanical stimulation and metabolic response it experiences during the training cycle is performed to quantify the region's load value, specifically including:

[0044] For each targeted body region, extract the relevant stress characteristic data from the stress data segment of the individual training file, and calculate the average pressure and number of impacts experienced by each targeted body region.

[0045] Extract heart rate cycle signals and blood oxygen saturation data for the corresponding time period from the physiological data segments of the individual training file, and calculate the average heart rate and blood oxygen saturation change rate during the active period of each targeted body area;

[0046] A regional load quantification model was established, using the activation intensity index, average pressure, number of impacts, average heart rate, and rate of change of blood oxygen saturation of the targeted body region as input parameters.

[0047] The regional load quantification model calculates a dimensionless comprehensive value through a weighted fusion algorithm, which is then quantified as the regional load value of the target body region.

[0048] Preferably, the step of retrieving stored user training files, comparing historical and current regional load values, and deriving the load evolution path specifically includes:

[0049] Retrieve the user's personal training profile generated from several recent training sessions from the user database at the remote computing center;

[0050] Extract the region load values ​​calculated from the same target body regions in historical training from these historical individual training records;

[0051] Arrange the current and historical regional load values ​​in chronological order to form a time series of regional load values ​​for each targeted body region;

[0052] Trend analysis tools were used to process the time series of regional load values ​​for each targeted body region to identify whether the overall trend of load changes was increasing, decreasing, or remaining stable.

[0053] By combining the specific type and intensity of each training session, key turning points in the trend are marked, thereby drawing a load evolution path that reflects the adaptation and change process of each targeted body region.

[0054] Preferably, the step of automatically arranging the components and execution parameters of subsequent training based on the load evolution path and the preset load reference boundary to form a customized training plan specifically includes:

[0055] Obtain the load reference boundary set for the user's physical fitness level. The load reference boundary includes a safe upper limit, an effective stimulation lower limit, and an optimal range.

[0056] Compare the current regional load value and the latest trend of the load evolution path of each targeted body region with the corresponding load reference boundary;

[0057] For targeted body regions where the regional load value is consistently below the effective stimulus lower limit and the trend is flat, add isolated training movements for the targeted body regions in subsequent training programs, and increase the number of training sets and load weight.

[0058] For target body regions whose regional load values ​​are close to or exceed the safety limit, or whose evolution path shows excessively rapid growth, their training priority should be reduced in subsequent training programs, and compound movements or active recovery phases should be introduced.

[0059] By integrating the above scheduling decisions, specific training items for subsequent training cycles, the number of repetitions and sets for each item, rest time between sets, and suggested intensity are specified, generating a detailed customized training plan document.

[0060] Preferably, the present invention also includes a sports and fitness training cloud data acquisition system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the sports and fitness training cloud data acquisition method described above.

[0061] Compared with the prior art, the beneficial effects of the present invention are:

[0062] By simultaneously acquiring electrical signals reflecting cardiac activity, inertial signals reflecting spatial changes in the limbs and trunk, and distribution signals reflecting contact pressure through a sensor network deployed on the body surface, and performing local real-time processing and time stamping, this technique ensures strict alignment of multi-source heterogeneous data in the temporal dimension, overcoming the signal fusion challenges caused by independent devices and asynchronous clocks in traditional solutions. The resulting preliminary processed signal set with high temporal consistency enables the remote system to accurately establish an instantaneous correlation model between the body's metabolic response and mechanical load at a specific moment, providing a reliable data foundation for subsequent in-depth analysis.

[0063] The remote computing center analyzes and reconstructs the received signals to build a structured personal training profile. Based on the kinematic information within, it identifies the target body regions that are actually activated in specific training types. The system comprehensively evaluates the mechanical stimulation and metabolic response experienced by each target region during the training cycle, quantifying it into specific region load values. By accessing the user's historical training profile, the system compares the current load value with historical records, automatically deriving the dynamic evolution path of the region's load over time. This process represents a paradigm shift from overall athletic performance evaluation to refined load monitoring of specific physiological structures.

[0064] Based on the generated individualized load evolution path and preset load reference boundaries, the system automatically arranges the components and execution parameters of subsequent training to form a customized training plan. This function directly transforms the dynamic load assessment results into executable training instructions, realizing a closed-loop adaptive training plan based on objective physiological and mechanical data. This replaces traditional planning methods that rely on general templates, fixed cycles, or subjective experience judgments, enabling training interventions to respond in real time to the user's physical adaptation status and changing trends, achieving truly personalized adaptive training. Attached Figure Description

[0065] Figure 1 This is a schematic diagram illustrating the working principle of the cloud data acquisition method for sports and fitness training described in this invention.

[0066] Figure 2 A flowchart illustrating how the data acquisition command is initiated and the sensor network is driven.

[0067] Figure 3A flowchart illustrating the transmission of a pre-processed signal set via a wireless network;

[0068] Figure 4 A comparative diagram of activation intensity and pressure distribution in different targeted body regions during sports and fitness training;

[0069] Figure 5 A bar chart showing the distribution of activation intensity indices for various targeted body regions during physical fitness training. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Please see Figure 1 This invention provides a cloud data acquisition method for sports and fitness training. The method includes: initiating an acquisition command adapted to the user's selected training type, which drives a sensor network deployed on the user's body surface to start working. The sensor network synchronously captures electrical signals reflecting cardiac activity, inertial signals reflecting spatial changes in the limbs and trunk, and distributed signals reflecting contact pressure. These captured raw electrical, inertial, and distributed signals are processed locally in real time to obtain a set of preliminary processed signals carrying a unified time identifier. This preliminary processed signal set is transmitted completely to a remote computing center via a wireless communication network. At the remote computing center, the received data is parsed and reconstructed to build a structured personal training profile. Based on the kinematic information in this profile, the target body regions activated in this specific training type are identified. For each identified target body region, the mechanical stimulation and metabolic response it experiences during the training cycle are comprehensively evaluated to quantify a region load value. The system retrieves the user's stored past training profiles, compares the historical and current region load values, and thus derives the evolution path of the region load. Based on the obtained load evolution path and the preset load reference boundary, the system automatically arranges the components and execution parameters of subsequent training to form a customized training plan.

[0072] In one embodiment of the present invention, see [reference] Figure 2The user operates an application installed on their smartphone. The application presents a graphical interface with options such as "Strength Training," "Endurance Training," and "Flexibility Training." The user selects one of these options as the type of physical fitness training via touch, for example, selecting "Strength Training." The application sends the selected training type information to the system backend. Based on the received "Strength Training" type, the system retrieves the corresponding sensor activation list and parameter initialization settings from a pre-built device configuration library. The sensor activation list clearly lists the sensor devices required for this training, including an electrocardiogram (ECG) recorder, a pulse oximeter probe, an inertial measurement unit (IMU) array deployed on the upper limbs, trunk, and lower limbs, a thin-film pressure sensor grid integrated into the grip handle, and a thin-film pressure sensor grid placed in the insole. The parameter initialization settings include setting the ECG recorder's sampling frequency to 250 Hz, the pulse oximeter probe's measurement interval to 1 second, the IMU array's data output frequency to 100 Hz, and the thin-film pressure sensor grid's scanning frequency to 50 Hz.

[0073] Based on the retrieved sensor activation list, the system sends digital wake-up and self-test commands via Bluetooth Low Energy to the ECG recorder, pulse oximetry probe, inertial measurement unit array, and thin-film pressure sensor grid listed in the list. Upon receiving the commands, the ECG recorder performs circuit self-tests and electrode contact impedance checks; the pulse oximetry probe performs light source and photodetector calibration; the inertial measurement unit array performs zero-bias calibration of the gyroscope and accelerometer; and the thin-film pressure sensor grid performs channel connectivity and reference pressure tests. After completing their self-tests, all sensors wirelessly transmit a signal containing a "ready" status code and device identifier to the system. Once the system confirms that all sensors in the list have returned "ready" signals, it generates and broadcasts a synchronization acquisition start pulse with a precise timestamp. The synchronous acquisition start pulse ensures that the ECG recorder begins recording ECG waveforms at a sampling rate of 250 Hz, the blood oxygen monitoring probe begins acquiring blood oxygen saturation percentage data at 1-second intervals, the inertial measurement unit array begins tracking the angular velocity and acceleration of the head, torso, and limbs at a frequency of 100 Hz, and the thin-film pressure sensing grid begins monitoring the distribution of hand grip pressure and foot support pressure at a frequency of 50 Hz.

[0074] In some embodiments, the device configuration library stores the mapping relationship between different training types and sensor configurations, which exists in the form of a database table. For the "Strength Training" type, the mapping relationship indicates that all types of sensors are enabled and high-frequency parameter initialization settings are performed; for the "Endurance Training" type, the mapping relationship indicates that the electrocardiogram recorder, blood oxygen monitoring probe, and main trunk inertial measurement unit are enabled, and the electrocardiogram sampling frequency in the parameter initialization settings is adjusted to 125 Hz; for the "Flexibility Training" type, the mapping relationship indicates that the joint inertial measurement unit array and part of the thin-film pressure sensing grid are enabled, and the inertial measurement unit output frequency in the parameter initialization settings is adjusted to 50 Hz. The process of retrieving the sensor enabled list is a database query operation, and the query condition is the training type identifier selected by the user.

[0075] Optionally, the user selects the type of sports and fitness training through a dedicated terminal interface. The dedicated terminal is a portable hardware device equipped with a display screen and input buttons. The user selects the training type via button navigation and confirmation on the dedicated terminal interface. The dedicated terminal encodes the selected training type into a data packet and transmits it to the local gateway device via a wireless LAN. The local gateway device then retrieves the corresponding sensor activation list and parameter initialization settings from the device configuration library. The sending of wake-up and self-test commands is coordinated by the local gateway device, and the synchronous acquisition start pulse is also generated and broadcast by the local gateway device to reduce reliance on the wide area wireless network and minimize command latency.

[0076] As is understandable, the sensor enablement list not only lists the device type but also includes the device's network address identifier. Parameter initialization settings are a configuration file containing a series of key-value pairs, which is transmitted to each sensor device and parsed and executed. The synchronization acquisition start pulse is a short, high-priority wireless data packet carrying globally unified start time information. All sensors begin recording data upon parsing the start time information, thereby achieving time-domain alignment of multi-source signals.

[0077] In one embodiment of the present invention, see [reference] Figure 3When the user performs a bench press exercise, the sensor network deployed on the user's body surface synchronously captures the raw signals. The processor built into the ECG recorder applies a filtering program to the acquired raw ECG waveform. The filtering program includes a 50 Hz power frequency notch filter and a 5-30 Hz bandpass filter. The 50 Hz AC interference and high-frequency electromyographic noise mixed in the raw ECG waveform are eliminated, and the output signal shows a clear R-wave peak sequence, thereby extracting the heart rate cycle signal composed of adjacent R-wave intervals in milliseconds. At the same time, the local processing unit of the inertial measurement unit array attached to the user's upper arm, forearm, and chest begins to work. The local processing unit reads the raw angular velocity data and raw acceleration data output by the inertial measurement units, and performs attitude calculation on the raw angular velocity and acceleration data through a quaternion update algorithm. The discretized implementation of the quaternion update algorithm is as follows:

[0078]

[0079] in: This represents the attitude quaternion at the k-th sampling time. Indicates the angular velocity at the current moment. Components on the axis, Indicates the sampling time interval, symbol This represents a quaternion multiplication operation, which iteratively updates to obtain real-time three-dimensional orientation data of each body segment relative to the global coordinate system. The real-time three-dimensional orientation data is expressed in Euler angles or rotation matrix form.

[0080] In some embodiments, the edge nodes of the thin-film pressure sensing grid integrated into the bench press barbell handle aggregate and characterize the raw pressure distribution data. The edge nodes scan the resistance value of each pressure sensing unit in the grid at a frequency of 50 Hz and convert it into a pressure value (in kilopascals). One scan yields a two-dimensional matrix containing 256 pressure values. The two-dimensional matrix is ​​then aggregated to calculate the pressure center coordinates. The calculation formula is:

[0081]

[0082]

[0083] in: This represents the pressure value of the cell in the i-th row and j-th column of the grid. and The preset planar coordinates of the unit are represented. The pressure center coordinates are connected over time to form the pressure center trajectory. The edge nodes simultaneously identify the maximum value in the two-dimensional matrix of each scan as the peak pressure point and record the curve of the total pressure changing over time. After completing the above local real-time processing, the high-precision clock source adds a 64-bit time identifier code to the processed heart rate cycle signal, real-time three-dimensional orientation data and pressure feature data. The time identifier code has a precision of microseconds. The heart rate cycle signal, real-time three-dimensional orientation data and pressure feature data after the time identifier code are added are packaged and integrated into a structured data block. The data block contains a type identifier, data length and binary load, which constitutes the preliminary processed signal set.

[0084] Optionally, when preparing to send the initial processed signal set, the system performs a wireless communication network status check. The wireless communication network status check is achieved by querying the link quality indicator and signal-to-noise ratio of the device's network interface. When the link quality indicator of the 5G network channel is detected to be higher than -70 dB and the signal-to-noise ratio is greater than 20 dB, the system prioritizes the 5G network channel as the data transmission link. After the data transmission link is selected, the system encapsulates the initial processed signal set. The encapsulation operation adds the source device identifier, training session identifier, and an incrementing data packet sequence number to the header of the data packet. The encapsulated data packet is continuously sent to the designated remote computing center Internet Protocol address through the selected 5G network channel in a User Datagram Protocol (UDP) streaming mode.

[0085] It is understandable that during streaming transmission, the system monitors the integrity and timing of data packet transmission in real time. The monitoring mechanism is implemented by maintaining a sending window and an acknowledgment timer. The sending window records the sequence number of data packets that have been sent but have not been acknowledged. If the acknowledgment timer times out or a duplicate acknowledgment packet is received, packet loss is detected. If the data packet sequence number reported by the receiver is not continuous, out-of-order data is detected. When packet loss or out-of-order events are detected, the system triggers a data packet retransmission mechanism. The data packet retransmission mechanism selectively retransmits lost data packets or adjusts the sending rate according to the triggering reason to ensure that the initial processing signal set is completely delivered to the remote computing center.

[0086] In some embodiments, the filtering program applied by the built-in processor of the ECG recorder compares the original ECG waveform with the processed heart rate cycle signal. The original ECG waveform shows baseline drift and high-frequency spikes, while the processed heart rate cycle signal presents a stable R-wave peak and a flat baseline. The real-time three-dimensional orientation data output by the local processing unit of the inertial measurement unit array is compared with the original angular velocity data. The original angular velocity data is a three-axis numerical stream that changes over time, while the real-time three-dimensional orientation data is converted into a chest displacement curve to display the reciprocating motion trajectory of the bench press. The pressure center trajectory output by the edge nodes of the thin-film pressure sensor grid is compared with the original two-dimensional pressure distribution matrix. The original two-dimensional pressure distribution matrix has 256 discrete values ​​per frame, while the pressure center trajectory is a continuous two-dimensional coordinate curve, which intuitively reflects the movement path of the grip center of gravity in the palm area.

[0087] Optionally, when the wireless communication network status check is unavailable or the quality is below the threshold, the fourth-generation mobile communication technology network channel is selected as the alternative data transmission link. In the data encapsulation process, a total data packet length field and a checksum field are added to the data packet header for the receiver to verify data integrity. The streaming transmission mode adopts constant bit rate encoding to adapt to network bandwidth fluctuations. The data packet retransmission mechanism adopts a selective retransmission strategy to only retransmit the data packets corresponding to the missing data packet sequence numbers.

[0088] In one embodiment of the invention, a data receiving service at a remote computing center listens to a specific network port and receives network data streams from user devices. The data receiving service sorts and reassembles the delivered preliminary processing signal set according to the data packet sequence number carried in the packet header. This sorting and reassembly process is achieved by maintaining a reordering buffer. The buffer rearranges out-of-order data packets according to their sequence numbers and discards duplicate data packets, thereby restoring the original timing sequence of the preliminary processing signal set at the sending end. The data receiving service parses the reassembled data blocks and separates the data payload into different processing pipelines based on the source device identifier in the data packet header. Heart rate cycle signals are routed to the physiological signal processing pipeline, real-time 3D orientation data is routed to the kinematics processing pipeline, and pressure feature data is routed to the mechanical processing pipeline. Within each processing pipeline, the system performs strict time-domain alignment on the data from different pipelines based on a unified microsecond-level time identifier code attached to each data point. The time-domain alignment operation uses the minimum time identifier code interval as a reference for interpolation or sampling to ensure that the heart rate cycle signal, chest inertial measurement unit orientation data, and handle pressure center coordinates corresponding to the same moment are accurately correlated.

[0089] In some embodiments, the time-domain aligned multimodal data is organized and encapsulated according to a preset structured format. This structured format includes a fixed-length data block header, a time axis array with microsecond precision, a physiological data segment storing heart rate and blood oxygen saturation data, a posture data segment storing the orientation matrices of each body segment, and a pressure data segment storing the pressure center trajectory and peak pressure. The data block header contains a version number, user identifier, training type code, and total data length field, thereby generating a complete personal training profile. From the posture data segment of the personal training profile, the system extracts real-time three-dimensional orientation data of body segments labeled "left forearm," "right forearm," "chest," and "upper back" throughout the training process. This real-time three-dimensional orientation data is stored in the form of quaternion sequences. For each body segment's quaternion sequence, the system calculates its motion trajectory, average motion amplitude, and motion frequency. The motion trajectory is calculated by converting quaternions into three-dimensional spatial coordinate points and connecting them to form a path. The average motion amplitude is calculated by statistically averaging the Euclidean distances between consecutive coordinate points along the path. The motion frequency is calculated by performing a Fast Fourier Transform on the angular velocity sequence of the body segment along a principal axis and extracting the dominant frequency.

[0090] Optionally, the calculated motion trajectory, average amplitude, and frequency of each body segment are compared with a predefined training type motion feature template library. This library is stored in a database at a remote computing center. For the "bench press" exercise, the template might include the reciprocating motion trajectory of the chest segment in the vertical direction, the average amplitude range of the forearm segment, and a characteristic frequency range for the chest segment. The comparison process calculates a similarity score between the current body segment's motion features and the template features. The score can be expressed as:

[0091]

[0092] in: This indicates the overall matching score. Indicates the number of features. Indicates the first Preset weight coefficients for each feature This indicates the number of points calculated from the current body segment. 1 eigenvalue, Indicates the first in the template eigenvalues, function Calculate the similarity between two feature values. The system sets a matching threshold and filters out body segments with a comprehensive matching score exceeding this threshold. These body segments are defined as the target body regions activated in this training, and an activation intensity index is recorded for each target body region. The activation intensity index can be a normalized value that is related to the product of the matching score and the feature amplitude.

[0093] It is understandable that during the parsing and reassembly process, the data receiving service immediately verifies the checksum field contained in the data packet header upon receiving the data packet to ensure that no errors occur during data transmission. The structured format for constructing personal training profiles uses binary encoding to improve storage and access efficiency. The timeline array, physiological data segments, posture data segments, and stress data segments are indexed in the profile using fixed offset addresses. When calculating the motion trajectory of body segments, the sequence of position changes of the origin of the local coordinate system of the body segment in the global coordinate system is used as the trajectory. The average amplitude of motion is a normalized measure of the trajectory length, and the motion frequency reflects the speed of action repetition.

[0094] In some embodiments, real-time 3D orientation data extracted from posture data segments in a personal training profile is compared with calculated motion features. For example, the extracted original "chest" segment quaternion sequence is a set of abstract numerical values, while the calculated chest motion trajectory is displayed as a curve that fluctuates regularly in the vertical direction, with an average motion amplitude quantized to 0.45 meters and a motion frequency quantized to 0.8 Hz. These calculated feature values ​​are compared with the reference trajectory pattern, amplitude range, and frequency range of the chest segment in the "bench press" movement template. The calculated comprehensive matching score is 0.88, exceeding the threshold of 0.75, therefore the "chest" is identified as the target body region. At the same time, the "left forearm" and "right forearm" segments are also identified as target body regions due to the high matching degree of motion features with the template, while the "lower limbs" segment is not identified because its features have a low matching degree with the template.

[0095] Optionally, the training type action feature template library is pre-built by analyzing a large amount of standard action data and can be updated based on user feedback. The comparison process can calculate not only the overall matching degree, but also sub-scores on multiple dimensions such as trajectory shape similarity, amplitude error ratio, and frequency difference. The system records the activation intensity index of each targeted body region. The calculation of the activation intensity index can combine the matching degree score, the movement amplitude and duration of that segment during training, to form a scalar value between 0 and 1, used to quantify the relative activation degree of that region.

[0096] In one embodiment of the present invention, for each identified target body region, the system extracts pressure feature data related to that region from the pressure data segment of the constructed personal training profile. For the "chest" region, this mainly refers to the pressure sensor grid data of the area where the bench press bench contacts the back, while for the "left forearm" and "right forearm" regions, it extracts the pressure sensor grid data of the thin film of the barbell handle. The system calculates the average pressure and number of impacts experienced by each target body region during the entire training cycle. The average pressure is obtained by statistically averaging the relevant pressure data sequences, and the number of impacts is obtained by detecting and counting the number of peak values ​​exceeding a preset threshold in the pressure sequence. Simultaneously, the system extracts the heart rate cycle signal and blood oxygen saturation data of the corresponding target body region marked as active from the physiological data segment of the personal training profile, and calculates the average heart rate and blood oxygen saturation change rate during this period. The average heart rate is obtained by converting all heart rate cycles during the active period into heart rate values ​​and then averaging them, and the blood oxygen saturation change rate is obtained by calculating the difference between the blood oxygen saturation values ​​at the beginning and end of the active period and dividing by the duration.

[0097] A regional load quantification model is established, which uses the activation intensity index, mean pressure, number of impacts, mean heart rate, and rate of change of blood oxygen saturation of the targeted body region as input parameters. The regional load quantification model calculates a dimensionless comprehensive value using a weighted fusion algorithm. One expression of the weighted fusion algorithm is as follows:

[0098]

[0099] in: This represents the area load value to be calculated. Indicates the activation intensity index. This represents the normalized average pressure. This represents the number of impacts after normalization. This represents the normalized mean heart rate. This indicates the rate of change in blood oxygen saturation. These are pre-defined weighting coefficients determined through data analysis, with a total value of 1. The calculated composite value... This is quantified as the regional load value of the targeted body region.

[0100] Table 1: Time Series Table of Load Values ​​in the "Chest" Region of Users

[0101] date Training type Regional load value (Q) 2024-01-02 bench press 72.3 2024-01-05 bench press 75.1 2024-01-09 bench press 78.5 2024-01-12 bench press 76.0 2024-01-16 bench press 80.2

[0102] In some embodiments, the regional load quantification model calculates example data as follows: For the "chest" region, the input parameters include an activation intensity index of 0.92, average pressure of 25.3 kPa, number of impacts of 85, average heart rate of 142 beats / minute, and blood oxygen saturation change rate of -0.8% / minute. The average pressure, number of impacts, and average heart rate are converted to a 0-1 scale using a preset normalization function and then substituted into the regional load quantification model formula, resulting in a final output regional load value of 80.2. For the "left forearm" region, the input parameters are different, but after the same weighted fusion calculation, a regional load value of 68.7 may be obtained.

[0103] Understandably, the system retrieves stored user training archives to deduce the load evolution path. It searches the user database at the remote computing center for recently completed personal training archives generated from several training sessions of the same or similar type as the current training. The search criteria are the user identifier and the training type code. From these historical personal training archives, the system extracts the region load values ​​for the same target body region, calculated using the same method in historical training. These region load values ​​are then arranged in chronological order of training date and time, forming a time series of region load values ​​for that target body region, as shown in the example above.

[0104] Optionally, a trend analysis tool can be applied to process the time series of regional load values ​​for each targeted body region. This tool can use linear regression to fit a trend line, with the slope indicating whether the overall trend of load change is upward, downward, or stable. Based on the specific type and intensity of each training session, the system marks key inflection points on the trend line. These inflection points are dates when regional load values ​​show a significant increase or decrease, thus plotting the load evolution path reflecting the adaptation and change process of each targeted body region. The load evolution path can be visually displayed in a graph, essentially a curve showing the regional load value changing over time, accompanied by a trend line and inflection point markers.

[0105] In some embodiments, the load values ​​of the regions extracted from historical personal training records are compared with the current values. In four historical bench press training sessions extracted from the database, the load values ​​for the "chest" region were 72.3, 75.1, 76.0, and 78.5, respectively, while the current calculated value is 80.2. These five values ​​are arranged in chronological order to form a time series. Applying linear regression to this time series yields a positive trend line slope, indicating that the load in the "chest" region generally shows an upward trend. Simultaneously, the increase from 76.0 to 78.5 is observed to be greater than the increase between other adjacent points in the series, and the system may mark this date as a key intensity adaptation inflection point. For the "left forearm" region, its historical and current load value series may show different trends, for example, values ​​fluctuating between 65 and 70, with a trend line slope close to zero, indicating that the load remains stable.

[0106] See Figure 4 This is a comparison chart of activation intensity and stress distribution across different target body regions during sports and fitness training. This chart, output from the analysis module of a sports and fitness training cloud data acquisition system, is used to identify the "activation-stress" matching relationship between various body regions during training, determining whether the load distribution of training movements is reasonable. It also helps to locate areas of "high activation but low stress" (such as the core) or "low activation but high stress," providing a basis for adjusting subsequent training plans. By comparing the two indicators of "activation intensity and stress," the load characteristics of each region can be quickly identified, clarifying whether the training movements provide balanced stimulation to different body regions. The chart visually demonstrates that the chest is the core load area for this training, providing a basis for the subsequent "primary-secondary" exercise arrangement.

[0107] In one embodiment of the invention, the system obtains a load reference boundary set for the user's physical condition from the user's personal profile. This load reference boundary is a set of values ​​pre-calculated based on the user's age, gender, training experience, and health assessment data, or set by the coach. The load reference boundary includes a safe upper limit, an effective stimulation lower limit, and an optimal range. For the "chest" targeted body region, the safe upper limit might be set to a regional load value of 85.0, the effective stimulation lower limit to 65.0, and the optimal range to 70.0 to 80.0. The system compares the current regional load value and the latest trend of its load evolution path for each targeted body region with the load reference boundary corresponding to that targeted body region stored in the database. This comparison involves numerical comparison and trend analysis.

[0108] In some embodiments, for a target body region whose regional load value is consistently below the effective stimulus lower limit and whose trend is flat, the system adds an isolation training exercise for that target body region in subsequent training programs and increases the number of training sets and the load weight. For example, the current regional load value of the "left forearm" target body region is 62.3, which is lower than its effective stimulus lower limit of 65.0, and its load evolution path shows that the regional load value of the past three training sessions fluctuates narrowly between 60.5 and 62.8, with the slope of the trend line close to zero. In this case, the system may add "dumbbell wrist curls" as an isolation training exercise in subsequent customized training programs and increase the preset number of sets for this exercise from 3 to 5, and increase the recommended load weight from 5 kg to 7 kg. For targeted body regions where the load value is close to or exceeds the safe upper limit, or where the load evolution path shows an excessively rapid increase, the system will reduce the training priority of these regions in subsequent training programs, arrange compound exercises, or introduce active recovery sessions. For example, if the current load value of the "chest" targeted body region is 80.2, which is close to its safe upper limit of 85.0, and the load evolution path shows that the load value of the region jumped from 76.0 to 80.2 in the last two training sessions, with a growth rate exceeding the historical average, the system may reduce the training priority of directly impacting the chest with "flat bench press" in subsequent plans, and instead arrange compound exercises such as "dumbbell incline bench press" that can distribute some of the load to the shoulders, or add "foam roller chest relaxation" as an active recovery session to the training day schedule.

[0109] Understandably, the system integrates the above-mentioned programming decisions based on comparison and analysis, specifically specifying the training items for the next training cycle, the number of repetitions and sets for each item, the rest time between sets, and the recommended intensity. The integration process, based on training science principles, coordinates programming decisions targeting different body regions into a coherent overall plan, avoiding conflicts and overtraining. Finally, the system generates a detailed customized training plan document in a structured text format, including fields such as training date, target muscle group, exercise name, load weight, repetitions, sets, rest time, and precautions. After generating the customized training plan document, the system pushes it to the user's terminal interface for execution and tracking.

[0110] Optional, dynamic adjustment formula between load reference edges:

[0111]

[0112] in: This represents the adjusted boundary value. This represents the baseline boundary value based on the user's physical condition. Indicates the adjustment factor. This represents the recent moving average of the load value in the targeted body region. This represents a longer-term moving average of the load values ​​for the targeted body region. This adjustment allows the load reference boundary to adapt to changes in the user's long-term adaptation level. The system quantifies the comparison between the regional load values ​​and the load reference boundary as a deviation index. Used to guide the intensity of arrangement. ,in This is the current regional load value. It is the median of the optimal interval range. This is the width of the optimal interval range. Deviation index. The larger the value, the further the current load deviates from the ideal range, and the greater the adjustment range in the scheduling decision may be.

[0113] In some embodiments, the comparison process and the scheduling decision involve data comparison. The system reads the current load value of the "chest" target body region as 80.2, with a safe upper limit of 85.0 and an optimal range of 70.0-80.0. Therefore, the current value is at the upper edge of the optimal range. At the same time, the load evolution path shows that it has increased too rapidly recently. Based on this information, the system makes scheduling decisions to "reduce training priority" and "incorporate compound movements." Specifically, in the customized training plan document, the "flat barbell bench press" for the next "chest" training day is changed from a core movement to a secondary movement, the number of sets is reduced from 4 to 3, and the "dumbbell incline bench press" is added as a core movement. For the "left forearm" targeted body region, the current load value of 62.3 is lower than the effective stimulation lower limit of 65.0, and the trend is flat. The system makes the programming decisions to "increase isolation exercises" and "increase the number of sets and load". A "barbell wrist curl with overhand grip" exercise is added to the "arms" training day section of the customized training plan document, set to 4 sets of 12 repetitions each, with a recommended load weight of 20 kg. All independent programming decisions for different targeted body regions are integrated by the system. After checking for time conflicts or excessive muscle group training frequency, a unified weekly training plan document is generated.

[0114] See Figure 5 This is a bar chart showing the distribution of activation intensity indices for various targeted body regions during sports and fitness training. This chart is output from the region activation analysis module of the fitness training cloud data acquisition system. It is used to clarify the stimulation priority of different body regions by training movements, determine the targeting accuracy of the training type, and identify areas with low activation intensity, providing direction for adjusting training movements. It quickly identifies the core stimulation areas of this training session, verifying the "target region coverage" of the training plan. If the training goal is to strengthen the core and trunk, this chart can directly prove that the training movements have achieved their targeting objectives; if the goal includes strengthening the limbs, it can reveal any targeting deviations in the current training.

[0115] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0116] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for collecting cloud data for sports and fitness training, characterized in that, The process includes the following steps: Initiate a data acquisition command that is compatible with the training type, and drive the sensor network deployed on the user's body surface; The sensor network synchronously captures electrical signals reflecting cardiac activity, inertial signals reflecting spatial changes in the limbs and trunk, and distribution signals reflecting contact pressure. The captured electrical signals, inertial signals, and distributed signals are processed locally in real time to obtain a preliminary processed signal set carrying time stamps; The initial processed signal set is transmitted completely to the remote computing center via a wireless communication network; At the remote computing center, the received preliminary processing signal set is parsed and reassembled to construct a structured personal training profile; Based on the kinematic information contained in an individual's training profile, target body regions that are activated during specific training types are identified; For each targeted body region, the mechanical stimulation and metabolic response it experiences during the training cycle are comprehensively evaluated, and the region load value is quantified. By retrieving stored user training files and comparing historical and current regional load values, the load evolution path can be deduced. Based on the load evolution path and the preset load reference boundary, the components and execution parameters of subsequent training are automatically arranged to form a customized training plan.

2. The method for collecting cloud data for sports and fitness training according to claim 1, characterized in that, The acquisition command, which is adapted to the startup and training type, drives the sensor network deployed on the user's body surface, specifically including: The system receives the type of sports and fitness training selected by the user through the terminal interface. The training type includes strength training, endurance training, or flexibility training. Based on the selected training type, retrieve the corresponding sensor enablement list and parameter initialization settings from the device configuration library; According to the sensor activation list, wake-up and self-test commands are sent to the electrocardiogram recorder, blood oxygen monitoring probe, inertial measurement unit array, and thin-film pressure sensing grid; After confirming that all sensors have returned a ready signal, a synchronization acquisition start pulse is sent to ensure that the ECG recorder starts recording ECG waveforms, the blood oxygen monitoring probe starts acquiring blood oxygen saturation, the inertial measurement unit array starts tracking the angular velocity of the head, torso, and limbs, and the thin-film pressure sensing grid starts monitoring the pressure distribution on the hands and soles of the feet.

3. The method for collecting cloud data for sports and fitness training according to claim 2, characterized in that, The process of locally and instantly processing the captured electrical signals, inertial signals, and distributed signals to obtain a preliminary processed signal set carrying a time stamp specifically includes: In the built-in processor of the electrocardiogram recorder, a filtering program is applied to the raw electrocardiogram waveform to eliminate power frequency interference and electromyographic noise, and to extract a clear heart rate cycle signal. In the local processing unit of the inertial measurement unit array, attitude calculation is performed on the raw angular velocity and acceleration data, and the real-time three-dimensional orientation data of each body segment relative to the global coordinate system is obtained through the quaternion update algorithm. In the edge nodes of the thin-film pressure sensing grid, the original pressure distribution data is aggregated and characterized to identify the pressure center trajectory, peak pressure point, and pressure change curve over time. A time identifier code generated by a high-precision clock source is uniformly added to the processed heart rate cycle signal, real-time three-dimensional orientation data, and pressure characteristic data. The various types of data, after being appended with time identifiers, are packaged and integrated to form the preliminary processing signal set.

4. The method for collecting cloud data for sports and fitness training according to claim 3, characterized in that, The step of transmitting the preliminary processed signal set completely to the remote computing center via a wireless communication network specifically includes: Check the current status of the wireless communication network and prioritize the 5G network channel as the data transmission link; The preliminary processed signal set is encapsulated, and the source device identifier, training session identifier, and data packet sequence number are added to the header of the data packet; Encapsulated data packets are continuously sent to the designated remote computing center Internet Protocol address via a selected 5G network channel in a streaming manner. During transmission, the integrity and timing of data packets are monitored in real time. If packet loss or out-of-order delivery is detected, a data packet retransmission mechanism is triggered to ensure that the initial processing signal set is delivered completely.

5. The method for collecting cloud data for sports and fitness training according to claim 4, characterized in that, The process of parsing and reassembling the received preliminary processing signal set at the remote computing center to construct a structured personal training profile specifically includes: The remote computing center's data receiving service sorts and reassembles the delivered preliminary processing signal set according to the data packet sequence number to restore its original timing. After analyzing the reconstructed data, the heart rate cycle signal, real-time three-dimensional orientation data, and pressure characteristic data are separated into different processing pipelines based on the source device identifier; Guided by time identification codes, data in different processing pipelines are strictly aligned in the time domain to ensure that physiological, posture and stress data at the same moment correspond precisely. The aligned data is organized according to a preset structured format, which includes a data block header, a timeline, physiological data segments, posture data segments, and stress data segments, thereby generating the personal training profile.

6. The method for collecting cloud data for sports and fitness training according to claim 5, characterized in that, The method of identifying target body regions activated during specific training types based on kinematic information contained in an individual's training profile includes: Real-time three-dimensional orientation data of each body segment throughout the training process are extracted from the posture data segment of the individual training file. Calculate the motion trajectory, average amplitude, and frequency of each body segment in three-dimensional space; The calculated motion trajectory, average motion amplitude, and motion frequency are compared with a predefined training type action feature template library. Based on the comparison results, body segments whose motion features match the selected training type's motion feature template with a set threshold are selected. These body segments are defined as the target body regions activated during this training, and the activation intensity index of each target body region is recorded.

7. The method for collecting cloud data for sports and fitness training according to claim 6, characterized in that, For each targeted body region, a comprehensive assessment of the mechanical stimulation and metabolic response it experiences during the training cycle is conducted to quantify the region's load value, specifically including: For each targeted body region, extract the relevant stress characteristic data from the stress data segment of the individual training file, and calculate the average pressure and number of impacts experienced by each targeted body region. Extract heart rate cycle signals and blood oxygen saturation data for the corresponding time period from the physiological data segments of the individual training file, and calculate the average heart rate and blood oxygen saturation change rate during the active period of each targeted body area; A regional load quantification model was established, using the activation intensity index, average pressure, number of impacts, average heart rate, and rate of change of blood oxygen saturation of the targeted body region as input parameters. The regional load quantification model calculates a dimensionless comprehensive value through a weighted fusion algorithm, which is then quantified as the regional load value of the target body region.

8. The method for collecting cloud data for sports and fitness training according to claim 7, characterized in that, The process of accessing stored user training data, comparing historical and current regional load values, and deriving the load evolution path specifically includes: Retrieve the user's personal training profile generated from several recent training sessions from the user database at the remote computing center; Extract the region load values ​​calculated from the same target body regions in historical training from these historical individual training records; Arrange the current and historical regional load values ​​in chronological order to form a time series of regional load values ​​for each targeted body region; Trend analysis tools were used to process the time series of regional load values ​​for each targeted body region to identify whether the overall trend of load changes was increasing, decreasing, or remaining stable. By combining the specific type and intensity of each training session, key turning points in the trend are marked, thereby drawing a load evolution path that reflects the adaptation and change process of each targeted body region.

9. The method for collecting cloud data for sports and fitness training according to claim 8, characterized in that, The step of automatically arranging the components and execution parameters of subsequent training based on the load evolution path and the preset load reference boundary to form a customized training plan specifically includes: Obtain the load reference boundary set for the user's physical fitness level. The load reference boundary includes a safe upper limit, an effective stimulation lower limit, and an optimal range. Compare the current regional load value and the latest trend of the load evolution path of each targeted body region with the corresponding load reference boundary; For targeted body regions where the regional load value is consistently below the effective stimulus lower limit and the trend is flat, add isolated training movements for the targeted body regions in subsequent training programs, and increase the number of training sets and load weight. For target body regions whose regional load values ​​are close to or exceed the safety limit, or whose evolution path shows excessively rapid growth, their training priority should be reduced in subsequent training programs, and compound movements or active recovery phases should be introduced. By integrating the above scheduling decisions, specific training items for subsequent training cycles, the number of repetitions and sets for each item, rest time between sets, and suggested intensity are specified, generating a detailed customized training plan document.

10. A cloud data acquisition system for sports and fitness training, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the sports and fitness training cloud data acquisition method according to any one of claims 1 to 9.

Images