Full-chain fat burning fitness system based on micro-unstable plane and adjustable damping technology
By integrating data acquisition, feature extraction, and neural network models into the fitness system, and combining exercise status and electrocardiogram signals, the system achieves accurate identification and dynamic adaptation of users' exercise patterns and energy consumption, solving the problem of difficulty in achieving accurate identification and guidance in existing systems and improving training effectiveness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 山西医科大学第二医院(山西医科大学第二临床医学院)
- Filing Date
- 2026-03-09
- Publication Date
- 2026-04-17
AI Technical Summary
Existing fitness systems struggle to integrate and analyze multi-dimensional data, including time-series data of a user's movement on a slightly unstable plane, electrocardiogram signals, and personal vital signs. This makes it difficult to accurately identify movement patterns, assess energy consumption, and provide exercise adaptation guidance, thus hindering the effective achievement of training goals.
The system uses a data acquisition module to obtain time-series data of motion status and electrocardiogram signals. Through a motion feature extraction module, an identification and evaluation module, and a motion adaptation evaluation module, it uses a motion neural network model to identify motion patterns and assess energy consumption. Combined with muscle fatigue assessment values, it provides motion adjustment prompts.
It enables accurate identification and dynamic adaptation guidance of users' exercise patterns and energy consumption, ensuring the effective achievement of training goals and avoiding the problem of mismatch between exercise intensity and physical endurance.
Smart Images
Figure CN121885091A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of regulation and control technology, specifically to a full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology. Background Technology
[0002] With the accelerating pace of modern urban life and the increasing health awareness of the public, there is a growing demand for efficient, convenient, and suitable fitness methods for small indoor spaces. The fitness industry is developing towards the cross-integration of exercise physiology, mechanical dynamics, and ergonomics. In-place fitness systems are suitable for various scenarios such as home and office because they do not require a large exercise space.
[0003] In-place exercises, with their controllable movements and minimal body displacement, can reduce the impact load on joints by adjusting the range of motion and force application, making them suitable for people with different fitness levels. Meanwhile, training methods on a slightly unstable plane can effectively activate deep core stabilizing muscle groups such as the transverse abdominis and multifidus, improving muscle engagement and spinal stability, allowing basic in-place exercises to have the dual value of core strengthening and physical training.
[0004] The development of adjustable damping technology has provided technical support for resistance control in stationary fitness systems, enabling fitness equipment to adapt to different needs from basic physical fitness activation to high-intensity explosive power training. At the same time, combined with the synergistic application of elastic resistance components, it can realize the linkage training of the upper and lower limbs, simulate the mechanical transmission of natural functional movements such as skiing and climbing, and improve the training effectiveness of stationary exercises.
[0005] The limitations of existing technologies include at least the following problems: Existing technologies struggle to integrate and analyze multi-dimensional data, such as time-series data of the user's motion state on a slightly unstable plane, electrocardiogram signals, and personal vital signs. This makes it difficult to simultaneously complete the entire process of motion pattern recognition, accurate energy consumption assessment, exercise suitability judgment, and exercise adjustment prompts based on the analysis results. Consequently, fitness systems struggle to accurately identify the user's actual motion state, physical condition, and personal vital signs, and to provide dynamic adaptation guidance during the exercise process. This can easily lead to a mismatch between the user's exercise intensity and their physical tolerance, making it difficult to ensure the effective achievement of training goals such as fat burning, and also making it difficult to provide appropriate exercise adjustment suggestions based on the user's real-time motion state and physical condition, resulting in a lack of precise support for fitness training. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides a full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology, which solves the problem that existing technologies are unable to comprehensively analyze exercise data, resulting in difficulties in achieving accurate identification and dynamic adaptation guidance.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a full-chain fat-burning fitness system based on a micro-unstable plane and adjustable damping technology, comprising: a data acquisition module for acquiring time-series data of the user's motion state on a micro-unstable plane and electrocardiogram signals within a set period; a motion feature extraction module for extracting the user's motion feature dataset based on the motion state time-series data; an identification and evaluation module for obtaining the user's vital signs information based on the motion feature dataset and inputting it into a preset motion neural network model to determine the user's exercise mode and energy consumption assessment value; an exercise adaptation assessment module for extracting the user's exercise adaptation assessment value based on the energy consumption assessment value and electrocardiogram signals; and a feedback output module for sending the exercise mode and exercise adaptation assessment value to the user terminal for display and providing exercise adjustment prompts.
[0008] Furthermore, the motion state timing data includes triaxial acceleration signals, triaxial angular velocity signals, damping adjustment level signals, and resistance band stretching displacement signals.
[0009] Furthermore, the specific steps for extracting the user's motion feature dataset are as follows: preprocess the motion state time series data; perform differential feature extraction processing on the preprocessed motion state time series data to obtain the user's motion feature dataset, including motion frequency, motion intensity, damping coefficient, and resistance band tension.
[0010] Furthermore, the specific steps for differential feature extraction are as follows: perform a fast Fourier transform on the triaxial acceleration signal to extract the motion frequency; perform time-domain integration on the triaxial acceleration signal to obtain the motion intensity; determine the damping coefficient based on the preset mapping relationship corresponding to the damping adjustment gear signal; and calculate the resistance band tension based on the resistance band tension displacement signal and in combination with the preset resistance band elastic coefficient.
[0011] Furthermore, the motion neural network model includes an input layer, a motion recognition layer, and an energy consumption assessment layer. The specific steps for determining the user's motion pattern and energy consumption assessment value are as follows: In the input layer, a fused feature vector of the user is constructed based on the motion feature dataset and vital sign information; in the motion recognition layer, the fused feature vector is subjected to classification mapping processing to generate the user's motion pattern; in the energy consumption assessment layer, the fused feature vector is subjected to regression mapping processing to generate the user's energy consumption assessment value.
[0012] Furthermore, the specific steps of the classification mapping process are as follows: perform nonlinear transformation processing on the fused feature vector to generate the user's classification feature vector; perform probability mapping processing on the classification feature vector to generate the user's recognition probability set; and perform category decision processing on the recognition probability set to generate the user's motion pattern.
[0013] Furthermore, the specific steps of the regression mapping process are as follows: attention weight generation processing is performed on the fused feature vector to generate the user's weight vector; the weight vector is fused with the fused feature vector to generate the user's energy consumption assessment value.
[0014] Further, the specific steps for extracting the user's exercise fit assessment value are as follows: Based on the electrocardiogram signal, extract the user's muscle fatigue assessment value; based on the muscle fatigue assessment value, determine the corresponding recommended energy consumption threshold from the preset energy consumption threshold rule base; use the ratio of the energy consumption assessment value to the recommended energy consumption threshold as the user's exercise fit assessment value.
[0015] Further, the specific steps for extracting the user's muscle fatigue assessment value are as follows: perform heart rate variability analysis on the electrocardiogram signal to extract the user's electromyographic fatigue assessment set, including heart rate fluctuation assessment value, electrocardiogram complex assessment value, and sympathetic balance assessment value; input the electromyographic fatigue assessment set into the pre-established fatigue assessment model to analyze the user's muscle fatigue assessment value.
[0016] Furthermore, the specific steps for motion adjustment prompts are as follows: compare the motion adaptation assessment value with the preset ideal adaptation range; based on the motion mode and the comparison result, match the corresponding prompt template from the preset prompt rule library to generate motion adjustment prompts.
[0017] The present invention has the following beneficial effects:
[0018] (1) The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology acquires the user's movement state time-series data and electrocardiogram signal on the micro-unstable plane through the data acquisition module. The movement feature extraction module can accurately extract core movement features such as movement frequency and movement intensity from the time-series data. The identification and evaluation module, combined with the user's vital signs information, completes the movement pattern identification and energy consumption evaluation through the movement neural network model. The movement adaptation evaluation module can also extract muscle fatigue evaluation value based on electrocardiogram signal, and then analyze the movement adaptation evaluation value that fits the user's physical condition. By deeply combining the above data, it can accurately determine whether the user's current movement mode matches their own state, and accurately evaluate the fit between actual energy consumption and physical tolerance. This allows the user to clearly know the matching status between their movement state and their body, avoid blind training, and improve the overall accuracy.
[0019] (2) This full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology combines the exercise feature dataset with the user's physical characteristics information to construct a fusion feature vector that fits the individual. Then, through a motion neural network model that includes an input layer, an exercise recognition layer, and an energy consumption assessment layer, it completes classification mapping and regression mapping respectively, and realizes exercise pattern recognition and energy consumption assessment simultaneously. At the same time, in the energy consumption assessment, attention weight generation processing is also used to reasonably reflect the influence of different exercise features on energy consumption. In this way, the user's exercise pattern can be accurately identified, and the energy consumption assessment value can be accurately matched with the user's own situation. This provides a reliable basis for subsequent exercise adaptation assessment and exercise adjustment prompts, so that the adjustment prompts are based on the user's actual situation.
[0020] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description
[0021] Figure 1 This is a block diagram of the full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology of the present invention.
[0022] Figure 2 This is a flowchart of the steps in the full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology of the present invention. Detailed Implementation
[0023] Please see Figure 1 This invention provides a technical solution: a full-chain fat-burning fitness system based on a micro-unstable plane and adjustable damping technology, comprising: a data acquisition module for acquiring time-series data of the user's motion state on a micro-unstable plane and electrocardiogram signals within a set period; a motion feature extraction module for extracting the user's motion feature dataset based on the motion state time-series data; an identification and evaluation module for obtaining the user's vital signs information based on the motion feature dataset and inputting it into a preset motion neural network model to determine the user's exercise mode and energy consumption assessment value; an exercise adaptation assessment module for extracting the user's exercise adaptation assessment value based on the energy consumption assessment value and electrocardiogram signals; and a feedback output module for sending the exercise mode and exercise adaptation assessment value to the user terminal for display and providing (corresponding) exercise adjustment prompts.
[0024] The motion state timing data includes triaxial acceleration signals, triaxial angular velocity signals, damping adjustment level signals, and resistance band stretching displacement signals.
[0025] Specifically, the steps for extracting the user's motion feature dataset are as follows: Preprocessing the motion state time-series data includes: applying Kalman filtering to the triaxial acceleration and angular velocity signals to eliminate sensor random noise and temperature drift; smoothing the damping adjustment signal and resistance band tension displacement signal using a sliding window mean filter to remove instantaneous jitter; aligning the timestamps of all filtered signals to ensure they are on the same time reference; and removing the DC component from the acceleration and angular velocity signals to eliminate gravity components and bias effects; and performing differential feature extraction on the preprocessed motion state time-series data to obtain the user's motion feature dataset, including motion frequency, motion intensity, damping coefficient, and resistance band tension.
[0026] The specific steps of differential feature extraction are as follows: Perform fast Fourier transform on the triaxial acceleration signal to extract the motion frequency. Specifically, synthesize the components of the preprocessed triaxial acceleration signal in the three directions, calculate the resultant acceleration value at each moment, and then convert the resultant acceleration value sequence over a continuous time period from the time domain to the frequency domain. Use the fast Fourier transform algorithm to calculate the energy distribution of different frequency components in the signal to form a spectrum. Search for the frequency component with the highest energy in the spectrum, that is, the frequency point with the largest amplitude, and determine the value of this frequency point as the current motion frequency.
[0027] The motion intensity is obtained by performing time-domain integration on the triaxial acceleration signal. Specifically, the absolute values of the components of the preprocessed triaxial acceleration signal in the three directions are taken to eliminate the influence of positive and negative directions. Within a fixed time window, the absolute values in the three directions are summed to obtain the motion intensity components in each of the three directions. The motion intensity components in the three directions are combined, and the motion intensity is calculated by taking the square root of the sum of squares. The larger the value, the more intense the user's motion.
[0028] Based on the preset mapping relationship corresponding to the damping adjustment gear signal, the damping coefficient is determined. Specifically, a reference table is pre-established in the system memory. This table records the damping coefficient values corresponding to each gear position from the minimum to the maximum gear. This reference table is obtained by calibrating the equipment before it leaves the factory. In the experiment, the damping knob is rotated to each gear, and the pulling force required to push the platform at that gear is measured with a force measuring instrument. The pulling force value is recorded as the damping coefficient of that gear in the table. When the Hall sensor detects the gear value currently adjusted by the user, the system directly looks up the corresponding damping coefficient in the reference table. If the detected gear is between two calibrated gears, the values of the two adjacent gears are used to perform a proportional calculation to obtain the accurate damping coefficient corresponding to the current gear.
[0029] Based on the resistance band's tensile displacement signal and combined with the preset resistance band elastic coefficient, the resistance band tension is calculated. Specifically, different elastic coefficients are preset for different resistance band models: a smaller elastic coefficient for lightweight resistance bands, a medium elastic coefficient for medium-weight resistance bands, and a larger elastic coefficient for heavyweight resistance bands. These elastic coefficients are obtained by conducting tensile tests on each specification of resistance band. During the test, the tensile force generated at different tensile lengths is measured, and the increase in tensile force per unit tensile length is calculated as the elastic coefficient. Before use, the user selects the currently installed resistance band model through a mobile application, and the system automatically retrieves the corresponding elastic coefficient value. The real-time tensile displacement is continuously measured by a displacement sensor. The system multiplies the current tensile displacement value by the preset elastic coefficient to obtain the real-time tension value generated by the resistance band at the current moment, i.e., the resistance band tension.
[0030] In this implementation plan, the preprocessing stage employs appropriate filtering methods for different signals, and also completes timestamp alignment and DC component removal processing to effectively eliminate various interference factors, making the original signal more closely resemble the actual motion state. During differential feature extraction, corresponding extraction methods are designed based on signal characteristics. After synthesizing the acceleration signal, frequency domain conversion is performed to extract the motion frequency. The motion intensity is obtained through multi-dimensional synthesis calculation. The damping coefficient is determined according to the factory-calibrated reference table, and the tension is calculated by combining the elastic coefficient of the resistance band calibration. This allows the extracted feature data to accurately reflect the user's actual motion state, making the results of subsequent motion pattern recognition and energy consumption assessment more closely resemble the user's actual training situation.
[0031] Specifically, the motion neural network model includes an input layer, a motion recognition layer, and an energy consumption assessment layer. The specific steps for determining the user's motion pattern and energy consumption assessment value are as follows:
[0032] In the input layer, a user's fusion feature vector is constructed based on the motion feature dataset and physical characteristics information (including age, weight, height, and gender). Specifically, the parameters in the motion feature dataset are normalized. The motion feature dataset includes motion frequency, motion intensity, damping coefficient, and resistance band tension. These parameters have different numerical ranges. By normalization, they are mapped to a unified interval, such as [0, 1], to eliminate the influence of dimensions. The normalization method can be min-max standardization or Z-score standardization.
[0033] The vital signs information is processed. This information includes age, weight, height, and gender. Age, weight, and height are continuous numerical values and are normalized accordingly. Gender is a categorical feature, which is converted into a binary vector using one-hot encoding; for example, males are represented as [1, 0] and females as [0, 1].
[0034] The normalized motion feature parameters, normalized age, weight, height, and gender one-hot encoded vectors are concatenated in a preset order to form a one-dimensional vector, which is the fusion feature vector. This vector integrates the user's motion state and physical characteristics.
[0035] In the motion recognition layer, the fused feature vectors are classified and mapped to generate the user's motion pattern; in the energy consumption assessment layer, the fused feature vectors are regressed and mapped to generate the user's energy consumption assessment value.
[0036] The specific steps of classification mapping are as follows:
[0037] The fused feature vector is subjected to a nonlinear transformation to generate a user's classification feature vector. Specifically, the fused feature vector is input into the first fully connected hidden layer of the motion recognition layer. This hidden layer contains multiple neurons. Each neuron performs a linear weighted summation of the input vector, and then transforms it through a nonlinear activation function (such as the ReLU function), outputting a nonlinearly transformed vector. This vector serves as the output of the first hidden layer. This vector can then pass through multiple similar hidden layers, each undergoing linear transformation and nonlinear activation, progressively extracting high-order abstract features related to motion pattern discrimination. After passing through all hidden layers, a final vector is output, which is the classification feature vector. This vector condenses the discriminative information used to distinguish different motion patterns.
[0038] The classification feature vector is subjected to probability mapping to generate the user's recognition probability set. Specifically, the classification feature vector is input to the output layer of the motion recognition layer. The output layer uses the softmax activation function to transform the classification feature vector into a probability distribution vector. The softmax function maps any real number vector to a positive real number vector whose sum of all components is 1. Each component corresponds to the probability of a preset motion mode category. The preset motion mode categories include at least the stationary gliding mode, the mountain climbing mode, and the resistance band coordinated action mode. This probability distribution vector is the recognition probability set, which reflects the likelihood of the current motion state belonging to each category.
[0039] The recognition probability set is processed by category decision to generate the user's motion pattern. Specifically, the component with the highest probability value is found from the recognition probability set, and the motion pattern category corresponding to the component is determined as the user's current motion pattern. For example, if the probability corresponding to the stationary gliding pattern is the highest, it is determined that the user is performing stationary gliding. This result is the motion pattern output by the motion recognition layer.
[0040] The specific steps of the regression mapping process are as follows: Attention weight generation is performed on the fused feature vector to generate the user's weight vector. Specifically, the fused feature vector is input into a sub-network of the energy consumption evaluation layer. This sub-network consists of a fully connected layer and a softmax activation function. Specifically, the fused feature vector undergoes a linear transformation through a fully connected layer to obtain the original weight scores with the same dimension as the fused feature vector. These scores are then normalized using the softmax function so that the sum of all weights is 1, thereby generating a weight vector. Each element in this weight vector represents the importance of the corresponding feature dimension in the fused feature vector, i.e., the attention weight. In this way, the network can automatically learn which features are more critical to energy consumption evaluation and assign them higher weights.
[0041] The weight vector and the fused feature vector are fused to generate the user's energy consumption assessment value. Specifically, the weight vector and the fused feature vector are multiplied element-wise to obtain a weighted feature vector. That is, the original features are weighted according to their importance. This operation achieves selective emphasis on features, so that important features play a greater role in subsequent calculations. The weighted feature vector is input into the subsequent fully connected hidden layer of the energy consumption assessment layer. After several layers of nonlinear transformation, it is finally mapped to a continuous value through the output layer (using a linear activation function), which is the user's energy consumption assessment value.
[0042] The pre-training steps for the motion neural network model are as follows:
[0043] A number of healthy subjects were recruited to undergo various exercise mode tests on the fitness system, including stationary gliding, mountain climbing, and resistance band coordinated movements. Simultaneously, time-series data of the subjects' exercise status (triaxial acceleration signal, triaxial angular velocity signal, damping adjustment level signal, resistance band stretching displacement signal) and vital signs information (age, weight, height, gender) were collected. Professionals labeled each segment of exercise data and recorded the exercise mode category (as the true label for the classification task) and the actual energy consumption value (as the true label for the regression task, obtained through indirect calorimetry or a portable metabolic meter).
[0044] The motion state time series data of each sample is preprocessed and differential features are extracted to obtain a motion feature dataset (including motion frequency, motion intensity, damping coefficient, and resistance band tension). The motion feature parameters and vital signs information are normalized and one-hot encoded and then concatenated to form a fused feature vector, which is used as the input of the model.
[0045] A multi-task neural network model is constructed, which includes an input layer, a shared hidden layer, a motion recognition layer, and an energy consumption evaluation layer. The motion recognition layer uses a softmax activation function to output the probability distribution of motion patterns, and the energy consumption evaluation layer uses a linear activation function to output continuous energy consumption values. A joint loss function is set, which includes a weighted sum of classification loss (cross-entropy loss function) and regression loss (mean squared error loss function). The training objectives of the two tasks are balanced by adjusting the weight coefficients.
[0046] The model is trained under end-to-end supervision by taking the fused feature vector as input and the labeled motion mode category and measured energy consumption value as output. The network weights are iteratively updated using the mini-batch stochastic gradient descent algorithm until the loss function converges. During training, the importance of each input feature to energy consumption assessment is automatically learned to achieve adaptive weighting of features.
[0047] The classification accuracy and energy consumption estimation accuracy of the model are evaluated on an independent validation set. When the model performance reaches the preset standard (e.g., motion pattern recognition accuracy ≥ 94%, energy consumption estimation error ≤ 10%), the trained network structure and weight parameters are solidified and deployed in the system database for motion pattern recognition and energy consumption assessment.
[0048] In this implementation scheme, the input layer processes different types of motion features and vital signs information through normalization and one-hot encoding, and splices them to form a fusion feature vector, which effectively integrates the user's motion state and individual basic information, and eliminates interference factors of various data. The motion recognition layer extracts the core discrimination information of the motion pattern through multi-layer nonlinear transformation, and combines probability mapping and category decision to accurately match the user's current motion pattern. The energy consumption assessment layer learns and highlights the key features for energy consumption assessment through attention weight generation and processing, and then obtains the energy consumption assessment value through weighted fusion and subsequent transformation. This fully combines the user's individual differences and actual motion state to complete the analysis, making the motion pattern recognition result more consistent with the actual training action and the energy consumption assessment result more consistent with the user's own situation.
[0049] Specifically, such as Figure 2 As shown, the specific steps for extracting a user's exercise adaptation assessment value are as follows: Based on the electrocardiogram signal, extract the user's muscle fatigue assessment value; based on the muscle fatigue assessment value, determine the corresponding recommended energy consumption threshold from the preset energy consumption threshold rule base, specifically as follows:
[0050] A number of healthy participants were recruited, covering different ages, genders, and fitness levels. Each participant underwent a progressive overload exercise test on a fitness system. The test started with a low resistance level, increasing by one level every 2 minutes, while participants were required to maintain a stable gliding frequency until they subjectively felt exhausted or fatigued and could not continue.
[0051] The following data were collected simultaneously throughout the entire movement:
[0052] Real-time energy consumption assessment value;
[0053] Real-time muscle fatigue assessment values;
[0054] Subjective fatigue ratings of the subjects (using the Borg CR-10 scale, recorded every 2 minutes);
[0055] Heart rate and blood lactate concentration (measured by collecting fingertip blood every 2 minutes);
[0056] Data from each subject was analyzed to determine the upper limit of safe exercise intensity under different levels of fatigue. The upper limit of safe exercise intensity is defined as the upper limit of exercise intensity that can be sustained without causing excessive fatigue accumulation.
[0057] The specific determination method adopts a comprehensive assessment using multiple indicators:
[0058] Physiological indicator determination: When the blood lactate concentration reaches 4 mmol / L (lactate threshold), the corresponding energy consumption assessment value is used as the upper limit of safe aerobic exercise.
[0059] Subjective feeling assessment: When the subjective fatigue score reaches 5 points (difficulty), the corresponding energy consumption assessment value is used as the upper limit of subjective safety.
[0060] Fatigue feedback determination: When the muscle fatigue assessment value increases by more than 0.3 within 3 consecutive minutes, the energy consumption assessment value at the starting point of the increase is taken as the fatigue threshold.
[0061] The minimum value of the above three judgment results shall be taken as the upper limit of the safe exercise intensity for the subject at that time point;
[0062] All time point data (muscle fatigue assessment value, upper limit of safe exercise intensity) of each subject during the entire exercise process were plotted into a scatter plot, with the horizontal axis representing the muscle fatigue assessment value (0-1) and the vertical axis representing the upper limit of safe exercise intensity (0-1).
[0063] Observing the scatter plot distribution trend, it was found that the upper limit of safe exercise intensity decreased as the muscle fatigue assessment value increased. A piecewise statistical method was used to statistically analyze a large amount of experimental data to obtain the correspondence between muscle fatigue assessment values and recommended energy consumption thresholds, for example:
[0064] When the muscle fatigue assessment value is between 0 and 0.3 (low fatigue level), the average value of the upper limit of safe exercise intensity is 0.9, which means that in a low fatigue state, users can safely maintain exercise energy consumption equivalent to 90% of the maximum intensity, and the corresponding recommended energy consumption threshold is 0.9.
[0065] When the muscle fatigue assessment value is between 0.3 and 0.6 (moderate fatigue level), the average value of the upper limit of safe exercise intensity is 0.7, which means that in a moderate fatigue state, the recommended exercise energy consumption should not exceed 70% of the maximum intensity, and the corresponding recommended energy consumption threshold is 0.7.
[0066] When the muscle fatigue assessment value is between 0.6 and 1.0 (high fatigue level), the average value of the upper limit of safe exercise intensity is 0.4, which means that in a state of high fatigue, the recommended exercise energy consumption should not exceed 40% of the maximum intensity, and the corresponding recommended energy consumption threshold is 0.4.
[0067] The ratio of the energy consumption assessment value to the recommended energy consumption threshold is used as the user's exercise adaptation assessment value.
[0068] The specific steps for extracting a user's muscle fatigue assessment value are as follows:
[0069] Heart rate variability analysis is performed on the electrocardiogram (ECG) signal to extract the user's electromyographic fatigue assessment set, including heart rate fluctuation assessment value, ECG complexity assessment value, and sympathetic balance assessment value. Specifically, the ECG signal is preprocessed by using a bandpass filter (which can be set to 0.5Hz~40Hz) to filter out power frequency interference, electromyographic noise, and baseline drift to obtain a clean ECG waveform. An adaptive threshold detection algorithm or the Pan-Tompkins algorithm is used to identify the position of the R wave in each heartbeat cycle, record the peak time of each R wave, and calculate the time interval between adjacent R waves to obtain the RR interval sequence.
[0070] After integrating the RR interval sequence, it is divided into equal time windows of different lengths. A local trend is fitted within each time window, and then the double logarithmic relationship between the detrended fluctuation function and the time window length is calculated to obtain a scaling index. This scaling index reflects the long-range correlation and self-similarity characteristics of heart rate fluctuations, that is, whether the fluctuation pattern of heart rate in the time series is regular. Under healthy conditions, heart rate fluctuations show moderate long-range correlation, and the scaling index is usually between 0.75 and 1.25. When muscle fatigue accumulates, the heart rate regulation capacity decreases, the fluctuation pattern changes, and the scaling index deviates from the healthy range. The degree of deviation of the calculated scaling index from the healthy baseline value (which can be set to 1) is used as the heart rate fluctuation assessment value.
[0071] Sample entropy is calculated for RR interval sequences. A pattern dimension *m* and a similarity tolerance *r* are set. The logarithms of the patterns matching *m* and *m+1* points in the sequence are then counted. Finally, the negative natural logarithm of the conditional probability is calculated. A higher sample entropy value indicates greater complexity of heart rate fluctuations, meaning more flexible regulation of the heart rhythm; a lower sample entropy value indicates more regular heart rate fluctuations and reduced complexity. When muscle fatigue accumulates, the autonomic nervous system's regulatory function of the heart is limited, leading to a decrease in the complexity of heart rate fluctuations and a lower sample entropy value. The calculated sample entropy value is used as an assessment value for electrocardiographic complexity.
[0072] The RR interval sequence was transformed from the time domain to the frequency domain using Fast Fourier Transform (FFT) to obtain the power spectral density. Low-frequency power (LF, 0.04-0.15 Hz) and high-frequency power (HF, 0.15-0.4 Hz) were extracted from the power spectrum, and the LF / HF ratio was calculated. This ratio reflects the balance between the sympathetic and parasympathetic nervous systems. An increased ratio indicates sympathetic dominance, while a decreased ratio indicates parasympathetic dominance. Exercise intensity was introduced as a correction factor. Specifically, the LF / HF ratio was divided by the normalized value of the current exercise intensity (exercise intensity ranges from 0 to 1) to obtain the corrected sympathetic balance assessment value. The higher the value, the more dominant the sympathetic nervous system is under the same exercise intensity, indicating a higher degree of muscle fatigue.
[0073] The above evaluation values were then normalized to obtain heart rate fluctuation evaluation values, ECG complexity evaluation values, and sympathetic balance evaluation values, which are between 0 and 1.
[0074] The electromyographic fatigue assessment set is input into a pre-established fatigue assessment model to analyze the user's muscle fatigue assessment values. The fatigue assessment model is as follows:
[0075] ;
[0076] in, For the user's muscle fatigue assessment value, Assess the user's heart rate fluctuation value. The heart rate variability adjustment coefficient is stored in the database. For the user's complex ECG assessment values, These are the complex regulation coefficients of electrocardiogram stored in the database. The user's sympathetic balance assessment value. The sympathetic balance adjustment coefficients are stored in the database. ;
[0077] It should be noted that, , , The steps to obtain it are as follows:
[0078] Participants were recruited to perform exercises of varying intensities on a fitness system, with simultaneous electrocardiogram (ECG) signals collected. Participants reported their current level of fatigue based on a subjective fatigue perception scale (Borg score). The collected ECG signals were preprocessed and feature extracted to obtain heart rate fluctuation assessment values, ECG complexity assessment values, and sympathetic balance assessment values for each time window. The three assessment values were normalized, and the ECG complexity assessment value was reciprocally transformed to make it positively correlated with the level of fatigue. The three processed feature values and their corresponding subjective fatigue scores were combined to form a sample, and a large number of samples were collected to form a training dataset.
[0079] Using three feature values from the training dataset as independent variables and subjective fatigue scores as dependent variables, a multiple linear regression model was established. The least squares method was used to fit the model, and the regression coefficients corresponding to the three feature values were solved. The absolute values of the three regression coefficients were divided by the sum of their absolute values to obtain three normalized regulation coefficients, and the sum of the three was 1. These three regulation coefficients correspond to the heart rate variability regulation coefficient, the electrocardiogram complex regulation coefficient, and the sympathetic balance regulation coefficient, respectively.
[0080] The calculated three adjustment coefficients are substituted into the fatigue assessment model, and the model's predictive performance is tested on an independent validation dataset to verify its accuracy and reliability. After successful validation, the three adjustment coefficients are stored in the system database.
[0081] In this implementation plan, physiological data from different groups during exercise is collected to determine the correlation between muscle fatigue and safe exercise intensity, forming a scientific and reliable energy consumption threshold rule base. Then, multiple physiological features are extracted based on electrocardiogram signals, and a specially constructed model is used to calculate muscle fatigue assessment values, ensuring that the assessment results are highly consistent with the user's actual fatigue experience. Based on the fatigue assessment results, a suitable recommended energy consumption threshold is determined, and the real-time energy consumption is compared with the recommended threshold to obtain the exercise adaptation assessment value. This makes the assessment results closer to the user's actual physical load, ensuring that the training intensity is always coordinated with the user's current physical state.
[0082] Specifically, the steps for providing exercise adjustment prompts are as follows: The exercise fit assessment value is compared with the preset ideal fit range. Specifically, this involves reading the user's currently set preset training goals, which include at least one of the following: fat-burning mode, rehabilitation mode, endurance improvement mode, or HIIT mode. Different training goals correspond to different exercise fit assessment threshold ranges, which are pre-stored in the system database.
[0083] For example, the ideal adaptation range for the fat-burning mode is 0.9~1.0, the ideal adaptation range for the recovery mode is 0.5~0.8, the ideal adaptation range for the endurance improvement mode is 0.7~1.0, and the HIIT mode sets different threshold ranges according to the high-intensity interval and rest period.
[0084] Compare the exercise fit assessment values with the ideal fit range corresponding to the preset training goals:
[0085] If the exercise fit assessment value is lower than the lower limit of the ideal fit range, it is judged as low intensity;
[0086] If the exercise fit assessment value is within the ideal fit range, the intensity is judged to be appropriate;
[0087] If the exercise fit assessment value is higher than the upper limit of the ideal fit range, it is judged as high intensity;
[0088] Based on the motion pattern and comparison processing results, the corresponding prompt template is matched from the preset prompt rule library to generate motion adjustment prompts, specifically as follows:
[0089] A prompt rule base is pre-built in the database. This rule base stores the corresponding prompt templates in tabular form for different motion modes, different comparison processing results, and different combinations of training objectives. The prompt template is a text statement containing placeholders, which are used to fill in specific adjustment parameters, including at least the damping gear adjustment amount.
[0090] The calculation method for damping level adjustment is as follows: When the judgment result is that the intensity is too low or too high, based on the degree of deviation between the current exercise adaptation assessment value and the boundary of the ideal adaptation range, combined with the system's preset damping adjustment step size, the recommended number of damping levels to be adjusted is calculated. For example, when the exercise adaptation assessment value is 0.7 and the ideal adaptation range of the fat-burning mode is 0.9~1.0, the deviation is 0.2, and the system can calculate that it is recommended to increase the damping by 2 levels according to the preset mapping relationship (such as 1 level of damping for every 0.1 deviation); when the exercise adaptation assessment value is 1.2 and the ideal adaptation range of the rehabilitation mode is 0.5~0.8, the deviation is 0.4, and it is recommended to decrease the damping by 4 levels.
[0091] For example, the suggestion rule base may contain the following rule entries:
[0092] When the exercise mode is stationary gliding mode, the comparison result is low intensity, and the training goal is fat burning mode, the corresponding prompt template is that the current intensity is low, and it is recommended to increase the X level of damping or increase the gliding frequency.
[0093] When the exercise mode is mountain running mode, the comparison result is that the intensity is too high, and the training goal is rehabilitation mode, the corresponding prompt template is that the current intensity is too high and there is a risk of over-fatigue. It is recommended to reduce the X-level damping or switch to the stationary gliding mode.
[0094] When the exercise mode is resistance band coordinated action mode, the comparison result is appropriate intensity, and the training goal is endurance improvement mode, the corresponding prompt template is to maintain the current intensity, the endurance improvement effect is the best, and the current damping setting is reasonable.
[0095] After obtaining the motion pattern and comparison processing results, based on these two results and the user's current preset training goal, a matching prompt template is searched in the prompt rule base, and the calculated damping gear adjustment amount is filled into the template placeholder to generate a complete motion adjustment prompt.
[0096] In this implementation plan, the ideal adaptation range is first retrieved based on the user's selected training goal. Then, the real-time calculated exercise adaptation assessment value is compared with this range to accurately determine the match between the current exercise intensity and the target requirements. Based on a pre-established prompt rule library, combined with the user's current exercise mode and the comparison results, the corresponding prompt content is automatically matched. At the same time, based on the deviation between the assessment value and the ideal range, a reasonable damping adjustment amount is calculated and filled into the prompt template to form a clear and actionable adjustment suggestion. This helps the user adjust the exercise intensity in a timely manner, ensuring that the training process always conforms to the preset goal, while avoiding the physical burden caused by inappropriate intensity and improving the safety of the entire fitness process.
[0097] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.
[0098] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A full-chain fat-burning fitness system based on micro-unstable planes and adjustable damping technology, characterized in that: include: The data acquisition module is used to acquire time-series data of the user's motion state on a slightly unstable plane and electrocardiogram signals within a set period. The motion feature extraction module is used to extract the user's motion feature dataset based on the time-series data of motion state. The identification and evaluation module is used to determine the user's exercise pattern and energy consumption assessment value based on the motion feature dataset and the user's vital signs information, and input them into the preset motion neural network model. The exercise adaptation assessment module is used to extract the user's exercise adaptation assessment value based on energy consumption assessment value and electrocardiogram signal; The feedback output module is used to send the exercise mode and exercise adaptation assessment value to the user terminal for display and to provide exercise adjustment prompts.
2. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 1, characterized in that, The motion state timing data includes triaxial acceleration signals, triaxial angular velocity signals, damping adjustment level signals, and resistance band stretching displacement signals.
3. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 2, characterized in that, The specific steps for extracting the user's motion feature dataset are as follows: Preprocess the motion state time series data; Differential feature extraction is performed on the preprocessed motion state time series data to obtain the user's motion feature dataset, including motion frequency, motion intensity, damping coefficient, and resistance band tension.
4. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 3, characterized in that, The specific steps for differential feature extraction are as follows: Perform a fast Fourier transform on the triaxial acceleration signal to extract the motion frequency; The motion intensity is obtained by integrating the triaxial acceleration signal in the time domain. The damping coefficient is determined based on the preset mapping relationship corresponding to the damping adjustment gear signal. The resistance band tension is calculated based on the resistance band tension displacement signal and the preset resistance band elastic coefficient.
5. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 1, characterized in that, The motion neural network model includes an input layer, a motion recognition layer, and an energy consumption assessment layer. The specific steps for determining the user's motion pattern and energy consumption assessment value are as follows: In the input layer, a fused feature vector of the user is constructed based on the motion feature dataset and vital sign information; In the motion recognition layer, the fused feature vectors are classified and mapped to generate the user's motion pattern; In the energy consumption assessment layer, the fused feature vector is subjected to regression mapping to generate the user's energy consumption assessment value.
6. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 5, characterized in that, The specific steps of classification mapping are as follows: The fused feature vector is subjected to a nonlinear transformation to generate the user's classification feature vector; Perform probability mapping on the classification feature vectors to generate a user recognition probability set; The probability set of identification is processed for category decision-making to generate the user's motion pattern.
7. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 5, characterized in that, The specific steps of regression mapping are as follows: Attention weight generation processing is performed on the fused feature vector to generate the user's weight vector; The weight vector and the fused feature vector are fused together to generate the user's energy consumption assessment value.
8. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 1, characterized in that, The specific steps for extracting a user's motion adaptation assessment values are as follows: Based on electrocardiogram signals, extract the user's muscle fatigue assessment value; Based on muscle fatigue assessment values, the corresponding recommended energy consumption thresholds are determined from a pre-defined energy consumption threshold rule base. The ratio of the energy consumption assessment value to the recommended energy consumption threshold is used as the user's exercise adaptation assessment value.
9. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 8, characterized in that, The specific steps for extracting a user's muscle fatigue assessment value are as follows: Heart rate variability analysis was performed on the electrocardiogram signal to extract the user's electromyographic fatigue assessment set, including heart rate fluctuation assessment value, electrocardiogram complexity assessment value, and sympathetic balance assessment value. The electromyographic fatigue assessment set is input into a pre-established fatigue assessment model to analyze the user's muscle fatigue assessment values.
10. The full-chain fat-burning fitness system based on micro-unstable plane and adjustable damping technology according to claim 1, characterized in that, The specific steps for motion adjustment prompts are as follows: The motion adaptation assessment value is compared with the preset ideal adaptation range; Based on the motion pattern and comparison processing results, the corresponding prompt template is matched from the preset prompt rule library to generate motion adjustment prompts.