Moxibustion temperature control method and system based on multi-sensor fusion

Abstract

The invention belongs to the technical field of information technology, and discloses a moxibustion temperature control method and system based on multi-sensor fusion, and the method comprises the steps: obtaining a biopotential signal and position offset data from an acupoint region through a sensor, processing the biopotential signal through a classification algorithm, and obtaining a classified potential distribution mode; on the basis of the classified potential distribution modes, heat conduction influence data between adjacent acupoints are obtained, the position offset data are integrated through an information fusion algorithm, fused physiological state indexes are determined, and the problem of measurement distortion caused by acupoint movement and equipment offset is solved through multi-source data fusion and dynamic error compensation; and meanwhile, the system constructs an accurate potential distribution model by utilizing a spatial calibration and classification algorithm, and temperature control accuracy reaches a millimeter level in combination with real-time position tracking, so that the integrity and reliability of data are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of information technology, specifically relating to a moxibustion temperature control method and system based on multi-sensor fusion. Background Technology

[0002] Acupoint therapy, as an important field combining traditional medicine and modern technology, has irreplaceable value in promoting human health and alleviating chronic diseases. Through precise intervention in acupoint areas, it can regulate the body's physiological state and improve health. However, current acupoint therapy techniques face many challenges in practical applications, particularly in achieving dynamic and precise temperature control, where significant technical bottlenecks remain. This directly affects the stability of treatment effects and the personalized experience for patients, necessitating innovative technological breakthroughs.

[0003] Current acupoint treatment methods typically rely on single preset parameters or static control strategies for temperature regulation. This makes it difficult to adapt to the dynamic needs of patients during treatment due to changes in their physical condition, environment, or treatment stage. The lack of real-time feedback in temperature control leads to uneven temperature distribution during treatment, with some acupoint areas potentially overheating or underheating, affecting the stability of the treatment effect. A deeper problem lies in the complexity of the physiological state of acupoint areas, involving the interaction of multiple physiological signals, and temperature regulation requires comprehensive consideration of the dynamic changes of these signals. It is impossible to capture and integrate multiple physiological information from acupoint areas in real time, such as bioelectric potential signals and positional offset data, to generate a comprehensive indicator reflecting the patient's current state. In actual treatment, patients may experience slight shifts in acupoint position due to changes in posture or muscle activity, and current technologies often cannot track these changes in real time. The lack of dynamic fusion of multidimensional physiological information makes it impossible to generate precise temperature control instructions based on the patient's real-time state during treatment, thus hindering the achievement of personalized and efficient treatment results. Summary of the Invention

[0004] To address the above problems, this invention provides a moxibustion temperature control method and system based on multi-sensor fusion, which has the advantages of precise temperature control and synergistic regulation of acupoints during moxibustion treatment.

[0005] In a first aspect, this application provides a moxibustion temperature control method based on multi-sensor fusion, the method comprising: Step 1: Obtain biopotential signals and positional offset data from acupoint areas using sensors, and process the biopotential signals using a classification algorithm to obtain the classified potential distribution pattern; Step 2: Based on the classified potential distribution pattern, obtain the heat conduction influence data between adjacent acupoints, and integrate the position offset data through an information fusion algorithm to determine the fused physiological state indicators; Step 3: Based on the relationship between the physiological state indicators and the preset threshold, a prediction model is used to determine the temperature demand change and obtain the adjustment demand vector; Step 4: Based on the adjustment demand vector, extract the real-time temperature output value from the multi-point collaborative device, and use a feedback control mechanism to calculate the deviation between the adjustment demand vector and the real-time temperature output value to obtain the deviation correction signal; Step 5: Generate a coordinated control command sequence based on the deviation correction signal, determine the conflict situation of the coordinated control command sequence, and obtain an optimized command set through optimization processing; Step 6: Transmit the optimized instruction set through the device interface to determine the updated value of the temperature output for each acupoint; Step 7: Monitor the biopotential feedback based on the updated value, and use an iterative method to optimize the signal stability to obtain the final temperature field coordination result.

[0006] Preferably, step 1 includes: Bioelectric potential signals and corresponding positional offset data of acupoint areas are collected using a multi-channel sensor array; The collected biopotential signals are preprocessed, and the fluctuation characteristics of the signals are extracted from them; The wave characteristics are classified using a support vector machine algorithm to generate multiple potential distribution sub-patterns; Spatial calibration of the potential distribution sub-pattern is performed based on the positional offset data to determine the potential distribution pattern of each acupoint region. The classification accuracy of the potential distribution pattern is verified, outlier data is removed, and the final classified potential distribution pattern is obtained.

[0007] Preferably, step 2 includes: Calculate the thermal conductivity coefficient between adjacent acupoints based on the described potential distribution pattern; Based on the aforementioned thermal conductivity coefficient, data on the impact of thermal conduction on each acupoint area were obtained. The heat conduction influence data and the position offset data are fused using the Kalman filter algorithm to generate a multidimensional physiological feature vector. Extract key physiological parameters from the multidimensional physiological feature vector; The key physiological parameters were integrated using a weighted average method to determine the fused physiological state indicators.

[0008] Preferably, step 3 includes: If the physiological state index exceeds a preset threshold, the physiological state index is input into the neural network model. The neural network model is used to predict the temperature requirement changes in each acupoint area. Based on the predicted changes in temperature demand, an initial adjustment vector is generated; The initial adjustment vector is normalized to generate a standardized adjustment requirement vector; Verify the degree of matching between the standardized adjustment demand vector and the current temperature output value to determine the final adjustment demand vector.

[0009] Preferably, step 4 includes: Based on the adjusted demand vector, the target output point in the multi-point collaborative device is determined; Obtain the real-time temperature output value of the target output point; The deviation between the real-time temperature output value and the adjustment demand vector is calculated using a proportional-integral-derivative control algorithm. Based on the aforementioned deviation, a preliminary correction signal is generated; The preliminary correction signal is filtered to remove high-frequency noise, resulting in the final deviation correction signal.

[0010] Preferably, step 5 includes: A preliminary coordinated control command sequence is generated based on the deviation correction signal; Conflict detection is performed on the preliminary coordinated control command sequence to determine whether there is mutual interference between commands; If a conflict exists, the instruction execution order is adjusted using a priority sorting algorithm; The adjusted instruction sequence is optimized to generate a balanced control instruction set. The stability of the balanced control instruction set is verified to obtain the optimized instruction set.

[0011] Preferably, step 6 includes: The optimized instruction set is converted into control signals that the device can recognize; The control signals are transmitted to the multi-point collaborative device via a standardized device interface; The target temperature output value for each acupoint area is determined based on the control signal. The target temperature output value is calibrated in real time to generate updated temperature output values ​​for each acupoint. Verify whether the updated value meets the device output range, and determine the final temperature output update value based on the verification result.

[0012] Preferably, step 7 includes: Based on the updated values, bioelectrical feedback signals are collected during the treatment process; Time-domain and frequency-domain analyses were performed on the biopotential feedback signal to extract signal stability characteristics; The stability features are optimized using a cyclic iterative method to generate a stable feedback signal; Determine whether the stabilized feedback signal meets a preset stability threshold; If so, the final temperature field coordination result is generated based on the stabilized feedback signal.

[0013] Preferably, step 3 includes: The physiological state indicators are compared with preset thresholds, and a decision is made on whether to trigger temperature adjustment based on the comparison results. If temperature adjustment is triggered, the physiological state indicators are input into a pre-trained deep neural network. The deep neural network is used to predict the dynamic temperature requirements of each acupoint area. Based on the aforementioned dynamic temperature requirements, a multi-dimensional adjustment requirement vector is generated; The multidimensional adjustment demand vector is smoothed to remove outliers, resulting in the final adjustment demand vector.

[0014] Secondly, this application provides a moxibustion control system based on multi-sensor fusion, used to implement a moxibustion temperature control method based on multi-sensor fusion, the system comprising: The sensing and classification module is used to acquire biopotential signals and positional offset data from acupoint areas through sensors, and to process the biopotential signals using a classification algorithm to obtain the classified potential distribution pattern. The fusion analysis module is used to obtain the heat conduction influence data between adjacent acupoints based on the classified potential distribution pattern, and to integrate the position offset data through an information fusion algorithm to determine the fused physiological state indicators. The demand forecasting module is used to determine changes in temperature demand based on the relationship between the physiological state indicators and preset thresholds using a forecasting model, and to obtain an adjustment demand vector. The feedback calculation module is used to extract real-time temperature output values ​​from the multi-point collaborative device according to the adjustment demand vector, and to calculate the deviation between the adjustment demand vector and the real-time temperature output values ​​using a feedback control mechanism to obtain a deviation correction signal. The optimization and coordination module is used to generate a coordination control command sequence based on the deviation correction signal, determine the conflict situation of the coordination control command sequence, and obtain an optimized command set through optimization processing. The instruction issuing module is used to transmit the optimized instruction set through the device interface to determine the updated value of the temperature output of each acupoint. The steady-state iteration module is used to monitor the biopotential feedback based on the updated value and to optimize the signal stability using an iterative method to obtain the final temperature field coordination result.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By integrating multi-source data and dynamic error compensation, the measurement distortion caused by acupoint movement and equipment offset is solved. At the same time, the system uses spatial calibration and classification algorithms to build an accurate potential distribution model. Combined with real-time position tracking, the temperature control accuracy reaches the millimeter level, which significantly improves the integrity and reliability of the data.

[0016] 2. Based on the fusion of thermal conductivity quantification model and multi-dimensional physiological characteristics, the system achieves accurate prediction of temperature demand. Through the synergistic effect of neural network feedforward control and PID feedback correction, it forms a forward-looking regulation capability, effectively suppresses temperature fluctuations, and ensures that the system maintains stable operation under complex physiological changes.

[0017] 3. By adopting digital twin simulation and dynamic priority scheduling mechanism, the system can identify and eliminate spatiotemporal conflicts between multiple devices. Through constraint optimization algorithm, the global optimization of instruction sequence is achieved, which significantly improves the efficiency of equipment collaboration and energy utilization while ensuring the treatment effect.

[0018] 4. Through the iterative optimization mechanism of bioelectric potential signals, the system establishes a temperature control mode guided by physiological feedback. By adopting time-frequency joint analysis and gradient optimization methods, the temperature stimulus and individual physiological response are dynamically matched, effectively preventing tissue damage and achieving a balance between treatment safety and effectiveness. Attached Figure Description

[0019] Figure 1 This is a flowchart of the moxibustion temperature control method based on multi-sensor fusion according to the present invention; Figure 2 This is a schematic diagram of the structure of the moxibustion temperature control system based on multi-sensor fusion of the present invention. Detailed Implementation

[0020] 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.

[0021] For ease of understanding, the specific process of the embodiments of this application is described below. Figure 1 The diagram shows a flowchart of the moxibustion temperature control method based on multi-sensor fusion provided by the present invention. The flowchart specifically includes the following steps: Step 1: Simultaneously acquire biopotential signals and position offset data from acupoint areas using a multi-channel sensor array, process the biopotential signals using a classification algorithm, and obtain the classified potential distribution pattern.

[0022] In one specific embodiment, the process of performing step 1 may specifically include the following steps: Bioelectric potential signals and corresponding positional offset data of acupoint areas are collected using a multi-channel sensor array; The collected biopotential signals are preprocessed, and the fluctuation characteristics of the signals are extracted from them; The wave characteristics are classified using a support vector machine algorithm to generate multiple potential distribution sub-patterns; Spatial calibration of the potential distribution sub-pattern is performed based on the positional offset data to determine the potential distribution pattern of each acupoint region. The classification accuracy of the potential distribution pattern is verified, outlier data is removed, and the final classified potential distribution pattern is obtained.

[0023] Specifically, biopotential signals are acquired from acupoint regions by a multi-channel sensor array and undergo preprocessing steps, including bandpass filtering and wavelet transform, to remove environmental noise and physiological artifacts, thereby extracting fluctuation features such as time-domain mean, variance, and frequency-domain power spectral density. The support vector machine algorithm uses kernel functions to map nonlinearly separable features to a high-dimensional space, optimizes the decision boundary through the maximum margin principle, and generates potential distribution sub-patterns representing clustering results for different physiological states, such as feature distributions corresponding to normal, overheated, or underheated states.

[0024] Position offset data is introduced to spatially calibrate the potential distribution sub-patterns to address pattern distortion caused by acupoint movement or sensor positioning errors. The position offset data is captured by an inertial measurement unit or optical sensor and converted into standardized spatial coordinates through coordinate transformation. Spatial calibration employs affine transformation or nearest neighbor interpolation methods to align the sub-patterns to a unified acupoint coordinate system, thus resolving pattern distortion caused by patient movement or device displacement. The calibrated potential distribution patterns reflect the correspondence between the potential dynamics and spatial positions of each acupoint region; for example, the spatial correlation between the potential patterns of adjacent acupoints can be calculated based on the position data.

[0025] Position offset data is used to perform geometric transformations on potential distribution sub-patterns. For example, through affine transformation or nearest neighbor interpolation, the sub-patterns are aligned to a standardized acupoint coordinate system, thereby generating potential distribution patterns for each acupoint region. This step ensures the correspondence between potential data and spatial location. For example, the potential patterns of adjacent acupoints can be spatially correlated based on location data, laying the foundation for subsequent heat conduction analysis and verifying the classification accuracy of potential distribution patterns.

[0026] In the data processing logic, there is a spatiotemporal correspondence between biopotential signals and location offset data. The signals collected by the multi-channel sensor array are synchronized through timestamps to ensure that the preprocessed fluctuation features correspond one-to-one with the spatial coordinates. When the support vector machine algorithm processes the fluctuation features, the feature vector is used as input and the output is the potential distribution sub-pattern. These sub-patterns are mapped in the spatial domain through the location offset data to form the calibrated potential distribution pattern. In the verification stage, the classification accuracy index and the anomaly detection results interact. For example, the output of the confusion matrix guides the adjustment of the anomaly threshold, while the data after anomaly removal is fed back to the spatial calibration step for re-alignment and optimization.

[0027] After removing outlier data, the system recalculates the statistical characteristics of the potential distribution pattern, such as the mean vector and covariance matrix, to ensure the representativeness and consistency of the pattern. The outlier removal logic relies on multi-dimensional verification, including the continuity of the time series and the rationality of the spatial distribution. For example, if the potential pattern of a certain acupoint fluctuates drastically in a short period of time or is highly uncorrelated with adjacent patterns, it is considered an anomaly. Through an iterative optimization process, the system reclassifies and calibrates the remaining data to generate the final classified potential distribution pattern. This pattern serves as the input for subsequent heat conduction analysis and the fusion of physiological state indicators, and its accuracy directly affects the reliability of temperature demand prediction.

[0028] Step 2: Based on the classified potential distribution pattern, obtain the thermal conduction influence data between adjacent acupoints, and integrate the position offset data through the information fusion algorithm to determine the fused physiological state indicators.

[0029] In one specific embodiment, the process of performing step 2 may specifically include the following steps: Calculate the thermal conductivity coefficient between adjacent acupoints based on the described potential distribution pattern; Based on the aforementioned thermal conductivity coefficient, data on the impact of thermal conduction on each acupoint area were obtained. The heat conduction influence data and the position offset data are fused using the Kalman filter algorithm to generate a multidimensional physiological feature vector. Extract key physiological parameters from the multidimensional physiological feature vector; The key physiological parameters were integrated using a weighted average method to determine the fused physiological state indicators.

[0030] Specifically, the thermal conductivity coefficient between adjacent acupoints is calculated based on the potential distribution pattern. This calculation involves physical modeling, where potential gradient data is combined with distance information between acupoints, and heat flow is simulated through linear regression or finite element analysis to quantify the thermal conductivity coefficient. For example, areas with high potential may indicate higher local temperature sensitivity, thus affecting the rate of heat propagation in tissues. Based on the thermal conductivity coefficient, data on the impact of thermal conduction on each acupoint area are obtained. These data represent the potential impact of heat propagation from one acupoint to another during moxibustion.

[0031] Specifically, after obtaining the heat conduction coefficients between acupoints, a heat conduction network model is constructed by combining the spatial location data of the acupoints. Each node in the model represents an acupoint, and the edge weights between nodes are the corresponding heat conduction coefficients. The spatial location data provides the distance parameter 'd' between nodes, used to calculate the length of the heat conduction path. This is based on the fundamental formula of heat conduction. Calculate the heat transfer rate, where Q is the heat transfer rate, S is the area of ​​the acupoint, T1 is the current temperature of the heat source acupoint (obtained in real time from a multi-point collaborative device), and T2 is the current temperature of the affected acupoint (also collected in real time).

[0032] For each acupoint, it is treated as a heat source acupoint. The heat transfer rate Q of the heat source acupoint to all adjacent acupoints is calculated through the heat conduction network model. Each adjacent acupoint corresponds to a Q value. At the same time, the heat transfer direction is recorded, that is, the correspondence between the heat source acupoint identifier and the affected acupoint identifier, forming a heat conduction influence dataset of the heat source acupoint.

[0033] To integrate the data on the effects of thermal conduction and the location offset, a Kalman filter algorithm was employed for fusion processing. The Kalman filter treats the thermal conduction data as a state variable and the location offset data as an observation variable, iteratively optimizing the data fusion through prediction and update steps. In other words, it predicts the current thermal conduction effect based on the state at the previous moment, and then uses the location offset data to correct the predicted value, thereby generating a multi-dimensional physiological feature vector. This feature vector contains fused spatiotemporal information, such as the thermal conduction intensity, location stability index, and dynamic changes in potential patterns for each acupoint. These dimensions collectively characterize multiple aspects of the physiological state.

[0034] After generating a multidimensional physiological feature vector, key physiological parameters are extracted. This step involves feature selection techniques, such as principal component analysis or correlation analysis, to identify variables most sensitive to temperature regulation. Principal component analysis reduces the dimensionality of high-dimensional data through linear transformation, retaining the components with the largest variance. Correlation analysis calculates the Pearson coefficient between parameters and historical temperature requirements, filtering out parameters such as heat conduction rate, potential fluctuation amplitude, or positional offset variance. These parameters are selected because they are directly related to the effectiveness and safety of moxibustion treatment; for example, heat conduction rate affects temperature diffusion efficiency, and potential fluctuation amplitude reflects the intensity of the physiological response.

[0035] Subsequently, a weighted average method was used to integrate these key physiological parameters to determine the fused physiological state index. The weighting coefficients were set based on historical data or expert knowledge; for example, the importance weights of each parameter were derived through machine learning model training to ensure that the index comprehensively reflects the overall physiological state of the acupoints. Specifically, the weighted average calculation assigned a weight factor to each parameter. These factors were optimized during the training phase using a gradient descent algorithm to minimize the error between the predicted temperature demand and the actual value. The integrated physiological state index was output as a scalar or vector for subsequent temperature demand prediction.

[0036] Step 3: Based on the relationship between physiological state indicators and preset thresholds, use a prediction model to determine changes in temperature demand and obtain an adjustment demand vector.

[0037] In one specific embodiment, the process of performing step 3 may specifically include the following steps: If the physiological state index exceeds a preset threshold, the physiological state index is input into the neural network model. The neural network model is used to predict the temperature requirements of each acupoint area. Based on the predicted changes in temperature demand, an initial adjustment vector is generated; The initial adjustment vector is normalized to generate a standardized adjustment requirement vector; Verify the degree of matching between the standardized adjustment demand vector and the current temperature output value, and determine the final adjustment demand vector; The physiological state indicators are compared with preset thresholds, and a decision is made on whether to trigger temperature adjustment based on the comparison results.

[0038] Specifically, the process of determining the adjustment demand vector begins with a threshold comparison of a physiological state indicator. This indicator is generated by fusing heat conduction influence data and positional offset data through Kalman filtering, and its value reflects the comprehensive physiological load of the acupoint area. The system compares this indicator with a preset clinical threshold, which is derived from a large amount of historical treatment data and represents the safe tolerance boundary of human tissue under moxibustion stimulation. When the physiological state indicator exceeds this critical value, the system triggers a temperature adjustment mechanism, inputting the indicator data into a pre-trained neural network model.

[0039] The neural network model employs a deep network architecture, having learned the complex mapping relationship between physiological parameters and ideal temperature settings during offline training. The model comprises multiple hidden layers, each employing a non-linear activation function to transform features. Neurons in the input layers receive physiological state indicators and their temporal differences to capture dynamic trends. The hidden layers abstract and reorganize the input features layer by layer, while the output layer generates predicted temperature requirements corresponding to the number of acupoints, forming a preliminary adjustment vector. Each element in this vector represents the required temperature change for a specific acupoint under the current physiological state, and its numerical range may fluctuate significantly due to individual differences and the stage of treatment.

[0040] The initial adjustment vector then enters the normalization stage, employing a Min-Max scaling algorithm to linearly transform the adjustment values ​​of each acupoint in the vector to a closed interval [0, 1]. The normalization process determines the maximum upper limit and minimum lower limit of adjustment for each acupoint based on the statistical distribution of temperature adjustment values ​​in historical treatment records. The vector after standardization is called the standardized adjustment demand vector, whose values ​​have uniform dimensions and comparability, facilitating subsequent control algorithm processing.

[0041] The system then verifies the matching degree between the standardized vector and the actual device state. The temperature output value of the multi-point collaborative moxibustion device is acquired in real time through the device interface. A current temperature vector with the same dimension as the adjustment demand vector is constructed, and the cosine similarity between the two vectors is calculated to measure their directional consistency in multi-dimensional space. Simultaneously, the numerical differences in each acupoint dimension are evaluated. If the deviation between the adjustment demand and the current temperature of a certain acupoint exceeds the safety tolerance, the value in that dimension is smoothed and corrected. It should be noted that the smoothing process uses a moving average filter, and its window size is dynamically adjusted according to the treatment stage. A larger window is used in the early stages of treatment to maintain stability, while a smaller window is used in the fine-tuning stage to improve response speed.

[0042] The vector that has undergone matching degree verification and outlier correction is confirmed as the adjustment demand vector. This vector serves as the setting input for the temperature control system. Each dimension value represents the temperature target that the corresponding acupoint needs to reach in the next control cycle. In the entire process, physiological state indicators serve as trigger signals and neural network inputs. The initial adjustment vector carries the neural network prediction results, the standardized vector provides intermediate data with unified dimensions, the current temperature vector serves as the verification benchmark, and the adjustment demand vector is the control command after multiple verification steps.

[0043] The neural network continuously optimizes the weight parameters through the backpropagation algorithm to make the predicted output approach the optimal temperature value verified in clinical practice; the normalized parameters are adaptively updated according to the characteristics of the equipment and the patient's physical condition; the matching degree verification process introduces a temperature change rate constraint to prevent excessive thermal gradients between adjacent acupoints.

[0044] In a preferred embodiment, the process of performing step 3 may further include the following steps: The physiological state indicators are compared with preset thresholds, and a decision is made on whether to trigger temperature adjustment based on the comparison results. If temperature adjustment is triggered, the physiological state indicators are input into a pre-trained deep neural network. The deep neural network is used to predict the dynamic temperature requirements of each acupoint area. Based on the aforementioned dynamic temperature requirements, a multi-dimensional adjustment requirement vector is generated; The multidimensional adjustment demand vector is smoothed to remove outliers, resulting in the final adjustment demand vector.

[0045] Specifically, the physiological state indicators are generated by fusing thermal conduction influence data and positional offset data through Kalman filtering. The data dimensions correspond to the number of acupoints, and each dimension reflects the comprehensive physiological load of the corresponding acupoint area, encompassing key parameters such as thermal conduction rate, potential fluctuation amplitude, and positional offset variance. Preset thresholds are derived from a large amount of clinical moxibustion treatment data and are set separately for each acupoint. Each acupoint corresponds to an independent threshold range, with the upper and lower limits determined based on the safe tolerance range of human tissue under moxibustion stimulation, the historical optimal treatment effect range, and the equipment's output capability boundaries. For example, the threshold range for acupoint A is [0.3, 0.8], and the threshold range for acupoint B is [0.25, 0.75]. Threshold data is stored in the system database and can be dynamically updated according to the treatment scenario and population characteristics.

[0046] The data comparison process is carried out on a per-acupoint basis. The system extracts physiological state indicator vectors from the fusion analysis module and simultaneously retrieves the preset threshold ranges for the corresponding acupoints from the database, establishing a one-to-one correspondence between indicator dimensions and acupoint thresholds. For each acupoint, its physiological state indicator value is compared with the corresponding threshold range. The comparison logic uses range inclusion judgment, that is, it determines whether the indicator value falls within the threshold range.

[0047] The neural network generates dynamic temperature requirements for each acupoint based on physiological state indicators, forming a preliminary adjustment vector containing multiple dimensions, where each element corresponds to the theoretical temperature adjustment amount for a specific acupoint. Because neural network predictions may be affected by signal transients or instantaneous model errors, the values ​​of some dimensions in the vector may exhibit abnormal fluctuations deviating from the normal range. Directly using these outliers for control may cause instability in the temperature field.

[0048] A sliding window averaging filter is used to process the initial adjustment vector. The filter window width W is dynamically configured according to the treatment stage: a larger window is used in the early stage of treatment to enhance stability, and a smaller window is used in the fine adjustment stage to preserve response sensitivity. For the adjustment value sequence x of each acupoint in the vector, the filter calculates the arithmetic mean of the current time and the previous W-1 time points to generate a smoothed sequence y. During processing, the system synchronously calculates the standard deviation σ of the adjustment value of each acupoint. When the adjustment value at a certain time point deviates from the window mean by more than 3σ, the value is marked as an outlier.

[0049] For identified outliers, the system uses linear interpolation for replacement. The normal values ​​adjacent to the outlier are used as interpolation base points, and the replacement value is calculated through linear fitting to ensure the continuity of the adjustment vector. After initial smoothing, the system further applies a low-pass digital filter with a Butterworth transfer function H(z). The cutoff frequency f is set according to the thermal response characteristics of human tissue, typically chosen as a critical value lower than the main frequency components of heat conduction. This filter is implemented through a difference equation, performing frequency domain shaping on the adjustment sequence for each acupoint, effectively suppressing high-frequency noise components.

[0050] The vectors, after undergoing dual filtering, enter the verification phase, where the system calculates their similarity to the real-time temperature output value. Cosine similarity is used to measure the directional consistency of the two vectors in multidimensional space. When the similarity is below a preset threshold, the adjustment vector is scaled. The scaling factor λ is determined based on the safe range of temperature changes in historical treatment data, ensuring that the adjustment amount meets treatment needs without exceeding physiological tolerance limits.

[0051] The vector resulting from all processing steps is identified as the adjustment demand vector, whose dimension strictly corresponds to the number of acupoints. Each element represents an optimized temperature adjustment command. In the data processing chain, the initial adjustment vector output by the neural network serves as input for smoothing. A sliding window filter provides temporal smoothing, a low-pass filter achieves frequency domain cleanup, outlier detection and interpolation ensure data integrity, and similarity verification maintains the consistency between control commands and system states.

[0052] Step 4: Based on the adjustment demand vector, extract the real-time temperature output value from the multi-point collaborative device, and use a feedback control mechanism to calculate the deviation between the adjustment demand vector and the real-time temperature output value to obtain the deviation correction signal.

[0053] In one specific embodiment, the process of performing step 4 may specifically include the following steps: Based on the adjusted demand vector, the target output point in the multi-point collaborative device is determined; Obtain the real-time temperature output value of the target output point; The deviation between the real-time temperature output value and the adjustment demand vector is calculated using a proportional-integral-derivative control algorithm. Based on the aforementioned deviation, a preliminary correction signal is generated; The preliminary correction signal is filtered to remove high-frequency noise, resulting in the final deviation correction signal.

[0054] Specifically, the process of obtaining the deviation correction signal begins with adjusting the demand vector. This vector, generated by a neural network model based on physiological state indicators, characterizes the temperature demand changes in each acupoint region. Based on the adjustment demand vector, target output points in the multi-point collaborative device are determined. These points correspond to the heating units in the acupoint regions, and their selection is based on the mapping relationship between the indices of elements in the vector and the physical locations of the devices. For example, each dimension value in the adjustment demand vector is associated with the heating unit identifier of a specific acupoint, and the target output point is located by querying the device configuration table. After determining the target output points, the system initiates a data request through the device communication interface to acquire the temperature sensor readings of these points in real time. This constructs a real-time temperature output value vector consistent with the dimensions of the adjustment demand vector, reflecting the current actual temperature state of each acupoint. Data acquisition is performed at a high sampling rate to ensure the continuity of the time series.

[0055] A proportional-integral-derivative (PID) control algorithm is used to calculate the deviation between the real-time temperature output value and the adjustment demand vector. This algorithm operates independently for each acupoint control channel, with its input being the difference between the target temperature value and the real-time temperature output value in the adjustment demand vector, referred to as the instantaneous error. The proportional term is calculated as the product of this error and a proportional gain coefficient, providing a rapid correction response. The integral term eliminates steady-state deviations by accumulating the product of historical error values ​​and an integral gain coefficient. The derivative term predicts future trends and suppresses oscillations based on the product of the error change rate and a derivative gain coefficient. The outputs of these three terms are linearly superimposed to generate a preliminary correction signal for each acupoint. This signal, as a multi-dimensional vector, indicates the amount of energy output adjustment required for each heating unit.

[0056] The initial correction signal may contain high-frequency components caused by sensor noise, electromagnetic interference, or thermal fluctuations. Therefore, the system performs filtering to improve signal quality. This filtering uses a digital low-pass filter, with its cutoff frequency set based on the thermal inertia characteristics of moxibustion treatment and the device's response time. For example, in the acupoint region, heat conduction has low-frequency characteristics; therefore, the filter is designed to attenuate frequency components above a certain threshold. The filtering process is implemented through convolution operations, performing a time-domain convolution between the initial correction signal and the filter kernel function to remove high-frequency noise while retaining effective control command components. After filtering, the signal becomes smooth and stable, forming the final deviation correction signal. This signal serves as the input to the coordinated control command sequence, driving the temperature adjustment of the multi-point collaborative device.

[0057] In the data processing logic, the adjustment demand vector serves as the setpoint input, and the real-time temperature output value serves as the feedback input. These two interact through a proportional-integral-derivative (PID) control algorithm to generate a preliminary correction signal. The gain coefficient in the PID algorithm is calibrated based on equipment characteristics and historical treatment data, for example, by tuning parameters using the Ziegler-Nichols method to ensure system stability under different physiological states. The preliminary correction signal is mapped to the filter kernel function in the frequency or time domain. The filtered signal has the same dimension as the preliminary correction signal, but high-frequency noise is suppressed. The data correspondence is as follows: each element of the adjustment demand vector corresponds to a target output point; the real-time temperature output value vector has the same dimension as the adjustment demand vector; the PID outputs the preliminary correction signal; and the filtering process outputs the deviation correction signal.

[0058] Step 5: Generate a coordinated control command sequence based on the deviation correction signal, determine the conflict situation of the coordinated control command sequence, and obtain the optimized command set through optimization processing.

[0059] In one specific embodiment, the process of performing step 5 may specifically include the following steps: A preliminary coordinated control command sequence is generated based on the deviation correction signal; Conflict detection is performed on the preliminary coordinated control command sequence to determine whether there is mutual interference between commands; If a conflict exists, the instruction execution order is adjusted using a priority sorting algorithm; The adjusted instruction sequence is optimized to generate a balanced control instruction set. The stability of the balanced control instruction set is verified to obtain the optimized instruction set.

[0060] Specifically, the system generates a preliminary coordinated control command sequence based on the deviation correction signal. This conversion process discretizes the continuous correction amount into control commands with time stamps. Each command contains an execution timestamp, a target acupoint identifier, and specific control parameters. The control parameters map the energy adjustment amount to the duty cycle of pulse width modulation or the heating power level through a lookup table method, forming a preliminary command sequence arranged in chronological order.

[0061] The system then performs conflict detection on the initial command sequence, a process implemented through digital twin simulation. The system constructs a thermal dynamics model of the acupoint region, which establishes partial differential equations based on the thermal conductivity characteristics of biological tissues to simulate the temperature distribution formed after the execution of each control command. In the simulation environment, the initial command sequence is executed at time steps, calculating the superposition effect of the thermal fields of adjacent acupoints. When the rate of temperature change in a local area exceeds a safety threshold or the thermal gradient exceeds the physiological tolerance range, it is determined that there is interference between commands. This interference may manifest as simultaneous heating of adjacent acupoints leading to heat accumulation, or the heating cycle of one acupoint overlapping with the heat dissipation phase of another acupoint.

[0062] When a conflict is detected, the system initiates a priority ranking algorithm to adjust the execution order of instructions. The algorithm assigns a priority weight to each instruction based on a multi-factor evaluation system. Evaluation factors include: the basic weight coefficient of acupoints in the treatment plan, determined according to meridian theory; the urgency of physiological state indicators, quantified by the magnitude of deviation from thresholds; thermal inertia parameters, calculated based on tissue specific heat capacity and blood perfusion rate; and factors influencing the uniformity of the global temperature field. The algorithm calculates the priority value of each instruction using a weighted summation method and rearranges the instruction execution sequence according to the priority value, with higher-priority instructions receiving earlier execution.

[0063] The adjusted instruction sequence is optimized. The system uses a constrained optimization algorithm to generate a balanced control instruction set. The objective function is set to minimize the weighted sum of the total system energy consumption and the temperature field non-uniformity. The constraints include the temperature safety boundary of each acupoint, the maximum heating rate limit, and the upper limit of equipment power. The multi-objective optimization problem is solved by the Lagrange multiplier method to obtain the Pareto optimal solution under all constraints. Based on this, the instruction parameters are adjusted, such as shortening the heating time or reducing the power level, to form a control instruction set that is evenly distributed in the time dimension and coordinated in the spatial dimension of the thermal field.

[0064] The system confirms the optimization results through rapid stability verification, which is carried out in a simplified heat conduction model to simulate the system's dynamic response when executing the optimized instruction set. The verification indicators include temperature overshoot, settling time, and steady-state error. When all indicators are within the preset range, the instruction set is confirmed as the optimized instruction set.

[0065] Step 6: Transmit the optimized instruction set through the device interface to determine the updated value of the temperature output for each acupoint.

[0066] In one specific embodiment, the process of performing step 6 may specifically include the following steps: The optimized instruction set is converted into control signals that the device can recognize; The control signals are transmitted to the multi-point collaborative device via a standardized device interface; The target temperature output value for each acupoint area is determined based on the control signal. The target temperature output value is calibrated in real time to generate updated temperature output values ​​for each acupoint. Verify whether the updated value meets the device output range, and determine the final temperature output update value based on the verification result.

[0067] Specifically, the system converts the optimized instruction set into control signals recognizable by the underlying hardware through the device driver layer. This conversion process encodes the instruction parameters into data frames of a specific format according to the device communication protocol. For example, heating power levels are converted into pulse width modulation duty cycle values, and timing instructions are converted into precise delay control codes. These structured control signals are transmitted to distributed multi-point collaborative moxibustion devices through standardized interfaces. The interfaces use serial communication or wireless transmission methods to ensure the integrity and timeliness of the instructions.

[0068] After the control signal reaches the execution unit, the system analyzes the signal content to determine the target temperature output value for each acupoint region. The analysis process uses a lookup table method or a linear mapping relationship to convert the control quantity into a theoretical temperature value. For example, there is a linear relationship between the duty cycle of pulse width modulation and the heating power; the corresponding theoretical temperature value is calculated using a pre-calibrated conversion coefficient. These target values ​​constitute a preliminary temperature setting vector, the dimension of which is consistent with the number of acupoints in the system.

[0069] Due to factors such as thermal inertia, individual tissue differences, and ambient temperature fluctuations in the actual treatment environment, the theoretical target value needs to be calibrated in real time. The system obtains the most recent temperature sensor readings through the device interface and compares these measured values ​​with the theoretical target value. The calibration algorithm employs a compensation mechanism, calculating the compensation amount based on a heat transfer model established from historical data. For example, when the thermal response rate of the target acupoint is detected to be lower than expected, the system adds a positive compensation value, which is calculated based on the deviation between the acupoint's historical thermal data and the standard response curve. The temperature values ​​after compensation are called the updated values ​​of the temperature output for each acupoint, and these values ​​are more closely aligned with actual treatment needs.

[0070] The system performs safety verification on these updated values, checking whether the updated value for each acupoint is within the device's allowed output range and human safety thresholds. The verification process is implemented using a boundary comparator. Each acupoint corresponds to a minimum and maximum allowable value, which are preset according to device specifications and clinical safety standards. When an updated value exceeds the allowable range, the system initiates clamping processing, forcibly adjusting the value to the nearest safety boundary; for example, an excessively high updated value will be reduced to the maximum allowable value, and the abnormal event will be recorded, potentially sending a warning signal to upstream modules.

[0071] Based on the verification results, the system determines the final temperature output update value. The verified update value is directly confirmed as the valid output, while the clamped update value is adjusted and used as the final output. These determined update values ​​are organized into an output vector that matches the control system dimension, ready to drive the actuators in the next control cycle.

[0072] Step 7: Monitor the biopotential feedback based on the updated values, and use an iterative method to optimize signal stability to obtain the final temperature field coordination result.

[0073] In one specific embodiment, the process of performing step 6 may specifically include the following steps: Based on the updated values, bioelectrical feedback signals are collected during the treatment process; Time-domain and frequency-domain analyses were performed on the biopotential feedback signal to extract signal stability characteristics; The stability features are optimized using a cyclic iterative method to generate a stable feedback signal; Determine whether the stabilized feedback signal meets a preset stability threshold; If so, the final temperature field coordination result is generated based on the stabilized feedback signal.

[0074] Specifically, when the multi-point collaborative device applies new thermal stimulation according to the updated temperature output value, the multi-channel sensor array deployed in the acupoint area immediately starts collecting bioelectrical feedback signals. These signals are continuously recorded at a high sampling rate, capturing the real-time changes in electrophysiological activity in the acupoint area under moxibustion thermal stimulation, forming a biopotential data stream that is time-synchronized with the temperature update value.

[0075] The acquired biopotential feedback signals undergo preprocessing, including bandpass filtering and baseline correction, to eliminate electromyographic interference and environmental noise. The preprocessed signals then proceed to the feature extraction stage, where the system performs time-domain and frequency-domain analyses in parallel. Time-domain analysis calculates the statistical characteristics of the signal within a sliding time window, including the variance δ and fluctuation range ρ, where δ reflects the dispersion of the signal amplitude and ρ characterizes the difference between signal extrema. Frequency-domain analysis transforms the signal to the frequency domain using a Fast Fourier Transform (FFT) to calculate the proportion of energy in each frequency band to the total energy, with particular attention to the low-frequency band energy ratio η and the high-frequency band energy ratio κ. These parameters collectively constitute the signal stability feature vector Ψ=(δ,ρ,η,κ).

[0076] The system compares the feature vector Ψ with a preset stability threshold, which is a multidimensional boundary condition derived from extensive clinical data. This threshold defines the stable range that a bioelectrical signal should satisfy under ideal physiological conditions. The comparison process employs a dimensional verification method. When all feature values ​​in Ψ fall between the upper and lower bounds of the dimension corresponding to the preset stability threshold, the stabilized feedback signal is deemed to meet the stability requirements. If the condition is not met, the system initiates an iterative optimization process, calculating the temperature adjustment increment ΔT based on the gradient descent algorithm. This increment is in the direction of the negative gradient that converges Ψ towards the preset stability threshold, and its magnitude is proportional to the Euclidean distance between Ψ and the preset stability threshold.

[0077] Each iteration generates a ΔT value, which is then superimposed on the current temperature update value to form a new temperature command for execution. The system then re-acquires the bioelectrical feedback signal, repeating feature extraction and threshold comparison to form an internal control loop focused on stability. This iterative process continues until all components of the feature vector Ψ enter the stable region defined by the preset stability threshold.

[0078] When the stability condition is met, the system generates the final temperature field coordination result based on the current stabilization feedback signal. This result is stored in the form of a data matrix, containing the final temperature setpoint for each acupoint, the corresponding bioelectrical characteristic value, and the timestamp of reaching a stable state. Simultaneously, the system records the combination of control parameters at the point of stabilization, including the duty cycle configuration of pulse width modulation and the heat output timing of each acupoint. These data constitute a complete temperature field coordination scheme.

[0079] The above describes the moxibustion temperature control method based on multi-sensor fusion in the embodiments of this application. The following describes the moxibustion temperature control system based on multi-sensor fusion in the embodiments of this application. Please refer to [link / reference]. Figure 2 The present invention provides a schematic diagram of a moxibustion temperature control system based on multi-sensor fusion, used to implement a moxibustion temperature control method based on multi-sensor fusion. The system includes: The sensing and classification module is used to acquire biopotential signals and positional offset data from acupoint areas through sensors, and to process the biopotential signals using a classification algorithm to obtain the classified potential distribution pattern. The fusion analysis module is used to obtain the thermal conduction influence data between adjacent acupoints based on the classified potential distribution pattern, and to integrate the position offset data through the information fusion algorithm to determine the fused physiological state indicators. The demand forecasting module is used to determine changes in temperature demand based on the relationship between physiological state indicators and preset thresholds using a forecasting model, and to obtain an adjustment demand vector. The feedback calculation module is used to extract real-time temperature output values ​​from multi-point collaborative devices based on the adjustment demand vector, and to calculate the deviation between the adjustment demand vector and the real-time temperature output values ​​using a feedback control mechanism to obtain a deviation correction signal. The optimization and coordination module is used to generate a coordination control command sequence based on the deviation correction signal, determine the conflict situation of the coordination control command sequence, and obtain an optimized command set through optimization processing; The instruction issuing module is used to transmit the optimized instruction set through the device interface to determine the updated value of the temperature output of each acupoint; The steady-state iteration module is used to monitor biopotential feedback based on updated values ​​and optimize signal stability using an iterative method to obtain the final temperature field coordination result.

[0080] It should be noted that, in this document, 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.

[0081] 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 moxibustion temperature control method based on multi-sensor fusion, characterized in that, include: Step 1: Obtain biopotential signals and positional offset data from acupoint areas using sensors, and process the biopotential signals using a classification algorithm to obtain the classified potential distribution pattern; Step 2: Based on the classified potential distribution pattern, obtain the heat conduction influence data between adjacent acupoints, and integrate the position offset data through an information fusion algorithm to determine the fused physiological state indicators; Step 3: Based on the relationship between the physiological state indicators and the preset threshold, a prediction model is used to determine the temperature demand change and obtain the adjustment demand vector; Step 4: Based on the adjustment demand vector, extract the real-time temperature output value from the multi-point collaborative device, and use a feedback control mechanism to calculate the deviation between the adjustment demand vector and the real-time temperature output value to obtain the deviation correction signal; Step 5: Generate a coordinated control command sequence based on the deviation correction signal, determine the conflict situation of the coordinated control command sequence, and obtain an optimized command set through optimization processing; Step 6: Transmit the optimized instruction set through the device interface to determine the updated value of the temperature output for each acupoint; Step 7: Monitor the biopotential feedback based on the updated value, and use an iterative method to optimize the signal stability to obtain the final temperature field coordination result.

2. The moxibustion temperature control method based on multi-sensor fusion according to claim 1, characterized in that, Step 1 includes: Bioelectric potential signals and corresponding positional offset data of acupoint areas are collected using a multi-channel sensor array; The collected biopotential signals are preprocessed, and the fluctuation characteristics of the signals are extracted from them; The wave characteristics are classified using a support vector machine algorithm to generate multiple potential distribution sub-patterns; Spatial calibration of the potential distribution sub-pattern is performed based on the positional offset data to determine the potential distribution pattern of each acupoint region. The classification accuracy of the potential distribution pattern is verified, outlier data is removed, and the final classified potential distribution pattern is obtained.

3. The moxibustion temperature control method based on multi-sensor fusion according to claim 1, characterized in that, Step 2 includes: Calculate the thermal conductivity coefficient between adjacent acupoints based on the described potential distribution pattern; Based on the aforementioned thermal conductivity coefficient, data on the impact of thermal conduction on each acupoint area were obtained. The heat conduction influence data and the position offset data are fused using the Kalman filter algorithm to generate a multidimensional physiological feature vector. Extract key physiological parameters from the multidimensional physiological feature vector; The key physiological parameters were integrated using a weighted average method to determine the fused physiological state indicators.

4. The moxibustion temperature control method based on multi-sensor fusion according to claim 1, characterized in that, Step 3 includes: If the physiological state index exceeds a preset threshold, the physiological state index is input into the neural network model. The neural network model is used to predict the temperature requirement changes in each acupoint area. Based on the predicted changes in temperature demand, an initial adjustment vector is generated; The initial adjustment vector is normalized to generate a standardized adjustment requirement vector; Verify the degree of matching between the standardized adjustment demand vector and the current temperature output value to determine the final adjustment demand vector.

5. The moxibustion temperature control method based on multi-sensor fusion according to claim 1, characterized in that, Step 4 includes: Based on the adjusted demand vector, the target output point in the multi-point collaborative device is determined; Obtain the real-time temperature output value of the target output point; The deviation between the real-time temperature output value and the adjustment demand vector is calculated using a proportional-integral-derivative control algorithm. Based on the aforementioned deviation, a preliminary correction signal is generated; The preliminary correction signal is filtered to remove high-frequency noise, resulting in the final deviation correction signal.

6. The moxibustion temperature control method based on multi-sensor fusion according to claim 1, characterized in that: Step 5 includes: A preliminary coordinated control command sequence is generated based on the deviation correction signal; Conflict detection is performed on the preliminary coordinated control command sequence to determine whether there is mutual interference between commands; If a conflict exists, the instruction execution order is adjusted using a priority sorting algorithm; The adjusted instruction sequence is optimized to generate a balanced control instruction set. The stability of the balanced control instruction set is verified to obtain the optimized instruction set.

7. The moxibustion temperature control method based on multi-sensor fusion according to claim 1, characterized in that, Step 6 includes: The optimized instruction set is converted into control signals that the device can recognize; The control signals are transmitted to the multi-point collaborative device via a standardized device interface; The target temperature output value for each acupoint area is determined based on the control signal. The target temperature output value is calibrated in real time to generate updated temperature output values ​​for each acupoint. Verify whether the updated value meets the device output range, and determine the final temperature output update value based on the verification result.

8. The moxibustion temperature control method based on multi-sensor fusion according to claim 1, characterized in that, Step 7 includes: Based on the updated values, bioelectrical feedback signals are collected during the treatment process; Time-domain and frequency-domain analyses were performed on the biopotential feedback signal to extract signal stability characteristics; The stability features are optimized using a cyclic iterative method to generate a stable feedback signal; Determine whether the stabilized feedback signal meets a preset stability threshold; If so, the final temperature field coordination result is generated based on the stabilized feedback signal.

9. The moxibustion temperature control method based on multi-sensor fusion according to claim 4, characterized in that, Step 3 includes: The physiological state indicators are compared with preset thresholds, and a decision is made on whether to trigger temperature adjustment based on the comparison results. If temperature adjustment is triggered, the physiological state indicators are input into a pre-trained deep neural network. The deep neural network is used to predict the dynamic temperature requirements of each acupoint area. Based on the aforementioned dynamic temperature requirements, a multi-dimensional adjustment requirement vector is generated; The multidimensional adjustment demand vector is smoothed to remove outliers, resulting in the final adjustment demand vector.

10. A moxibustion control system based on multi-sensor fusion, used to implement the moxibustion temperature control method based on multi-sensor fusion as described in any one of claims 1-9, characterized in that, include: The sensing and classification module is used to acquire biopotential signals and positional offset data from acupoint areas through sensors, and to process the biopotential signals using a classification algorithm to obtain the classified potential distribution pattern. The fusion analysis module is used to obtain the heat conduction influence data between adjacent acupoints based on the classified potential distribution pattern, and to integrate the position offset data through an information fusion algorithm to determine the fused physiological state indicators. The demand forecasting module is used to determine changes in temperature demand based on the relationship between the physiological state indicators and preset thresholds using a forecasting model, and to obtain an adjustment demand vector. The feedback calculation module is used to extract real-time temperature output values ​​from the multi-point collaborative device according to the adjustment demand vector, and to calculate the deviation between the adjustment demand vector and the real-time temperature output values ​​using a feedback control mechanism to obtain a deviation correction signal. The optimization and coordination module is used to generate a coordination control command sequence based on the deviation correction signal, determine the conflict situation of the coordination control command sequence, and obtain an optimized command set through optimization processing. The instruction issuing module is used to transmit the optimized instruction set through the device interface to determine the updated value of the temperature output of each acupoint. The steady-state iteration module is used to monitor the biopotential feedback based on the updated value and to optimize the signal stability using an iterative method to obtain the final temperature field coordination result.

Images