Adaptive massage method and system based on multi-modal physiological signal dynamic feature fusion

Abstract

The invention discloses a self-adaptive massage method and system based on multi-modal physiological signal dynamic feature fusion. Firstly, a surface electromyogram signal, a skin conductance signal, a heart rate variability signal and a body pressure distribution signal of a user are synchronously collected through a multi-sensor array; then, three types of dynamic features including the instantaneous feature, the trend feature and the accumulated feature of each type of signal are extracted in parallel; then, performing normalization and weighted fusion on the dynamic characteristics from different signals, and inputting the dynamic characteristics into an innovative dynamic tensity index model to calculate a comprehensive evaluation value; and finally, inputting the index and the dynamic characteristics of the index into a fuzzy control rule base, resolving an optimal massage parameter in real time according to a preset intelligent rule, and driving an execution mechanism to realize accurate, self-adaptive and predictive massage intervention. Therefore, the problems that existing intelligent massage equipment is not comprehensive in evaluation, poor in self-adaptive capacity and lack of predictive intervention due to the fact that the existing intelligent massage equipment depends on single physiological signals and static threshold judgment are solved.

Description

Technical Field

[0001] This invention relates to the field of massage control technology, and in particular to an adaptive massage method and system based on the fusion of dynamic features of multimodal physiological signals. Background Technology

[0002] With the accelerating pace of society and increasing work pressure, the public's demand for health management and relaxation is growing. As an important health aid, intelligent massage devices have evolved from traditional, simple mechanical massage to an intelligent stage with certain sensing and mode selection functions. However, existing intelligent massage devices still have significant shortcomings in achieving true personalization, adaptability, and intelligence. Their technological limitations are mainly reflected in the following aspects:

[0003] First, they rely on a single physiological signal, resulting in a limited assessment dimension. Most massage devices on the market that claim to be "intelligent" typically rely on only one physiological signal for status assessment, such as using heart rate alone to evaluate a user's recovery status. Human tension or fatigue is a complex physiological and psychological state involving coordinated changes in multiple systems, including the muscular, nervous, and cardiovascular systems. A single physiological signal (such as heart rate) cannot comprehensively and accurately reflect the complex levels of musculoskeletal tension, psychological stress, and the cumulative effects of poor posture, leading to biased and distorted assessment results that are unreliable for precise intervention.

[0004] Second, the use of static threshold judgments lacks dynamic perception capabilities. Existing technologies generally employ simple "IF-THEN" logic based on fixed thresholds for control. For example, when the amplitude of the surface electromyography signal exceeds a certain preset threshold, a fixed "strong massage" mode is triggered. This method is inherently static and lagging. It cannot perceive the dynamic changing trends of the user's tension state (e.g., whether the tension is intensifying or easing) or the continuous cumulative effect of tension. This "rigid" control strategy cannot adapt to the continuous and dynamic evolution of the tension state, resulting in a rigid intervention and a poor user experience.

[0005] Third, the intervention lacks foresight and cannot prevent problems before they occur. Due to the limitations of the aforementioned static judgment mechanism, existing equipment can only initiate intervention when the user's physiological signals clearly indicate a state of high tension or fatigue. This is a passive, "remedial" response. It cannot identify the rapid rise in tension, thus missing the optimal opportunity for predictive and alleviating intervention in the early stages of tension accumulation, before it reaches its peak. This results in the user experiencing unnecessary tension peaks, reducing the effectiveness of relief.

[0006] In summary, there is an urgent need in this field for a breakthrough technical solution that can dynamically perceive the tension state of the human body from multiple dimensions, not only focusing on the instantaneous level of physiological signals, but also keenly capturing their changing trends and historical accumulation, thereby achieving accurate quantification and forward-looking prediction of the tension level, and based on this, realizing real-time, accurate, and adaptive adjustment of massage parameters.

[0007] The information disclosed in this background section is included only to enhance the understanding of the context of this disclosure, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] One objective of this invention is to provide an adaptive massage method and system based on the fusion of dynamic features of multimodal physiological signals, which can solve the problems of incomplete assessment, poor adaptability, and lack of predictive intervention caused by existing intelligent massage devices relying on a single physiological signal and static threshold judgment.

[0009] To achieve the above objectives, on the one hand, the present invention provides an adaptive massage method based on the fusion of dynamic features of multimodal physiological signals, comprising:

[0010] Step S1: Collect various physiological signals of the user in real time through a multi-sensor array. The physiological signals include at least surface electromyography signals, skin conductance signals, heart rate variability signals, and body pressure distribution signals.

[0011] Step S2: Perform dynamic feature extraction on each physiological signal to obtain the instantaneous features, trend features, and cumulative features corresponding to the signal. The instantaneous features reflect the instantaneous intensity level of the signal. The trend features are obtained by differentiating the instantaneous features or calculating their short-time linear regression slope to reflect the rate of change of the signal. The cumulative features are obtained by calculating the integral of the instantaneous features exceeding the resting baseline value over a period of time and are used to reflect the sustained load of the signal.

[0012] Step S3: Based on the dynamic characteristics of the extracted physiological signals, perform weighted fusion to calculate and generate a comprehensive dynamic tension index;

[0013] Step S4: Input the dynamic tension index and its corresponding dynamic features into the fuzzy control rule base, and use the fuzzy control rule base to calculate the control parameters for the massage actuator, and control the massage actuator to perform adaptive massage for the user accordingly.

[0014] Optional,

[0015] The instantaneous feature is a time-domain or frequency-domain feature value used to reflect the instantaneous intensity level of the signal;

[0016] The trend feature is a first-order derivative or regression slope value used to reflect the rate of change of the instantaneous feature;

[0017] The cumulative feature is an integral value used to reflect the sustained load that exceeds the resting reference value of the instantaneous feature;

[0018] The resting reference value is a reference benchmark defined for each physiological signal, reflecting the level of the signal's characteristics when the user is in a relaxed state.

[0019] Optional,

[0020] Within a sliding time window of a specified length, the root mean square value of all acquired surface electromyography (sEMG) signals is calculated as the instantaneous feature I_EMG(t) of the surface electromyography (sEMG).

[0021] Optional,

[0022] The amplitude value of the skin conductance signal (SC) after filtering and smoothing is used as the instantaneous feature I_SC(t) of the skin conductance signal (SC).

[0023] Optional,

[0024] The ratio of low-frequency power to high-frequency power of heart rate variability (LF / HF Ratio) is used as the instantaneous feature I_HRV(t) of heart rate variability signal (HRV).

[0025] Optional,

[0026] The asymmetry index of body pressure distribution is used as the instantaneous feature I_P(t) of the body pressure distribution signal (P).

[0027] Optionally, the dynamic tension index mentioned in step S3 is calculated using the following formula:

[0028] DTI(t) = K1*I(t) + K2*F_D(D(t)) + K3*F_S(S(t));

[0029] in,

[0030] DTI(t) is the dynamic tension index at time t;

[0031] I(t) is the comprehensive value of the instantaneous features at time t, which is obtained by weighted summation of the instantaneous features corresponding to each of the physiological signals;

[0032] D(t) is the comprehensive value of the trend characteristics at time t, which is obtained by weighted summation of the trend characteristics corresponding to each of the physiological signals;

[0033] S(t) is the cumulative feature value at time t, which is obtained by weighted summation of the cumulative features corresponding to each physiological signal;

[0034] K1, K2, and K3 are weighting coefficients.

[0035] F_D is a nonlinear function that performs saturation processing on the composite value of trend characteristics, and F_S is a nonlinear function that performs marginal effect processing on the composite value of cumulative characteristics.

[0036] Optional,

[0037] The function F_D is the arctangent function: F_D(D(t)) = arctan(C * D(t)), where C is a constant for adjusting the sensitivity;

[0038] The function F_S is a logarithmic function: F_S(S(t)) = log(1 + S(t)).

[0039] Optionally, the fuzzy control rule base in step S4 includes multiple fuzzy logic rules based on dynamic tension index and dynamic features, wherein the fuzzy logic rules include at least:

[0040] Predictive intervention rule: When the instantaneous feature composite value is "medium", the trend feature composite value is "rapidly rising", and the cumulative feature composite value is "high", then control the massage actuator to perform massage with high intensity and high frequency rhythm;

[0041] Soothing and recovery rule: When the instantaneous feature composite value is "low", the trend feature composite value is "slowly decreasing", and the cumulative feature composite value is "extremely high", the massage actuator is controlled to perform massage with low intensity and slow rhythm.

[0042] On the other hand, an adaptive massage system is provided for any of the aforementioned adaptive massage methods, comprising:

[0043] A multi-sensor array is used to collect various physiological signals from the user in real time, including at least surface electromyography signals, skin conductance signals, heart rate variability signals, and body pressure distribution signals.

[0044] The dynamic feature extraction module is used to extract dynamic features from each physiological signal to obtain instantaneous features, trend features, and cumulative features corresponding to the signal. The instantaneous features reflect the instantaneous intensity level of the signal. The trend features are obtained by differentiating the instantaneous features or calculating their short-time linear regression slope to reflect the rate of change of the signal. The cumulative features are obtained by calculating the integral of the instantaneous features exceeding the resting baseline value over a period of time to reflect the sustained load of the signal.

[0045] The dynamic tension index calculation module is used to perform weighted fusion based on the dynamic characteristics of multiple extracted physiological signals to calculate and generate a comprehensive dynamic tension index.

[0046] The fuzzy adaptive control module has a built-in fuzzy control rule library. It is used to input the dynamic tension index and its corresponding dynamic features into the fuzzy control rule library. The fuzzy control rule library calculates the control parameters for the massage actuator and controls the massage actuator to perform adaptive massage on the user accordingly.

[0047] A massage actuator configured to perform adaptive massage actions based on the control parameters.

[0048] The beneficial effects of this invention are: it provides an adaptive massage method and system based on the dynamic feature fusion of multimodal physiological signals.

[0049] First, by synchronously collecting various physiological signals such as surface electromyography, skin conductance, heart rate variability, and body pressure distribution through a multi-sensor array, a multi-dimensional data foundation is provided for a comprehensive assessment of stress.

[0050] Next, dynamic feature extraction is performed in parallel for each signal to obtain instantaneous features reflecting the current intensity level, trend features reflecting the rate of change, and cumulative features reflecting the sustained load. These three types of features together constitute a dynamic and three-dimensional description of the physiological state.

[0051] Then, the dynamic features from these different signals are weighted and fused to calculate a comprehensive dynamic tension index, which can quantify the overall tension level of the human body.

[0052] Finally, the dynamic tension index and its corresponding dynamic characteristics are input into the fuzzy control rule base. By leveraging the advantages of fuzzy logic in handling uncertainty problems, precise, nonlinear control parameters are calculated, and the massage actuator is driven to complete adaptive massage.

[0053] The adaptive massage method provided by this invention constructs a complete, closed-loop adaptive massage control process. By introducing multimodal signals, it overcomes the shortcomings of one-sided assessment by a single physiological signal; by extracting three types of dynamic features—instantaneous, trend, and cumulative—it achieves comprehensive perception of tension states from static to dynamic and from the present to the past; by fusing them to form a dynamic tension index, it achieves precise quantification of complex physiological states; and finally, through a fuzzy control rule base, it achieves intelligent decision-making, making massage intervention no longer a rigid "on / off" response, but capable of smoothly and intelligently adapting to continuous changes in the body's tension state, significantly improving the accuracy and comfort of massage.

[0054] Therefore, the adaptive massage method and system based on the fusion of dynamic features of multimodal physiological signals provided by the present invention can solve the problems of incomplete evaluation, poor adaptability and lack of predictive intervention caused by existing intelligent massage devices relying on a single physiological signal and static threshold judgment. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 The flowchart illustrates an adaptive massage method based on the fusion of dynamic features of multimodal physiological signals, provided for an embodiment.

[0057] Figure 2 A structural block diagram of the adaptive massage system provided in the embodiment. Detailed Implementation

[0058] In this invention, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment can be included in at least one embodiment of the invention. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this invention, as long as there is no technical contradiction or conflict, the technical features mentioned in each embodiment can be combined in any way to form a corresponding implementable technical solution.

[0059] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit the invention.

[0060] In the description of this invention, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " generally indicates that the preceding and following objects have an "or" logical relationship.

[0061] In this invention, terms such as “first” and “second” are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy, or order between these entities or operations.

[0062] Without further limitations, the use of terms such as “comprising,” “including,” “having,” or other similar expressions in this invention is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes the stated elements, such that a process, method, or product that includes a list of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such a process, method, or product.

[0063] Similar to the understanding in the Examination Guidelines, in this invention, expressions such as "greater than," "less than," and "exceeding" are understood to exclude the stated number; expressions such as "above," "below," and "within" are understood to include the stated number. Furthermore, in the description of the embodiments of this invention, "multiple" means two or more (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups" and "multiple times," unless otherwise explicitly specified.

[0064] In the description of the embodiments of the present invention, the spatial related expressions used, such as "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "vertical," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," "circumferential," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiments or drawings. They are only for the purpose of describing the specific embodiments of the present invention or for the reader's understanding, and do not indicate or imply that the device or component referred to must have a specific position, a specific orientation, or be constructed or operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.

[0065] Unless otherwise explicitly stated or limited, the terms "installation," "connection," "linking," "fixing," and "setting," as used in the description of the embodiments of this invention, should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral arrangement; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this invention according to the specific circumstances.

[0066] The direct drive mechanism in this invention can be a linear motor, a cylinder, a hydraulic cylinder, or a motor lead screw and slider assembly, etc.; the rotary drive mechanism can be a servo motor, a stepper motor, or a rotary cylinder, etc.

[0067] This invention provides an adaptive massage method and system based on the fusion of dynamic features of multimodal physiological signals, which can solve the problems of incomplete assessment, poor adaptability and lack of predictive intervention caused by existing intelligent massage devices that rely on a single physiological signal and static threshold judgment.

[0068] Example 1

[0069] This embodiment provides an adaptive massage method based on the fusion of dynamic features of multimodal physiological signals, which can solve the problems of incomplete assessment, poor adaptability and lack of predictive intervention caused by existing intelligent massage devices that rely on a single physiological signal and static threshold judgment.

[0070] See Figure 1 The adaptive massage method includes the following steps:

[0071] Step S1: Collect various physiological signals from the user in real time using a multi-sensor array. The physiological signals include at least surface electromyography (sEMG), skin conductance (SC), heart rate variability (HRV), and body pressure distribution (P).

[0072] Step S2: Perform dynamic feature extraction on each physiological signal to obtain the instantaneous features, trend features, and cumulative features corresponding to the signal. The instantaneous features reflect the instantaneous intensity level of the signal. The trend features are obtained by differentiating the instantaneous features or calculating their short-time linear regression slope to reflect the rate of change of the signal. The cumulative features are obtained by calculating the integral of the instantaneous features exceeding the resting baseline value over a period of time and are used to reflect the sustained load of the signal.

[0073] Step S3: Based on the dynamic characteristics of the extracted physiological signals, perform weighted fusion to calculate and generate a comprehensive dynamic tension index;

[0074] Step S4: Input the dynamic tension index and its corresponding dynamic features into the fuzzy control rule base, and use the fuzzy control rule base to calculate the control parameters for the massage actuator, and control the massage actuator to perform adaptive massage for the user accordingly.

[0075] In the above process:

[0076] First, by synchronously collecting various physiological signals such as surface electromyography, skin conductance, heart rate variability, and body pressure distribution through a multi-sensor array, a multi-dimensional data foundation is provided for a comprehensive assessment of stress.

[0077] Next, dynamic feature extraction is performed in parallel for each signal to obtain instantaneous features reflecting the current intensity level, trend features reflecting the rate of change, and cumulative features reflecting the sustained load. These three types of features together constitute a dynamic and three-dimensional description of the physiological state.

[0078] Then, the dynamic features from these different signals are weighted and fused to calculate a comprehensive dynamic tension index, which can quantify the overall tension level of the human body.

[0079] Finally, the dynamic tension index and its corresponding dynamic characteristics are input into the fuzzy control rule base. By leveraging the advantages of fuzzy logic in handling uncertainty problems, precise, nonlinear control parameters are calculated, and the massage actuator is driven to complete adaptive massage.

[0080] The adaptive massage method provided by this invention constructs a complete, closed-loop adaptive massage control process. By introducing multimodal signals, it overcomes the shortcomings of one-sided assessment by a single physiological signal; by extracting three types of dynamic features—instantaneous, trend, and cumulative—it achieves comprehensive perception of tension states from static to dynamic and from the present to the past; by fusing them to form a dynamic tension index, it achieves precise quantification of complex physiological states; and finally, through a fuzzy control rule base, it achieves intelligent decision-making, making massage intervention no longer a rigid "on / off" response, but capable of smoothly and intelligently adapting to continuous changes in the body's tension state, significantly improving the accuracy and comfort of massage.

[0081] Therefore, the adaptive massage method and system based on the fusion of dynamic features of multimodal physiological signals provided by the present invention can solve the problems of incomplete evaluation, poor adaptability and lack of predictive intervention caused by existing intelligent massage devices relying on a single physiological signal and static threshold judgment.

[0082] In step S1 above, optionally, an adaptive massage system can be used as an example, such as a massage chair:

[0083] ① Use surface electromyography (sEMG) sensors that can be contacted by the user's target muscle groups such as the shoulders and back (located in the shoulder and back area of ​​the massage chair back) to collect surface electromyography signals (sEMG).

[0084] ② Use skin conductivity sensors (located on the armrest of the massage chair) that can be touched by the user's palm or fingertips to collect skin conductivity signals (SC);

[0085] ③ Collect heart rate signals using an optical heart rate sensor or contact electrodes (located on the back of the massage chair) and extract heart rate variability (HRV) signals from them;

[0086] ④ Use a pressure distribution sensor array (laid on the massage chair seat and backrest) to collect body pressure distribution signals (P) when the user is sitting or lying down.

[0087] In step S2 above, after preprocessing (such as filtering and denoising) each raw signal acquired in step S1, three types of dynamic features are extracted: instantaneous features, trend features, and cumulative features. The specific dynamic feature extraction method is as follows:

[0088] The instantaneous feature is a time-domain or frequency-domain feature value used to reflect the instantaneous intensity level of the signal;

[0089] The trend feature is a first-order derivative or regression slope value used to reflect the rate of change of the instantaneous feature;

[0090] The cumulative feature is an integral value used to reflect the sustained load that exceeds the resting reference value for the instantaneous feature.

[0091] The following sections will introduce the dynamic feature extraction methods for surface electromyography (sEMG), skin conductance (SC), heart rate variability (HRV), and body pressure distribution (P):

[0092] I. Dynamic Feature Extraction of Surface Electromyography (sEMG) Signals:

[0093] (1) Instantaneous feature extraction: Within a sliding time window of a specified length (e.g., 200ms), the root mean square value of the sEMG signal is calculated as the instantaneous feature I_EMG(t) to reflect the muscle activation intensity at the current moment: I_EMG(t) = RMS(sEMG(t)).

[0094] (2) Trend feature extraction: The instantaneous feature I_EMG(t) is numerically differentiated, and its rate of change is calculated as the trend feature D_EMG(t). In practical applications, the first-order difference approximation can be used:

[0095] D_EMG(t) ≈ [I_EMG(t) - I_EMG(t-Δt)] / Δt

[0096] Where Δt is the sampling interval.

[0097] (3) Cumulative feature extraction: Set a backtracking time T (e.g., 10 minutes), and calculate the instantaneous feature I_EMG(t) exceeding its resting baseline value (Threshold) within this time period. rest The integral area of ​​) is used as the cumulative feature S_EMG(t) to quantify the sustained load of muscle tension.

[0098]

[0099] in,

[0100] S_EMG(t): The cumulative characteristic value calculated at time t, whose physical meaning is the quantitative sum of muscle tension load;

[0101] This indicates an integration operation, where the integration interval is from the current time t back to a past time t−T of length T. This means the system only considers historical data within the most recent time period T.

[0102] I_EMG(t): The instantaneous eigenvalue of the integral variable at time t;

[0103] Threshold rest The user's resting baseline value represents the baseline electrical activity level of muscles in a relaxed state.

[0104] This function indicates that the instantaneous eigenvalue I_EMG(t) is only valid if it is higher than the resting reference value Threshold. rest Only when the threshold is reached is the excess portion (I_EMG(t) −Threshold) calculated. rest >0); if the instantaneous characteristic value is lower than or equal to the resting baseline value, then this contribution is 0. This ensures that only meaningful tension signals are accumulated, while ignoring basic physiological activities in the relaxed state; the design of the integral formula for the accumulated characteristic is particularly critical. The max(0,...) function ensures that only effective tension load is accumulated, while the integral based on the time window T accurately quantifies the duration and intensity of tension, effectively capturing the cumulative effects of fatigue and load, and providing a reliable data basis for identifying long-term tension states.

[0105] dt: Time integration variable.

[0106] It should be noted that, in this invention, the "resting baseline value" is an individualized reference standard defined for each physiological signal, reflecting the characteristic level of that signal when the user is in a relaxed state. This value is used to determine whether the real-time signal has deviated from the relaxation baseline when calculating cumulative characteristics, thereby ensuring that only effective stress loads are accumulated.

[0107] The resting reference value is obtained through a user calibration process, the main methods of which include, but are not limited to:

[0108] (1) Individualized calibration: When a user first uses the system or before each session, the system guides the user into a relaxed state and collects physiological signal data for a period of time. Based on this data, the resting baseline value of each signal is calculated. For example:

[0109] For surface electromyography (EMG) signals, the resting baseline value Threshold_rest_EMG can be the average value of the root mean square value of the signal during the relaxation phase plus twice the standard deviation.

[0110] For heart rate variability signals, the resting baseline Threshold_rest_HRV can be the average heart rate during the relaxation phase (e.g., 70 beats / minute) or the baseline value of its high-frequency power (HF).

[0111] For skin conductance signals, the resting reference value Threshold_rest_SC can be the minimum or average value of skin conductance level during the relaxation phase.

[0112] (2) Default value setting: When individualized calibration is not possible, the system can enable the default resting baseline value based on population statistics.

[0113] By employing individualized, sub-signal resting baseline values, the accuracy and personalization of stress assessment are ensured, effectively eliminating interference from individual differences and baseline physiological fluctuations.

[0114] Furthermore, in practical digital signal processing systems, Continuous integrals can be approximated by discrete summation, i.e.:

[0115] S_EMG(t) =

[0116] Where n represents the current sampling point; N corresponds to the number of sampling points within the backtracking time window T; Δt is the sampling time interval; k is the summation variable, representing the sequence number of each sampling point traversed within the backtracking time window, and the value of k changes from (nN) to the current point n.

[0117] II. Dynamic Feature Extraction of Skin Conductivity (SC) Signals

[0118] Skin conductance signals directly reflect the level of sympathetic nerve excitation, and the signal manifests as fluctuations at the baseline level.

[0119] (1) Extraction of instantaneous feature I_SC(t): The instantaneous feature is the instantaneous amplitude value of the skin conductance signal. This value can be obtained by directly reading the signal value after filtering and smoothing, and is used to reflect the instantaneous excitation level of the sympathetic nervous system. The higher the I_SC(t) value, the higher the level of psychological stress or alertness.

[0120] (2) Extraction of trend feature D_SC(t): The instantaneous feature I_SC(t) is differentiated over time, i.e., D_SC(t) ≈ d[I_SC(t)] / dt. This trend feature reflects the rate of change of skin conductance level. D_SC(t) > 0 indicates that the excitation level is rising, which may indicate that stress or tension is accumulating rapidly.

[0121] (3) Extraction of cumulative feature S_SC(t): The cumulative feature S_SC(t) is obtained by calculating the integral of the instantaneous feature I_SC(t) exceeding its resting baseline value (Threshold_rest_SC) within a backtracking time T. Its discrete calculation formula is as follows:

[0122]

[0123] in,

[0124] S_SC[n]: The cumulative characteristic value of the skin conductance signal calculated at the discrete sampling point sequence number n. Its physical meaning is the total physiological load generated by psychological stress or sympathetic nerve excitation in a recent period. This value is non-negative; a larger value indicates a heavier cumulative psychological stress load. This cumulative characteristic quantifies the duration and intensity of psychological stress, reflecting long-term sympathetic nerve load.

[0125] The summation symbol represents the accumulation of values ​​over N+1 sampling points from sequence number k=n−N to the current sequence number k=n. This defines a backtracking time window of length N;

[0126] k: The summation index variable. It takes all integer values ​​from n−N to n in sequence during the summation process, used to iterate through each discrete sampling point within the time window.

[0127] n: The sequence number of the current sampling point. Represents the latest data point at the current moment.

[0128] N: The total number of sampling points contained within the backtracking time window T. Its value is determined by the formula N=T / Δt. For example, if the time window T=300 seconds and the sampling interval Δt=0.1 seconds, then N=300 / 0.1=3000 points.

[0129] n−N: The sequence number of the starting point of the backtracking time window. That is, the starting point for calculating the cumulative load.

[0130] The core expression to be accumulated:

[0131] The instantaneous characteristic value of the skin conductance signal calculated at sampling point k, i.e., the instantaneous amplitude of the signal;

[0132] Threshold_rest_SC: The resting baseline value of the skin conductance signal, representing the basic skin conductance level when the user is in a relaxed state;

[0133] max(0,...): The function that retrieves the maximum value. This function ensures that the maximum value is retrieved only if the instantaneous eigenvalue is... The excess portion (positive value) is only calculated when the instantaneous value is higher than the resting baseline value (Threshold_rest_SC); if the instantaneous value is lower than or equal to the resting baseline value, this contribution is 0. This ensures that only meaningful excitation loads exceeding the baseline level are accumulated.

[0134] Δt: Sampling time interval. That is, the time length between two consecutive samples (unit: seconds). Multiplying the load increment at each point by Δt is equivalent to calculating the load "area" within each tiny time interval, so that the final summation result S_SC[n] has the dimension of [signal strength] × [time], which correctly represents the cumulative load.

[0135] This formula approximates the total "area" of skin conductance levels that consistently exceed their resting baseline value within the most recent time window T by using discrete summation. This total area represents the quantified cumulative psychological stress load.

[0136] III. Dynamic Feature Extraction of Heart Rate Variability (HRV) Signals

[0137] Heart rate variability reflects the balance of the autonomic nervous system, and features need to be extracted from the heart rate signal (e.g., through frequency domain analysis).

[0138] (1) Extraction of instantaneous feature I_HRV(t): The instantaneous feature is the ratio of low-frequency power to high-frequency power of heart rate variability (LF / HF Ratio). This ratio is a key indicator for assessing autonomic nervous system balance, and its instantaneous value I_HRV(t) directly characterizes the instantaneous balance between sympathetic and parasympathetic nervous system activity. An increased ratio usually indicates that sympathetic nervous system activity is dominant (state of tension).

[0139] (2) Extraction of trend feature D_HRV(t): The instantaneous feature I_HRV(t) is differentiated over time, i.e., D_HRV(t) ≈ d[I_HRV(t)] / dt. This trend feature reflects the rate of change of the autonomic nervous system balance state. D_HRV(t) > 0 indicates that the balance state is accelerating towards the direction of sympathetic dominance.

[0140] (3) Extraction of cumulative feature S_HRV(t): The cumulative feature S_HRV(t) is obtained by calculating the integral of the instantaneous feature I_HRV(τ) exceeding its resting baseline value (Threshold_rest_HRV, representing the baseline of the equilibrium state) over a backtracking time T. Its calculation formula is as follows:

[0141]

[0142] in,

[0143] S_HRV[n]: The cumulative characteristic value of the heart rate variability signal calculated at the discrete sampling point sequence number n. Its physical meaning is the total load of the autonomic nervous system (sympathetic dominance) imbalance state over a recent period. This value is non-negative; a larger value indicates a longer and more severe period of sympathetic dominance, i.e., a higher cumulative neural stress load. This cumulative characteristic quantifies the duration of the autonomic nervous system imbalance state.

[0144] I_HRV[k]: The instantaneous characteristic value of the heart rate variability signal calculated at sampling point k, i.e., the ratio of low-frequency power to high-frequency power (LF / HF Ratio). An increase in this ratio indicates a relative increase in sympathetic nerve activity.

[0145] Threshold_rest_HRV: The resting baseline value of the heart rate variability signal, representing the baseline level of autonomic balance when the user is in a relaxed state (the baseline of the LF / HF ratio).

[0146] `max(0,...)`: The maximum value function. This function ensures that the excess (positive value) is calculated only when the instantaneous characteristic value `I_HRV[k]` (i.e., the LF / HF ratio) is higher than the resting baseline value `Threshold_rest_HRV`, indicating a meaningful sympathetic dominance; if the instantaneous value is lower than or equal to the resting baseline value, this contribution is 0. This guarantees that only the load of autonomic imbalance is accumulated.

[0147] The meanings of the other symbols are the same as those in the aforementioned discrete calculation formula.

[0148] This formula approximates the total "area" of the LF / HF ratio of heart rate variability exceeding its resting equilibrium baseline within the most recent time window T by using discrete summation. This total area is the quantified cumulative autonomic imbalance load.

[0149] IV. Dynamic Feature Extraction of Body Pressure Distribution Signal (P)

[0150] Body pressure distribution signals are acquired through a pressure sensor array and used to assess user posture.

[0151] (1) Extraction of instantaneous feature I_P(t): The instantaneous feature is an index of the asymmetry of body pressure distribution, for example, calculating the absolute difference between the total pressure on the left and right halves of the body. This index I_P(t) directly quantifies the degree of instantaneous asymmetry of sitting posture. The larger the I_P(t) value, the more serious the poor posture.

[0152] (2) Extraction of trend feature D_P(t): The instantaneous feature I_P(t) is differentiated over time, i.e., D_P(t) ≈ d[I_P(t)] / dt. This trend feature reflects the deterioration or improvement trend of poor posture. D_P(t) > 0 indicates that the body is tilting at an accelerating rate.

[0153] (3) Extraction of cumulative feature S_P(t): The cumulative feature S_P(t) is obtained by calculating the integral of the instantaneous feature I_P(τ) exceeding its resting reference value (Threshold_rest_P, representing the reference of a good symmetrical posture) within a backtracking time T. The calculation formula is as follows:

[0154]

[0155] S_P[n]: The cumulative characteristic value of the body pressure distribution signal calculated at the discrete sampling point sequence number n. Its physical meaning is the total static load caused by poor posture (measured by the asymmetry of body pressure distribution) over a recent period. This value is non-negative; a larger value indicates a longer cumulative time the user maintains the poor posture or a more severe asymmetry, resulting in a higher static load on the musculoskeletal system. This cumulative characteristic is essentially the cumulative time of the poor posture, directly related to the tension and fatigue of the musculoskeletal system caused by long-term static load.

[0156] I_P[k]: The instantaneous characteristic value of the body pressure distribution signal calculated at sampling point k, i.e., the body pressure asymmetry index (such as the absolute difference between the total pressure on the left and right halves of the body). The larger this value, the more asymmetrical the instantaneous posture.

[0157] Threshold_rest_P: The resting baseline value of the body pressure distribution signal, representing the baseline level of the body pressure asymmetry index when the user is in a good, symmetrical sitting posture.

[0158] `max(0,...)`: The maximum value function. This function ensures that the excess (positive value) is calculated only when the instantaneous characteristic value I_P[k] (i.e., the asymmetry index) is higher than the resting baseline value Threshold_rest_P, indicating a meaningful undesirable posture; if the instantaneous value is lower than or equal to the resting baseline value, this contribution is 0. This guarantees that only the load caused by undesirable posture is accumulated.

[0159] This formula approximates the total "area" of body pressure asymmetry index that continuously exceeds its good posture baseline within the most recent time window T by using discrete summation. This total area is the quantified cumulative poor posture load, which is directly related to muscle tension and fatigue caused by prolonged static posture.

[0160] In step S3 above, the dynamic tension index is calculated using the following formula:

[0161] DTI(t) = K1*I(t) + K2*F_D(D(t)) + K3*F_S(S(t))

[0162] DTI(t) is the dynamic tension index at time t;

[0163] I(t) is the comprehensive value of the instantaneous features at time t, which is obtained by weighted summation of the instantaneous features corresponding to each of the physiological signals;

[0164] D(t) is the comprehensive value of the trend characteristics at time t, which is obtained by weighted summation of the trend characteristics corresponding to each of the physiological signals;

[0165] S(t) is the cumulative feature value at time t, which is obtained by weighted summation of the cumulative features corresponding to each physiological signal;

[0166] K1, K2, and K3 are weighting coefficients.

[0167] F_D is a nonlinear function that performs saturation processing on the composite value of trend characteristics, and F_S is a nonlinear function that performs marginal effect processing on the composite value of cumulative characteristics.

[0168] The three characteristic values ​​I(t), D(t), and S(t) mentioned above refer to a comprehensive scalar value that represents the overall instantaneous tension level after normalizing and weighting the time-domain or frequency-domain characteristic values ​​of all the individual signals mentioned above.

[0169] For example, the instantaneous feature composite value I(t) = α1 * I_EMG(t) + α2 * I_SC(t) + α3 * I_HRV(t) + α4 * I_P(t) (where α1-α4 are definable weight ratios).

[0170] The DTI(t) calculation model does not simply linearly add the combined values ​​of the three feature classes (I(t), D(t), S(t)), but introduces targeted nonlinear functions for processing. Specifically, F_D is a nonlinear function designed to saturate the trend feature value D(t), aiming to constrain the infinite amplification of the trend value; F_S is a nonlinear function designed to handle the marginal effect of the cumulative feature value S(t), aiming to simulate the physiological effect of diminishing sensitivity to continuous stimulation. Weighting coefficients K1, K2, and K3 are used to adjust the contribution of the three feature classes to the final index.

[0171] By introducing nonlinear functions F_D and F_S, the calculation of the dynamic tension index is made more consistent with physiological and psychological principles. Saturation processing makes the system insensitive to slight fluctuations in the signal (noise), enhancing the model's robustness while still enabling a clear response to significant trends. Marginal effect processing allows the model to better reflect real human feelings, where the initial accumulation of tension has the greatest impact, and the additional impact of subsequent accumulations gradually weakens. This makes the calculated dynamic tension index not only a mathematical indicator but also a physiological measure of tension that closely reflects real human feelings.

[0172] 1. Saturation processing function F_D for trend characteristics

[0173] Functional Purpose: "Saturation Processing" aims to prevent unreasonable large jumps in the Dynamic Tension Index (DTI) due to drastic or instantaneous fluctuations in the trend value D(t), thereby enhancing the system's stability and anti-interference capabilities. It simulates the physiological characteristic that the human body has an upper limit to its perception of the rate of change.

[0174] Specific examples:

[0175] Optionally, the function F_D is the arctangent function: F_D(D(t)) = arctan(C * D(t)), where C is a constant for adjusting the sensitivity;

[0176] Working process: Suppose that due to sudden fright, the user's muscle tension tendency D(t) instantly becomes very large (e.g., D(t) = 10). Without saturation treatment, its contribution K2·D(t) will be very large, causing the massage intensity to spike to an unnecessary or even uncomfortable level.

[0177] The effect of saturation: When using the arctan function, regardless of the size of the input D(t), the output value of arctan(C⋅D(t)) will be limited to a finite interval of (−2 / π, 2 / π). For example, let C=0.5:

[0178] When D(t) = 2 (moderate increase), arctan(1) ≈ 0.79;

[0179] When D(t) = 10 (a sharp increase), arctan(5) ≈ 1.37;

[0180] As D(t)→∞, arctan(C⋅D(t))→1.57;

[0181] Beneficial effects: The input D(t) increases fivefold from 2 to 10, but after processing by F_D, its contribution to DTI only increases from 0.79 to 1.37, a significant slowdown in the increase. This ensures that the system can respond sensitively to changing trends without being "biased" by extreme values, outputting stable and reasonable control commands.

[0182] 2. Marginal effect treatment function F_S for cumulative characteristics

[0183] Functional Purpose: "Marginal Effect Processing" aims to simulate the physiological and psychological phenomenon that the human body's sensitivity to continuous stimulation gradually decreases (i.e., "diminishing marginal effect"). Initial accumulation of tension can make people feel very tired, but the same amount of additional accumulation will result in a smaller increase in subjective feelings on top of already being very tired.

[0184] Specific examples:

[0185] Optionally, the function F_S is a logarithmic function: F_S(S(t)) = log(1 + S(t)).

[0186] Working process: Assume that the cumulative feature S(t) quantifies the continuous load on the muscle (unit: microvolts × seconds).

[0187] The effect of marginal processing: The characteristic of logarithmic functions is that the larger the input value, the slower the output value grows.

[0188] When S(t) = 1 (initial load), log(2) ≈ 0.69;

[0189] When S(t) = 10 (the load increases by 10 times), log(11) ≈ 2.40 (the load increases by about 3.5 times);

[0190] When S(t) = 100 (the load increases by 10 times), log(101) ≈ 4.62 (the load increases by less than 2 times);

[0191] Beneficial effects: Increasing the load from 1 to 10 increased the subjective perception index (function output) by approximately 1.71; while increasing it from 10 to 100, with the same increase of 90 units of objective load, only increased the subjective perception index by approximately 2.22. This accurately simulates the marginal effect of "the more tired you are, the more numb you become." This ensures that the massage system does not linearly increase the massage intensity to an unbearable level when the user is already extremely fatigued, but instead provides a gentler, more comfortable intervention intensity.

[0192] In summary, by combining the saturation treatment of F_D and the marginal effect treatment of F_S, the model (DTI) described in this invention elevates a primary model based on linear weighted fusion into an intelligent model capable of simulating complex human response characteristics. This ensures that the final calculated dynamic tension index is not merely a mathematically weighted sum, but a high-level quantification of tension that is physiologically reasonable, stable in control, and comfortable for the user experience.

[0193] Regarding step S4 above, the fuzzy control rule base contains multiple fuzzy logic rules based on dynamic tension index and dynamic features. These fuzzy logic rules include at least:

[0194] Predictive intervention rule: When the instantaneous characteristic composite value is "medium", the trend characteristic composite value is "rapidly rising", and the cumulative characteristic composite value is "high", it indicates that the user's tension may develop to a very high level in the short term. At this time, the system will not passively wait, but will actively trigger a high-intensity, high-frequency massage, aiming to intervene before the tension fully erupts. Therefore, it will actively control the massage actuator to perform the massage with high intensity and high frequency.

[0195] Soothing and Recovery Rules: When the instantaneous feature composite value is "low", the trend feature composite value is "slowly decreasing", and the cumulative feature composite value is "extremely high", it indicates that the user is in a fatigue recovery period. The system should use low-intensity, slow-paced massage to promote recovery, rather than stimulating intervention. Therefore, actively control the massage actuator to perform massage with low intensity and slow pace.

[0196] Furthermore, high threshold intervals, medium threshold intervals, and low threshold intervals can be preset, and the high, medium, or low value of the feature composite value can be determined based on the threshold interval into which the feature composite value falls.

[0197] Predictive intervention rules demonstrate the system's forward-looking approach, enabling it to "prevent problems before they arise" and intervene earlier, greatly enhancing the user experience. Soothing and recovery rules reflect the system's deep understanding of the user's state, distinguishing different types of tension and adopting the most appropriate massage strategies. Together, these rules transform the massage system from a simple actuator into an intelligent agent capable of deeply understanding the user's state and providing personalized, intelligent services.

[0198] Example 2

[0199] This embodiment provides an adaptive massage system for performing any of the adaptive massage methods described in Embodiment 1, possessing the same functions and beneficial effects, specifically including:

[0200] The multi-sensor array 1 is used to collect various physiological signals from the user in real time, including at least surface electromyography signals, skin conductance signals, heart rate variability signals, and body pressure distribution signals.

[0201] The dynamic feature extraction module 2 is used to extract dynamic features from each physiological signal to obtain instantaneous features, trend features, and cumulative features corresponding to the signal. The instantaneous features reflect the instantaneous intensity level of the signal. The trend features are obtained by differentiating the instantaneous features or calculating their short-time linear regression slope to reflect the rate of change of the signal. The cumulative features are obtained by calculating the integral of the instantaneous features exceeding the resting baseline value over a period of time to reflect the sustained load of the signal.

[0202] The dynamic tension index calculation module 3 is used to perform weighted fusion based on the dynamic characteristics of the extracted multiple physiological signals to calculate and generate a comprehensive dynamic tension index.

[0203] The fuzzy adaptive control module 4 has a built-in fuzzy control rule library, which is used to input the dynamic tension index and its corresponding dynamic features into the fuzzy control rule library, and the fuzzy control rule library calculates the control parameters of the massage actuator, and controls the massage actuator to perform adaptive massage on the user accordingly.

[0204] Massage actuator 5 is configured to perform adaptive massage actions according to the control parameters.

[0205] Optionally, the massage actuator 5 includes at least one of the following:

[0206] Servo motors are used to drive the massage heads to perform kneading and tapping motions.

[0207] The air pump and solenoid valve assembly are used to control the inflation and deflation of the airbag to achieve squeezing massage.

[0208] After the user activates the massage system, the multi-sensor array continuously collects physiological signals. The modules within the signal processing and computation unit operate sequentially: the dynamic feature extraction module calculates various dynamic features in real time; the dynamic tension index calculation module periodically (e.g., once per second) fuses features and updates the DTI value; the fuzzy adaptive control module calculates control parameters in real time based on the latest DTI and dynamic features, driving the massage actuator to adjust its movements. This cycle repeats, forming a closed-loop adaptive massage control system.

[0209] Based on Example 1, features not explained in this example will be explained using the methods described in Example 1, and will not be repeated here.

[0210] Finally, it should be noted that although the above embodiments have been described in the text and drawings of this application, this should not limit the scope of patent protection of this application. Any technical solutions that are based on the essential concept of this application and utilize the content described in the text and drawings of this application, resulting in equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this application.

Claims

1. An adaptive massage method based on multi-modal physiological signal dynamic feature fusion, characterized in that, Comprising: Step S1: Collecting multiple physiological signals of a user in real time through a multi-sensor array, the physiological signals at least including surface electromyography signals, skin conductance signals, heart rate variability signals and body pressure distribution signals; Step S2: Extracting dynamic features for each physiological signal respectively, obtaining instantaneous features, trend features and cumulative features corresponding to the signal; the instantaneous features are used to reflect the instantaneous intensity level of the signal, the trend features are obtained by deriving or calculating the short-time linear regression slope of the instantaneous features, and are used to reflect the change rate of the signal, and the cumulative features are obtained by calculating the integral of the instantaneous features exceeding a resting baseline value within a backtracking time, and are used to reflect the sustained load of the signal; Step S3: Based on the extracted dynamic features of multiple physiological signals, performing weighted fusion to calculate a comprehensive dynamic tension index; Step S4: Inputting the dynamic tension index and its corresponding dynamic features into a fuzzy control rule base, obtaining control parameters of a massage execution mechanism by solving the fuzzy control rule base, and controlling the massage execution mechanism to perform adaptive massage on the user according to the control parameters.

2. The adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 1, characterized in that: the instantaneous features are time-domain or frequency-domain feature values used to reflect the instantaneous intensity level of the signal; the trend features are first-order differential or regression slope values used to reflect the change rate of the instantaneous features; the cumulative features are integral values used to reflect the sustained load of the instantaneous features exceeding a resting baseline value; wherein the resting baseline value is a reference baseline defined for each physiological signal, reflecting the signal feature level when the user is in a relaxed state.

3. The adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 2, characterized in that: within a specified length of sliding time window, the root mean square value of all collected surface electromyography signals (sEMG) is calculated as the instantaneous feature I_EMG(t) of the surface electromyography signals (sEMG).

4. The adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 2, characterized in that: the amplitude value of the skin conductance signal (SC) after filtering and smoothing processing is taken as the instantaneous feature I_SC(t) of the skin conductance signal (SC).

5. The adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 2, characterized in that: the ratio (LF / HF Ratio) of the low-frequency power to the high-frequency power of the heart rate variability is taken as the instantaneous feature I_HRV(t) of the heart rate variability signal (HRV).

6. The adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 2, characterized in that: the asymmetry index of the body pressure distribution is taken as the instantaneous feature I_P(t) of the body pressure distribution signal (P).

7. The self-adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 2, characterized in that, The dynamic tension index in step S3 is calculated by the following formula: DTI(t) = K1 * I(t) + K2 * F_D(D(t)) + K3 * F_S(S(t)); wherein, DTI(t) is a dynamic tension index at time t; I(t) is an instantaneous feature integrated value at time t, which is obtained by weighting and summing the instantaneous features corresponding to each of the physiological signals; D(t) is a trend feature integrated value at time t, which is obtained by weighting and summing the trend features corresponding to each of the physiological signals; S(t) is a cumulative feature integrated value at time t, which is obtained by weighting and summing the cumulative features corresponding to each of the physiological signals; K1, K2, K3 are weight coefficients, F_D is a nonlinear function for saturation processing of the trend feature integrated value, and F_S is a nonlinear function for marginal effect processing of the cumulative feature integrated value.

8. The self-adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 7, characterized in that, the function F_D is an arctangent function: F_D(D(t)) = arctan(C * D(t)), wherein C is a constant for adjusting sensitivity; the function F_S is a logarithmic function: F_S(S(t)) = log(1 + S(t)).

9. The self-adaptive massage method based on multi-modal physiological signal dynamic feature fusion according to claim 7, characterized in that, The fuzzy control rule base in step S4 contains multiple fuzzy logic rules based on the dynamic tension index and dynamic features, and the fuzzy logic rules at least include: a predictive intervention rule: when the instantaneous feature integrated value is "medium", the trend feature integrated value is "rapidly rising", and the cumulative feature integrated value is "high", the massage execution mechanism is controlled to perform massage at high intensity and high frequency rhythm; a soothing recovery rule: when the instantaneous feature integrated value is "low", the trend feature integrated value is "slowly falling", and the cumulative feature integrated value is "extremely high", the massage execution mechanism is controlled to perform massage at low intensity and slow rhythm.

10. An adaptive massage system for performing the adaptive massage method of any one of claims 1-9, characterized by, It comprises: a multi-sensor array for real-time acquisition of multiple physiological signals of a user, the physiological signals at least including surface electromyography signals, skin conductance signals, heart rate variability signals, and body pressure distribution signals; a dynamic feature extraction module for extracting dynamic features of each physiological signal respectively, to obtain instantaneous features, trend features, and cumulative features corresponding to the signal; the instantaneous features are used to reflect the instantaneous intensity level of the signal, the trend features are obtained by differentiating or calculating the short-time linear regression slope of the instantaneous features, and are used to reflect the change rate of the signal, and the cumulative features are obtained by calculating the integral of the instantaneous features exceeding the resting baseline value within a backtracking time, and are used to reflect the sustained load of the signal; a dynamic tension index calculation module for calculating a comprehensive dynamic tension index by weighted fusion based on the extracted dynamic features of multiple physiological signals; a fuzzy adaptive control module with a fuzzy control rule base, for inputting the dynamic tension index and its corresponding dynamic features into the fuzzy control rule base, and obtaining the control parameters of the massage execution mechanism by solving the fuzzy control rule base, and accordingly controlling the massage execution mechanism to perform adaptive massage on the user. A massage actuator configured to perform an adaptive massage action in accordance with the control parameters. A massage actuator configured to perform an adaptive massage action in accordance with the control parameters.

Images