Equipment self-adaptive regulation and control system based on multi-mode neural perception and user feedback

Abstract

The invention discloses an equipment adaptive regulation and control system based on multi-modal neural perception and user feedback, which realizes millisecond-level adaptive regulation of equipment parameters through a multi-modal neural perception feedback and dynamic analysis model, solves the problems of insufficient dynamic evaluation, algorithm lag and the like of traditional equipment, and improves the equipment performance. The user experience (the pressure comfort is improved by 40%, and the temperature stability reaches + / -0.3 DEG C) and the energy efficiency (the energy is saved by 21.3%) are remarkably improved, the response speed is improved by 52.7% compared with a traditional scheme, the method is suitable for scenes such as massage chairs and smart homes, key indexes are improved by more than 40% compared with the prior art through CNAS authentication, and the method has remarkable technical advancement and market value and is worthy of popularization and application. The problems that on one hand, an existing massage equipment regulation and control system lacks dynamic evaluation on real-time states of muscle tension and fatigue degree of a user, and on the other hand, change of nonlinear characteristics of physiological signals is difficult to deal with are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and intelligent control, and particularly to a device adaptive regulation system based on user pain feedback and multimodal neural perception feedback. Background Art

[0002] Traditional massage devices mainly rely on preset programs for operation, and have the following defects:

[0003] Lack of personalization: There is a lack of dynamic assessment of the real-time state of the user's muscle tension and fatigue level;

[0004] Algorithm lag: Existing devices mostly use fixed rules for control and are difficult to cope with the non-linear characteristics of changing physiological signals;

[0005] For example, there are massage machines proposed in existing patents based on pressure feedback, but multi-modal data is not integrated; there are those that propose introducing electromyogram signals as a reference, but a dynamic feedback mechanism is not established either. Therefore, there is an urgent need for an adaptive regulation system that integrates modal perception and intelligent algorithms. Summary of the Invention

[0006] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this part, the abstract, and the title. However, such simplifications or omissions cannot be used to limit the scope of the present invention.

[0007] In view of the problems existing in the above-mentioned existing massage device regulation system, the present invention is proposed.

[0008] Therefore, the technical problem solved by the present invention is: to solve the problem that the existing massage device regulation system lacks dynamic assessment of the real-time state of the user's muscle tension and fatigue level on the one hand, and is difficult to cope with the non-linear characteristics of changing physiological signals on the other hand.

[0009] To solve the above technical problem, the present invention provides the following technical solution: An adaptive regulation system for a device based on multimodal neural perception and user feedback, including the following components: a data signal acquisition module that acquires signal data during the massage process based on a configured sensor array; a data processing module that is wirelessly connected to the data signal acquisition module to obtain each signal data; a data analysis module that is data-connected to the data processing module, establishes an analysis model based on the processed signal data, inputs each signal data, outputs an analysis value, and transmits the analysis value to an execution module; and an execution module that adaptively adjusts relevant parameters according to the analysis value output by the data analysis module.

[0010] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: the signal data collected by the data signal acquisition module specifically includes: pressure sensing data: pressure sensors are distributed on the surface of the massage head, with a sampling frequency ≥ 100 Hz, for detecting muscle contact pressure and obtaining the pressure sensing data; temperature sensing data: temperature sensors are embedded inside the massage pad to monitor the skin surface temperature in real time, with an accuracy of ±0.1 °C, for obtaining the temperature sensing data; electromyogram (EMG) signal sensing data: EMG signal sensors collect the EMG signals of the target muscle group through electrode patches, with a bandwidth of 5 Hz - 500 Hz, for obtaining the EMG signal sensing data; among them, a set of pressure sensors, temperature sensors, and EMG signal sensors are configured in each muscle group.

[0011] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: the data processing module specifically includes: after collecting the EMG signal sensing data, conditioning steps such as high-pass filtering, high-gain amplification, and low-pass filtering are required.

[0012] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: based on the processed signal data, the analysis module establishes the analysis model and outputs the analysis value, which specifically includes the following steps: S1: During the detection period of the EMG signal sensor, the EMG signal voltage values δ are collected at uniform intervals; S2: A two-dimensional coordinate system is constructed, the collected EMG signal voltage values δ are input, and each reference point is obtained. Based on each reference point, a time-voltage fluctuation curve is obtained; S3: The analysis model is established, and the fluctuation characteristic value η is extracted based on the time-voltage fluctuation curve; S4: The fluctuation characteristic value η is defined as the analysis value.

[0013] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: when collecting the EMG signal voltage values δ, they are obtained at intervals of 1 ms; when establishing the time-voltage fluctuation curve, the time is used as the horizontal axis and the voltage is used as the vertical axis to depict discrete points, and each reference point is sequentially connected by a continuous and smooth curve to form the time-voltage fluctuation curve.

[0014] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: the established analysis model is specifically:

[0015]

[0016] Among them, η is the analysis value; k i is the transformation slope at the i-th reference point; n is the number of reference points; δi is the ordinate at the i-th reference point; 1.02 is the robust adjustment constant.

[0017] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: the specific steps of the execution module for adaptively adjusting relevant parameters according to the analysis value include: regulating the applied pressure according to the analysis value, and the magnitude of the applied pressure is sensed by a pressure sensor; regulating the applied temperature according to the analysis value, and the magnitude of the applied temperature is sensed by a temperature sensor.

[0018] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: the regulation of the applied pressure according to the analysis value is specifically based on the following model:

[0019]

[0020] Wherein, a' is the adjusted applied pressure, N; a is the applied pressure before adjustment, N; η is the analysis value.

[0021] As a preferred embodiment of the device adaptive regulation system based on multi-modal neural perception and user feedback according to the present invention, wherein: the regulation of the applied temperature according to the analysis value is specifically based on the following model:

[0022]

[0023] Wherein, b' is the adjusted applied temperature, °C; b is the applied temperature before adjustment, °C; η is the analysis value; t is the temperature of the current environment, °C.

[0024] Advantages of the present invention: The present invention provides a device adaptive regulation system based on multi-modal neural perception and user feedback. Through multi-modal neural perception feedback and dynamic analysis models, it realizes millisecond-level adaptive adjustment of device parameters, solves problems such as insufficient dynamic evaluation and algorithm lag in traditional devices, significantly improves the user experience (the pressure comfort is increased by 40%, and the temperature stability reaches ±0.3 °C) and energy efficiency (energy saving of 21.3%), the response speed is increased by 52.7% compared with traditional solutions, is applicable to scenarios such as massage chairs and smart homes, and the key indicators are improved by more than 40% compared with the prior art after being certified by CNAS, with significant technological advancement and market value, and solves the problems that on the one hand, the existing massage device regulation system lacks dynamic evaluation of the real-time state of the user's muscle tension and fatigue degree, and on the other hand, it is difficult to cope with the non-linear characteristics of changing physiological signals. Description of the Drawings

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0026] Figure 1 It is a system module diagram of the device adaptive regulation system based on multi-modal neural perception and user feedback provided by the present invention.

[0027] Figure 2 It is a method flow chart for the data analysis module provided by the present invention to establish an analysis model and output analysis values based on the processed signal data. Specific Embodiments

[0028] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0029] The existing massage device regulation system has problems in that on the one hand, it lacks dynamic assessment of the real-time state of the user's muscle tension and fatigue degree, and on the other hand, it is difficult to cope with the non-linear characteristics of changing physiological signals.

[0030] Specifically, the defects of the existing device regulation system (with experimental data):

[0031] (1) Parameter curing problem:

[0032] Comparative Experiment 1: In the test of 10 subjects for the traditional massage chair fixed program group:

[0033] Pressure distribution matching degree: 62.3% ± 8.7%;

[0034] Frequency of active user adjustment: Manual intervention is required on average every 8.2 minutes;

[0035] Energy consumption fluctuation range: 35 - 72W (volatility rate 103%);

[0036] (2) Defects in the regulation algorithm:

[0037] Comparative Experiment 3: In the dynamic load test of traditional PID control:

[0038] Overshoot: 28.6% (industry standard ≤ 15%);

[0039] Adjustment time: 1.8 s (target value 0.3 s);

[0040] Anti-interference ability: fails when the load mutates by ±10%.

[0041] Therefore, the present invention provides a device adaptive regulation system based on multi-modal neural perception and user feedback, specifically an adaptive regulation system that integrates bioelectrical signal analysis and physical constraint optimization, and is applicable to scenarios such as massage chairs (Example 1), smart homes (Example 2), industrial robots (Example 3), etc.

[0042] Refer to Figure 1 , the device adaptive regulation system based on multi-modal neural perception and user feedback includes the following components:

[0043] The data signal acquisition module 100 collects signal data during the massage process based on the configured sensor array;

[0044] The data processing module 200 is wirelessly connected to the data signal acquisition module 100 to obtain each signal data;

[0045] The data analysis module 300 is data-connected to the data processing module 200. Based on the processed signal data, an analysis model is established, the signal data is input, an analysis value is output, and the analysis value is transmitted to the execution module;

[0046] The execution module 400 adaptively adjusts relevant parameters according to the analysis value output by the data analysis module 300.

[0047] Specifically, the signal data collected by the data signal acquisition module 100 specifically includes:

[0048] Pressure sensing data: Pressure sensors are distributed on the surface of the massage head, and the sampling frequency ≥ 100 Hz, which is used to detect the muscle contact pressure and obtain the pressure sensing data;

[0049] Temperature sensing data: Temperature sensors are embedded inside the massage pad to continuously monitor the skin surface temperature with an accuracy of ±0.1 °C to obtain the temperature sensing data;

[0050] Electromyogram signal sensing data: Electromyogram signal sensors collect the electromyogram signals (EMG) of the target muscle groups through electrode patches, with a bandwidth of 5 Hz - 500 Hz, to obtain the electromyogram signal sensing data;

[0051] Among them, a set of pressure sensors, temperature sensors, and electromyogram signal sensors are configured in each muscle group.

[0052] It should be noted that the pressure sensors, temperature sensors, and electromyogram signal sensors involved in the present invention are all direct applications of existing conventional electronic sensors, and no redundant description will be given here.

[0053] Furthermore, the data processing module 200 specifically includes: after collecting the sEMG sensing data, conditioning steps such as high-pass filtering, high-gain amplification, and low-pass filtering are required.

[0054] It should be noted that: surface electromyogram (sEMG) signal is the electrical signal accompanied by muscle contraction, and it is an important method for non-invasive detection of muscle activity on the body surface. Its application background mainly focuses on two major fields: rehabilitation medicine and sports science.

[0055] With the development of detection technology and signal processing means, using sEMG to replace needle electrode EMG for comprehensive clinical non-destructive diagnosis has become one of the hot issues in biomedical and medical research. The sEMG of the human body is very weak, vulnerable to interference, and difficult to measure. How to effectively collect and extract sEMG has become one of the key technologies for sEMG application.

[0056] Most of the applied sEMG acquisition devices transmit the collected sEMG to a data acquisition card or a specific medical instrument through wires for analysis and processing. The sampling electrodes are pasted on the skin surface, and the electrode power supply indirectly comes from the mains. The significant deficiencies of this method are as follows: (1) Special instruments are required to complete signal acquisition, and the application scenarios are limited. For example, it is impossible to collect the sEMG of athletes' limbs during sprinting; (2) Due to the connection of wires, when the limb moves, the relative movement between the electrode and the skin is likely to occur, affecting the acquisition results; (3) The power supply is obtained after processing the mains, introducing a large amount of power frequency interference and increasing the difficulty of data processing; (4) If the wires are long, it is very easy to introduce other interferences in the environment, such as electromagnetic interference, etc.

[0057] After reviewing the prior art, the present invention also provides a miniature, wireless, battery-powered sEMG acquisition device with data storage function, which preferably solves the problems commonly existing in current sEMG acquisition devices.

[0058] Hardware Circuit Design

[0059] According to neurophysiological knowledge, muscle action potential will generate a potential difference of -90 mV to 30 mV. Since the human body is a poor conductor of electricity (with an internal resistance of the order of 1 MΩ), only a peak value of about 1 mV can be obtained from the patch electrodes on the body surface. According to the literature, sEMG often mixes with very low-frequency (close to direct current) and high-frequency interference signals, and the effective sEMG signal spectrum is distributed between 10 - 500 Hz. Therefore, the signal detected from the patch electrodes needs to go through signal conditioning processes such as high-pass filtering (DC blocking), high-gain amplification, and low-pass filtering (filtering out high-frequency interference).

[0060] The sampling electrodes of the SEMG acquisition device adopt Ag / AgCl surface electrodes, which have a small polarization voltage and can quickly obtain stable myoelectric signals. In the present invention, one electrode is composed of three mutually parallel Ag / AgCl electrodes. The middle electrode is the reference ground terminal, and the two on both sides form differential input terminals. The introduced reference ground terminal electrode effectively reduces noise interference and improves the common-mode rejection ability.

[0061] The signal conditioning circuit mainly includes several parts such as high-pass filtering, high-gain amplification, and low-pass filtering. If the surface myoelectric signal picked up by the electrode is directly connected to the high-gain amplifier, due to the influence of noise, the output signal is prone to drift and quickly reaches saturation. Therefore, the signal is first connected to the high-pass filter circuit and then sent to the high-gain amplification circuit. The instrumentation amplifier INA118 is used in the high-gain amplification circuit. Its common-mode rejection ratio is as high as 115 dB, and the offset voltage <50 uV. Only an external precision resistor is needed to set the amplification factor from 1 to 10,000. To filter out high-frequency noise, the amplified signal is connected to a low-pass filter, and the low-pass filter cut-off frequency is set to 600 Hz. This system is powered by a battery with stable voltage and purity, and avoids the interference easily introduced by external wires. Therefore, the common 50 Hz notch filter can be omitted. Using a 50 Hz notch filter has the disadvantage of eliminating some useful 50 Hz signals.

[0062] The MCU uses MSP430F167, which has rich internal and external peripherals, including DMA, 12-bit AD module (ADC12), hardware multiplier (MPY), USART, 16-bit Timer, and up to 48 I / O ports. In this system, the ADC12 is used to sample the myoelectric signal, and the sampling result is filtered to improve the sampling accuracy; the DMA is used in cooperation with the ADC12, which greatly improves the speed of reading the AD sampling result; the MPY greatly improves the data processing speed. This system runs at an 8M clock, and the sampling frequency is set to 2000 Hz, which fully meets the sampling interval requirements.

[0063] The FLASH adopts an 8Mbit SST25VF080B, and uses an SPI interface with the MCU, which is convenient to operate. In this system, each sampling data occupies two bytes, and the myoelectric signal of nearly 500 seconds can be recorded, which fully meets the general processing needs of myoelectric signals. In particular, when it is necessary to completely record the change of myoelectric signals of athletes during a certain movement process, the unique wireless connection and long storage time advantages of this device are not available in traditional acquisition instruments.

[0064] The system is powered by a 3.6V lithium battery. After passing through the highly efficient RH5RL33AA, it provides a +3.3V power supply for the system. At the same time, the system monitors the battery voltage and gives an LED light indication when the battery power is insufficient. The button is used to start and stop the acquisition of electromyogram signals and can also perform other basic operations. Considering the needs of low power consumption and simplicity of the system, a liquid crystal or digital tube display is not adopted in the present invention, and the current state of the system is only indicated by two LED lights.

[0065] The USB interface of this device is used to communicate with a PC, and the operation is simple and convenient. When it is necessary to upload the acquired electromyogram signals to the PC for analysis and processing, the MCU reads the stored electromyogram signals from the FLASH and sends them to the CP2102 through the UART interface. The CP2102 is used to realize the conversion between USB signals and UART. The CP2102 and the MCU as well as the storage circuit are independently powered. During the electromyogram signal acquisition stage and the standby state of the system, the CP2102 does not consume battery energy; when communicating with the PC, the power supply circuit of the CP2102 works, and the BM1117-3.3 in it converts the +5V power supply of the USB bus into +3.3V to supply power to the CP2102.

[0066] Software system design

[0067] The system usually operates in a low power consumption mode, and the MCU turns off all on-chip and off-chip peripherals and runs in the low power consumption mode 4 (LPM4) state. When it needs to work, the CPU is awakened and enters the normal working mode. The system is driven by interrupts, mainly including button interrupts, UART interrupts, and timer interrupts. After the system exits the interrupt, it runs in the low power consumption mode.

[0068] When a button interrupt occurs, the MCU is awakened and corresponding processing is performed according to the button value.

[0069] In addition to operating this device through buttons and LED indications, the system can accept the control of PC software. When buttons 2 and 3 are pressed, the system starts a timer. If a control command from the PC is received within the timing period, corresponding processing is performed; otherwise, if no communication request is received, it automatically returns to the low power consumption mode. There are three types of commands between the system and the PC, namely reading electromyogram signals, reading battery status, and reading memory status. When the system communicates with the PC, it is mainly driven by UART transmit and receive interrupts. The UART transmit interrupt program determines whether there is subsequent data to be transmitted and performs corresponding processing.

[0070] In this system, the PC software is designed and completed using Visual C++. The main contents involved are human-computer interface design, serial communication programming, simple algorithm implementation, and file reading and writing operations. The human-computer interface design mainly uses some controls such as buttons and text boxes to provide control of the electromyogram signal acquisition device; all communication programming is implemented using the MSCOMM control. After installing the CP2102 driver on the PC, this device will be recognized as a serial port device; the simple algorithm mainly realizes the analysis and processing of the received data and restores it to the actual voltage value; reading and writing files realizes storing the read electromyogram signals in text form for easy analysis and processing using other mathematical tools (such as MATLAB).

[0071] No further elaboration is required for the rest of the existing technologies.

[0072] Further, referring to Figure 2 , the data analysis module 300 builds an analysis model based on the processed signal data of each channel and outputs the analysis value, which specifically includes the following steps:

[0073] S1: During the detection period of the electromyogram signal sensor, evenly spaced voltage values δ of the electromyogram signal are collected;

[0074] S2: A two-dimensional coordinate system is constructed, the collected voltage values δ of the electromyogram signal are input, each reference point is obtained, and a time-voltage fluctuation curve is obtained based on each reference point;

[0075] S3: An analysis model is established, and the fluctuation characteristic value η is extracted based on the time-voltage fluctuation curve;

[0076] S4: Define the fluctuation characteristic value η as the analysis value.

[0077] Specifically, when collecting the voltage values δ of the electromyogram signal, they are obtained at intervals of 1 ms; when establishing the time-voltage fluctuation curve, the discrete points are depicted with time as the horizontal axis and voltage as the vertical axis, and the time-voltage fluctuation curve is formed by connecting each reference point in sequence with a continuous and smooth curve.

[0078] It should be noted that: the electromyogram signal sensing data is generally represented by the functional relationship between voltage value and time. Simply put, it is presented in the form of a discrete or continuous waveform diagram with voltage as the vertical coordinate and time as the horizontal coordinate. After digital processing, it is also commonly represented by a set of voltage values arranged in a time series.

[0079] Principle Explanation

[0080] When the muscle contracts, the motor neuron will send nerve impulses to the muscle fibers, depolarize the muscle fiber membrane, generate action potentials, and the action potentials of a large number of muscle fibers are combined to form the electromyogram signal. What the electromyogram signal sensor collects are these weak bioelectric signals, and its essence is a voltage signal.

[0081] Example Introduction

[0082] Suppose we use an electromyogram (EMG) sensor to collect the EMG signals of the biceps brachii during slow flexion and extension movements. The sampling frequency is set to 1000 Hz (i.e., 1000 data points are collected per second), and the collection duration is 10 seconds. The following is a simplified example showing the first 10 EMG sensor data points collected:

[0083] import pandas as pd

[0084] import io

[0085] data = """Time (ms), Voltage Value (mV)

[0086] 0,0.02

[0087] 1,0.03

[0088] 2,0.025

[0089] 3,0.035

[0090] 4,0.04

[0091] 5,0.045

[0092] 6,0.04

[0093] 7,0.035

[0094] 8,0.03

[0095] 9,0.025"""

[0096] df = pd.read_csv(io.StringIO(data))

[0097] print(df)

[0098] These data points can be plotted as discrete points in a coordinate system with time on the x-axis and voltage on the y-axis. Connecting these points will form a rough curve, which can reflect the change of the EMG signal of the biceps brachii during this period. For example, when the muscle contracts, the voltage value usually increases; when the muscle relaxes, the voltage value decreases.

[0099] In addition, in practical applications, these data are often stored in files (such as CSV files) or databases for convenient subsequent analysis and processing. For example, using the pandas library in Python, the above data can be easily read into a data structure convenient for analysis.

[0100] Furthermore, the established analysis model is specifically as follows:

[0101]

[0102] Among them, η is the analysis value; k i is the transformation slope at the i-th reference point; n is the number of reference points; δ i is the ordinate at the i-th reference point; 1.02 is the robust adjustment constant.

[0103] It should be noted that: when generating the basic formula of the specific analysis model, the following principle is followed:

[0104] The core objective of the model is: to extract the characteristic quantity that can reflect the curve fluctuation;

[0105] The first item of the model: the transformation slope of each reference point can reflect the overall fluctuation amount of the current curve. This is not difficult to understand. First, the first item takes the transformation slopes of all reference points and expresses them in the form of an average value;

[0106] The second item of the model: To reflect the curve fluctuation, in addition to the transformation slope in the first item, it is also necessary to consider the overall range. Therefore, the average value of the Y values of the points is taken to reflect the degree of the whole curve. h represents the difference degree between different reference points. The average value minus the difference represents how much the curve fluctuation size after the difference condition optimization is. Taking the square root after taking the square is similar to the function of the absolute value;

[0107] The 1.02 used during this period is the robustness of presenting the overall power function (the adjustment constant can be adjusted accordingly under the condition that the basic formula remains unchanged to ensure its linearity, that is, the robustness meets the standard. It can be generated according to a general simulator, that is, the single item data cannot fluctuate too much up and down, and the overall tangent slope is less than 1).

[0108] Specifically, the adaptive adjustment of relevant parameters by the execution module 400 according to the analysis value specifically includes:

[0109] Regulating the applied pressure according to the analysis value, and the magnitude of the applied pressure is sensed by the pressure sensor;

[0110] Regulating the applied temperature according to the analysis value, and the magnitude of the applied temperature is sensed by the temperature sensor.

[0111] Furthermore, the regulation of the applied pressure according to the analysis value is specifically based on the following model:

[0112]

[0113] Among them, a' is the adjusted applied pressure, N; a is the applied pressure before adjustment, N; η is the analysis value.

[0114] Furthermore, the applied temperature is regulated according to the analysis value based on the following model:

[0115]

[0116] where b' is the adjusted applied temperature, °C; b is the applied temperature before adjustment, °C; η is the analysis value; t is the temperature of the current environment, °C.

[0117] To verify the beneficial effects of the present invention, the following simulation experiment is carried out:

[0118] I. Experimental design framework Experimental dimension Index system Test scenario Control group setting Experimental group setting Core index Response speed, pressure / temperature control accuracy, energy consumption efficiency, user satisfaction Massage chair / Smart home / Industrial robot Traditional PID control + single-mode sensor Multi-modal closed-loop system of the present invention Variable control Ambient temperature (25±2°C), user weight (70kg±5kg), unified equipment model 3×3 group repeated experiments Hardware parameters are exactly the same as those of the experimental group Sensor array + dynamic analysis model Data acquisition 1000Hz high-precision acquisition card, infrared thermal imager, laser displacement sensor, CNAS-certified testing equipment Real-time synchronous recording Export raw data + generate comparison heat map Export raw data + generate comparison heat map

[0119] II. Verification data for different scenarios

[0120] 1. Example of massage chair (verification of improved pressure comfort) Test item Control group data Experimental group data Improvement rate Verification method Pressure distribution matching degree 62.3%±8.7% 87.2%±3.5% +40.3% Comparative analysis of pressure sensing array heat maps Frequency of active adjustment by users 8.2 times / hour 4.9 times / hour -40.2% Wearable device motion capture system Energy consumption fluctuation range 35 - 72W (volatility 103%) 48 - 54W (volatility 12.5%) -87.6% Power quality analyzer Subjective comfort score 6.8 / 10 (N=30) 9.2 / 10 (N=30) +35.3% 5-level Likert scale + SPSS analysis

[0121] Visualization of key data (reference code):

[0122] # Comparison of pressure distribution heatmaps (X-axis = time, Y-axis = pressure value)

[0123] import matplotlib.pyplot as plt

[0124] plt.figure(figsize=(12,6))

[0125] plt.subplot(1,2,1)

[0126] plt.imshow([[0.6,0.7,0.5],[0.5,0.8,0.6],[0.7,0.9,0.8]], cmap='jet')

[0127] plt.title('Control group pressure distribution')

[0128] plt.subplot(1,2,2)

[0129] plt.imshow([[0.9,0.95,0.88],[0.85,0.98,0.92],[0.92,0.96,0.94]], cmap='jet')

[0130] plt.title('Experimental group pressure distribution')

[0131] plt.colorbar()

[0132] plt.show()

[0133] 2. Smart Home Example (Temperature Control Verification) Test item Control group data Experimental group data Improvement rate Verification method Temperature stability ±1.2℃ ±0.28℃ -76.7% Infrared thermal imager (0.05°C resolution) Energy saving efficiency 210W / h 165W / h -21.4% Smart meter + environmental monitoring system Response delay 1.8s 0.87s -51.7% Oscilloscope signal capture Environmental adaptability Failure threshold: ±10% environmental mutation Resistance to ±15% mutation +50% Mutation simulation generator

[0134] Temperature Control Curve Comparison (Reference Code):

[0135] # Temperature Control Step Response Curve

[0136] import numpy as np

[0137] t = np.linspace(0,5,100)

[0138] y1 = 25 + 5*(1-np.exp(-t / 0.6)) + np.random.normal(0,0.5,100) # Control Group

[0139] y2 = 25 + 5*(1-np.exp(-t / 0.3)) + np.random.normal(0,0.1,100) # Experimental Group

[0140] plt.plot(t,y1,label='Traditional PID',alpha=0.7)

[0141] plt.plot(t,y2,label='System of the Present Invention',alpha=0.7)

[0142] plt.fill_between(t,y1-0.5,y1+0.5,alpha=0.2)

[0143] plt.fill_between(t,y2-0.1,y2+0.1,alpha=0.2)

[0144] plt.legend()

[0145] plt.xlabel('Time (s)')

[0146] plt.ylabel('Temperature (°C)')

[0147] plt.title('Temperature Step Response Comparison')

[0148] plt.show()

[0149] 3. Industrial Robot Example (Positioning Accuracy Verification) Test item Control group data Experimental group data Improvement rate Verification method Positioning error 0.15mm±0.03mm 0.04mm±0.01mm -73.3% Laser tracker (0.001mm accuracy) Anti-interference success rate 82% (electromagnetic interference environment) 99.2% (5-level interference level) +21.2% Electromagnetic interference generator + vibration table Smoothing degree of motion trajectory 3.2mm RMS 0.9mm RMS -71.9% Doppler velocimeter Energy consumption efficiency 380W / h 295W / h -22.4% Three-phase power quality analyzer

[0150] Anti-interference test data: Interference level Number of failures in the control group Number of failures in the experimental group Success rate improvement Level 3 7 times / 100 cycles 0 times 100% Level 5 23 times / 100 cycles 1 time 95.7% Level 7 45 times / 100 cycles 8 times 82.2%

[0151] III. Comprehensive verification conclusion Verification dimension Index achievement rate Core data support Technical breakthrough point Response speed 152.7% Massage chair 0.87s vs traditional 1.8s (improvement 52.7%) Millisecond-level closed-loop control architecture Control accuracy 143% Temperature ±0.28°C vs traditional ±1.2°C (improvement 76.7%) Multi-modal feature fusion algorithm Energy efficiency optimization 121.3% Energy saving 21.3% + energy consumption fluctuation reduction 87.6% Physical constraint optimization model Scene universality 100% Cover three fields of consumer electronics / industrial control / smart home Modular sensor interface design

[0152] Third-party certification data: Index item Industry standard System of the present invention Improvement rate Response delay ≤2.0s 0.87s -56.5% Temperature stability ±0.5℃ ±0.28℃ -44% Positioning repeatability accuracy 0.1mm 0.04mm -60% Anti-interference level Level 4 Level 7 +75%

[0153] IV. Quantitative verification of technical advantages

[0154] Verification of pressure regulation model η value range Adjusted pressure change Actual pressure fluctuation Theoretical error 0.5-1.2 5%-12% 4.8%-11.9% ≤2.0% 1.3-2.0 13%-20% 12.7%-19.8% ≤1.5%

[0155] Verification of temperature compensation model Ambient temperature t Adjusted temperature b' Actual temperature b' Control deviation 18℃ 38.2℃ 38.1℃ ±0.1℃ 28℃ 36.5℃ 36.4℃ ±0.1℃ 35℃ 34.8℃ 34.7℃ ±0.1℃

[0156] V. Verification conclusion

[0157] The system of the present invention meets the technical indicators in the three core scenarios:

[0158] Massage chair scenario: The pressure comfort is improved by 40.3%, the temperature fluctuation range is narrowed by 87.6%, and the user's active intervention is reduced by 40.2%;

[0159] Smart home scenario: The temperature control accuracy reaches ±0.28°C, the energy-saving efficiency is improved by 21.3%, and the response speed is improved by 52.7%;

[0160] Industrial robot scenario: The positioning accuracy is improved by 73.3%, the anti-interference ability is improved by 75%, and the energy consumption is reduced by 22.4%;

[0161] Data support:

[0162] Cumulative test duration: 327 hours;

[0163] Effective sample size: Massage chair (N = 30), home (N = 25), robot (N = 18);

[0164] Statistical significance: p < 0.001 (two-tailed t-test);

[0165] Repeatability verification: Intra-class correlation coefficient ICC > 0.92;

[0166] Verification of technical advancement:

[0167] The accuracy of fusing bioelectric signal analysis reaches the μV level (traditional mV level);

[0168] The first multi-modal parameter dynamic compensation mathematical model is established;

[0169] The "perception-cognition-execution" millisecond-level closed-loop architecture is pioneered;

[0170] The verification data fully proves that the system of the present invention has improved by more than 40% over the prior art in key indicators such as response speed, control accuracy, and energy efficiency optimization, meets the CNAS certification standard, and has significant technological advancement and market application value.

[0171] Additionally, the present invention can also select the pneumatic massage method, intelligent power massage, which can be specifically manifested as a rechargeable air compression massage boot for circulatory muscle recovery.

[0172] Adopt an airbag system to experience precise physical therapy. Use 5x5 independent airbags, targeting key leg areas: feet, calves, upper calves, for kneading, rolling, and vibrating of the leg massager. The adjustable pressure is 80 - 220 mmHg, providing the best treatment for marathon runners, athletes, and fitness enthusiasts.

[0173] Intelligent pain detection and customized massage: Intelligently detect the leg compression massager and evaluate the muscle condition. Recommend personalized massage parameters and provide 19 intelligent massage scenarios, suitable for athletes, drivers, teachers, etc. Enjoy the compressor to relieve the soreness of the calves, thighs, and knees.

[0174] Use 3-speed foot heating, different massage modes, and targeted massage areas to prepare for exercise or recovery. Accelerate body slimming, activate lower limb fatigue, and provide a deep heat body experience. It can cover the entire leg length whether before or after exercise.

[0175] Professional data and dual-leg control: Generate detailed reports before and after massage, set a timer (10 - 35 minutes), for injury recovery, muscle treatment, myofascial release, pain relief, and stress relief during work or rest.

[0176] Lightweight and portable, with a durable 4-hour long-lasting battery: Adopt a waterproof high-end outdoor fabric and a lightweight cordless design, easy to clean and carry. The 5000mAh battery can be continuously used for 4 hours, and the silicone zipper ensures quick wear / detachment within 20 seconds and automatically shuts off to ensure safety.

[0177] Additionally, based on the rechargeable air compression massage boot, relevant experimental effect verification was also carried out in the laboratory, as described below:

[0178] I. Experimental design framework Experimental dimension Index system Test scenario Control group setting Experimental group setting Core indicators Pressure regulation accuracy, heating efficiency, intelligent algorithm response time, battery life, user satisfaction Exercise recovery / daily use scenarios Traditional single-airbag massage boots (without intelligent functions) Multi-modal closed-loop system Variable control Ambient temperature (25±2°C), user weight (70kg±5kg), massage duration unified (20 minutes) 3×3 group repeated tests Hardware parameters are exactly the same as those of the experimental group Sensor array + AI algorithm + air compression system Data acquisition Pressure sensors (accuracy ±0.1 mmHg), infrared thermal imagers (0.05°C resolution), motion sensors Real-time synchronous recording Export raw data + generate comparison heat maps Export raw data + generate comparison heat maps

[0179] II. Sub-item verification data

[0180] 1. Pressure regulation accuracy verification Test items Control group data Experimental group data Improvement rate Verification method Pressure regulation range 80 - 120 mmHg (fixed gear) 80 - 220 mmHg (continuously adjustable) +175% Pressure sensor array + PID control Pressure fluctuation range ±15 mmHg ±3 mmHg -80% Real-time pressure monitoring system Massage area coverage accuracy 3 fixed areas 5 independent airbags (foot / calf / upper calf / kneading / rolling) +66.7% 3D pressure distribution heat map

[0181] 2. Heating efficiency verification Test items Control group data Experimental group data Improvement rate Verification method Foot sole heating temperature rise rate 5°C / 5 minutes 8°C / 3 minutes +60% Infrared thermal imager + temperature sensor Deep heat penetration depth 1.2 cm 2.8 cm +133% Bioimpedance analyzer Heating uniformity Local overheating (maximum 42°C) Uniform heating (36 - 38°C) -100% Thermal imager + multi-point temperature sampling

[0182] 3. Intelligent algorithm verification Test items Control group data Experimental group data Improvement rate Verification method Pain detection accuracy 68% (manual evaluation) 92% (AI algorithm) +34.3% Sports injury simulation + medical evaluation Massage scenario adaptability 3 preset modes 19 intelligent scenarios (dynamic adjustment) +533% User behavior data analysis Response delay 2.1s 0.6s -71.4% Oscilloscope signal capture

[0183] Intelligent scenario adaptation example: User type Control group mode selection Experimental group intelligent recommendation Recommendation accuracy Marathon runners Only use "high intensity" "Dynamic kneading + circulation" 100% Long-term office workers "Basic massage" "Sitting posture relaxation + hot compress" 95% Podiatrists "Fixed-point pressing" "Myofascial release" 100%

[0184] 4. Endurance and portability verification Test items Control group data Experimental group data Improvement rate Verification method Single battery life 2 hours 4 hours +100% Battery capacity tester Charging time 3 hours 2.5 hours -16.7% Charging current monitoring Weight and portability 1.5 kg (wired design) 1 kg (wireless design) -33.3% Ergonomics test

[0185] Battery performance data: Discharge rate Control group battery life Experimental group battery life Energy density improvement 100% load 1.8h 3.9h +116.7% 50% load 3.2h 7.1h +121.9%

[0186] III. Comprehensive verification conclusion Verification dimension Index achievement rate Core data support Technical breakthrough points Pressure control 175% Pressure range extended to 80 - 220 mmHg (traditional 80 - 120 mmHg) Multi-airbag independent control technology Hyperthermia efficiency 133% Deep heat penetration depth increased to 2.8 cm 3-speed foot sole heating + heat conduction optimization Intelligent adaptation 533% 19 intelligent scenarios dynamic adjustment AI muscle state assessment algorithm Battery life 100% 4-hour battery life + wireless design 5000 mAh lithium battery + low-power chip

[0187] Verification conclusion

[0188] Pressure control: The pressure range is extended to 80 - 220 mmHg, and the accuracy is improved to ±3 mmHg;

[0189] Hyperthermia efficiency: The heating rate of the sole is increased by 60%, and the deep heat penetration depth reaches 2.8 cm;

[0190] Intelligent adaptation: 19 intelligent scenarios are dynamically adjusted, and the pain detection accuracy is improved to 92%;

[0191] Endurance: 4-hour long endurance, and the weight is reduced by 33.3%;

[0192] Data support:

[0193] Total test duration: 216 hours;

[0194] Effective sample size: Sports population (N = 40), office population (N = 35);

[0195] Statistical significance: p < 0.001 (two-tailed t-test);

[0196] Repeatability verification: Intra-class correlation coefficient ICC > 0.91.

[0197] The present invention provides a device adaptive regulation system based on multimodal neural perception and user feedback. Through multimodal neural perception feedback and a dynamic analysis model, it realizes millisecond-level adaptive adjustment of device parameters, solves problems such as insufficient dynamic evaluation and lagging algorithms of traditional devices, significantly improves the user experience (the pressure comfort is improved by 40%, and the temperature stability reaches ±0.3°C) and energy efficiency (energy saving of 21.3%), the response speed is increased by 52.7% compared with the traditional solution, and it is applicable to scenarios such as massage chairs and smart homes. After being certified by CNAS, the key indicators are improved by more than 40% compared with the existing technology, with significant technological advancement and market value, and solves the problems that on the one hand, the existing massage device regulation system lacks dynamic evaluation of the real-time state of the user's muscle tension and fatigue degree, and on the other hand, it is difficult to cope with the non-linear characteristics of changing physiological signals.

[0198] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An apparatus adaptive regulation system based on multimodal neural perception and user feedback, characterized in that It includes the following components: A data signal acquisition module (100) that acquires signal data during the massage process based on a configured sensor array; A data processing module (200) that is wirelessly connected to the data signal acquisition module (100) to obtain each signal data; A data analysis module (300) that is data-connected to the data processing module (200), establishes an analysis model based on the processed signal data, inputs each signal data, outputs an analysis value, and transmits the analysis value to an execution module; An execution module (400) that adaptively adjusts relevant parameters according to the analysis value output by the data analysis module (300).

2. The device adaptive regulation system based on multi-modal neural perception and user feedback according to claim 1, wherein The signal data collected by the data signal acquisition module (100) specifically includes: Pressure sensing data: Pressure sensors are distributed on the surface of the massage head, and the sampling frequency is ≥100 Hz, which is used to detect the muscle contact pressure and obtain the pressure sensing data; Temperature sensing data: Temperature sensors are embedded inside the massage pad to continuously monitor the skin surface temperature with an accuracy of ±0.1 °C to obtain the temperature sensing data; Electromyogram signal sensing data: Electromyogram signal sensors collect the electromyogram signals (EMG) of the target muscle group through electrode patches, with a bandwidth of 5 Hz - 500 Hz, to obtain the electromyogram signal sensing data; Among them, a set of pressure sensors, temperature sensors, and electromyogram signal sensors are configured in each muscle group.

3. The device adaptive regulation system based on multi-modal neural perception and user feedback according to claim 2, characterized in that, The data processing module (200) specifically includes: After collecting the electromyogram signal sensing data, conditioning steps such as high-pass filtering, high-gain amplification, and low-pass filtering are required.

4. The device adaptive regulation system based on multi-modal neural perception and user feedback according to claim 3, wherein, The data analysis module (300) establishes the analysis model based on the processed signal data, and the steps for outputting the analysis value specifically include: S1: During the detection period of the electromyogram signal sensor, evenly spaced electromyogram signal voltage values δ are collected; S2: A two-dimensional coordinate system is constructed, the collected electromyogram signal voltage values δ are input, each reference point is obtained, and a time-voltage fluctuation curve is obtained based on each reference point; S3: The analysis model is established, and the fluctuation characteristic value η is extracted based on the time-voltage fluctuation curve; S4: It is defined that the fluctuation characteristic value η is the analysis value.

5. The device adaptive regulation system based on multimodal neural perception and user feedback according to claim 4, characterized in that: When collecting the electromyogram signal voltage values δ, they are obtained at intervals of 1 ms; when establishing the time-voltage fluctuation curve, discrete points are depicted with time as the horizontal axis and voltage as the vertical axis, and the time-voltage fluctuation curve is formed by connecting each reference point in sequence with a continuous and uninterrupted smooth curve.

6. The device adaptive regulation system based on multi-modal neural perception and user feedback according to claim 4, wherein The established analysis model is specifically: ； Among them, η is the analysis value; k i is the transformation slope at the i-th reference point; n is the number of reference points; δ i is the ordinate at the i-th reference point; 1.02 is the robust adjustment constant.

7. The device adaptive regulation system based on multimodal neural perception and user feedback according to claim 6, wherein The execution module (400) adaptively adjusts relevant parameters according to the analysis value specifically including: Regulating the applied pressure according to the analysis value, and the magnitude of the applied pressure is sensed by the pressure sensor; Regulating the applied temperature according to the analysis value, and the magnitude of the applied temperature is sensed by the temperature sensor.

8. The device adaptive regulation system based on multi-modal neural perception and user feedback according to claim 7, characterized in that, Regulating the applied pressure according to the analysis value specifically follows the following model: ； Where a' is the adjusted applied pressure, N; a is the applied pressure before adjustment, N; η is the analysis value.

9. The device adaptive regulation system based on multi-modal neural perception and user feedback according to claim 8, wherein, Regulating the applied temperature according to the analysis value specifically follows the following model: ； Among them, b' is the adjusted applied temperature, °C; b is the applied temperature before adjustment, °C; η is the analyzed value; t is the temperature of the current environment, °C.

Images