Olfactory closed-loop feedback regulation method and system based on multi-modal physiological signals

Abstract

The application discloses a kind of based on multimodal physiological signal's olfactory closed loop feedback regulation method and system, the described including: step 1: the multimodal physiological data of user is obtained;Step 2: based on the preset deep learning evaluation model, the multimodal physiological data is carried out feature extraction and attention allocation, and output original prediction score;The dynamic correction coefficient is weighted with the original prediction score Calculation, obtain continuous instantaneous state score;Step 3: based on the instantaneous state score, the release power of different smell base and the proportion of dynamic adjustment are formed.The application is by real-time acquisition user's brain wave and respiration, heart rate and so on Multimodal physiological characteristics, utilize special interactive program to carry out state comprehensive evaluation, and drive the olfactory execution module configured with multiple independent basic smell chamber, form "physiological state drive, dynamic proportion release fragrance" olfactory stimulation closed loop feedback.

Description

Technical Field

[0001] This invention relates to the field of smart wearable device technology, and in particular to an olfactory closed-loop feedback control method and system based on multimodal physiological signals. Background Technology

[0002] With the fast pace of modern life, sub-health conditions such as anxiety and overwork are becoming increasingly common. Mindfulness meditation and psychological adjustment exercises have been proven to effectively improve these problems. In traditional methods, olfactory guidance through incense and essential oils is widely used to aid in entering a state of tranquility, possessing a strong subconscious calming effect. However, traditional methods have high barriers to entry and their effects are difficult to quantify. Existing portable brainwave devices and olfactory-assisted devices mainly suffer from the following technical problems: Single monitoring dimension and distorted state assessment: Most existing meditation headbands or relaxation devices rely only on single-channel EEG or simple optical heart rate, without combining electromyography, respiration and full-band EEG for multimodal fusion, resulting in low accuracy in recognizing the user's deep emotions and cognitive fatigue.

[0003] Olfactory interventions are static, unidirectional, and prone to causing physiological fatigue: Existing aromatherapy diffusers or scent diffusion devices mostly rely on continuous physical evaporation or fixed-time spraying after parameters are set, lacking a closed-loop regulation method that converts brainwave states into dynamic scent combinations in real time. More critically, human olfaction is highly susceptible to physiological adaptation (i.e., olfactory fatigue). Existing devices continuously release a single scent at a fixed concentration, causing a rapid decline in the user's perception and a significant decrease in intervention effectiveness over time. Furthermore, indiscriminately diffusing scents into the environment not only easily leads to cross-contamination and waste of consumables but also fails to achieve precise synchronization with the user's deep neural state and respiratory rhythm. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of the present invention is to provide an olfactory closed-loop feedback regulation method and system based on multimodal physiological signals to realize closed-loop digital olfactory nerve regulation.

[0005] To address the aforementioned technical problems, this invention proposes an olfactory closed-loop feedback modulation method based on multimodal physiological signals, comprising: Step 1: Acquire the user's multimodal physiological data, which includes at least full-band EEG signals and at least one non-EEG physiological auxiliary feature; Step 2: Based on the preset deep learning evaluation model, feature extraction and attention allocation are performed on the multimodal physiological data to output the original predicted score; at the same time, the physiological baseline of the user in the preset initial time period is extracted, and the deviation between the current multimodal physiological data and the physiological baseline is calculated to generate a dynamic correction coefficient; the dynamic correction coefficient and the original predicted score are weighted and calculated to obtain a continuous instantaneous state score. Step 3: Based on the instantaneous state score, generate and execute olfactory control instructions for the user, control the olfactory feedback execution terminal configured with multiple independent odor chambers, and dynamically adjust the release power and ratio of different odor bases.

[0006] Accordingly, embodiments of the present invention also provide an olfactory closed-loop feedback control system based on multimodal physiological signals, including a wearable device and an edge computing device, and further including: Data acquisition and transmission module: Located on the wearable device, it is used to acquire the user's multimodal physiological data in real time and transmit it to the edge computing device. Adaptive correction evaluation module: set on the edge computing device, used to output the original predicted score based on the preset deep learning evaluation model, and generate dynamic correction coefficients by combining the deviation of the user's initial physiological baseline, and then calculate the continuous instantaneous state score through the two. The multi-channel dynamic olfactory feedback execution module includes multiple internally isolated independent fragrance chambers, an independent controlled release array, and a directional fluid component. The multi-channel dynamic olfactory feedback execution module is used to receive the instantaneous state score and execute dynamic odor combinations by independently adjusting the physical release power and ratio of different fragrance chambers.

[0007] The beneficial effects of this invention are as follows: 1. This invention breaks through the traditional static aromatherapy's "perception-evaluation-dynamic fragrance blending" olfactory closed loop (derived from method step 3 and the product solution's multi-channel dynamic olfactory feedback execution terminal): Existing scent-guided devices and meditation aromatherapy products mostly release scents statically, unidirectionally, and at fixed concentrations, resulting in extremely limited intervention methods. This invention uniquely introduces a dynamic combination fragrance algorithm driven by high-frequency EEG characteristics. It dynamically maps the user's real-time brainwave state score to independent release ratios of different basic scent chambers (such as soothing and activating phases), forming an olfactory closed loop of "precise brainwave-controlled fragrance." This deep neuro-olfactory bidirectional interaction can precisely and seamlessly change the scent combination and concentration according to the user's real-time anxiety or relaxation level, significantly shortening the time required for the user to enter a state of mindfulness or deep relaxation.

[0008] 2. The anti-fatigue and adaptive extinction mechanism of this invention overcomes the physiological adaptation of olfaction: The human olfactory system is highly susceptible to physiological adaptation (i.e., olfactory fatigue), causing continuously emitted odors to lose their guiding effect. This invention pioneers an "intermittent pulse-type" fragrance release and sensory adaptive extinction mechanism based on EEG state verification. When the system objectively determines that the user has entered the target state, it actively cuts off the continuous high-concentration odor supply, switching to maintaining the state only with micro-pulses when the state declines, or releasing a neutralizing agent to clear residual olfactory receptors. This technology completely solves the technical pain points of traditional aromatherapy devices, such as "not recognizing the fragrance after prolonged exposure" and the ease with which continuous overstimulation can interrupt deep meditation.

[0009] 3. This invention combines precise odor targeting and synchronous release based on respiratory phase: Unlike traditional aroma diffusers that indiscriminately disperse odors into the environment (leading to cross-contamination and wasted consumables), this invention uses "real-time respiratory rhythm" from multimodal physiological data as the trigger point for fluid physics control. The system controls a directional fan or atomizing component to precisely deliver odor molecules only at the beginning of the user's inhalation phase and immediately stop at the end of the exhalation phase. This not only greatly improves the effective utilization rate of odor molecules and the subconscious neural anchoring effect, but also achieves precise odor isolation and rapid dissipation at the physical level.

[0010] 4. The multimodal feature fusion and edge computing of this invention balance high-precision evaluation with lightweight equipment: Single EEG signals are highly susceptible to external interference. This invention synchronously fuses full-band EEG features with non-EEG signals such as heart rate and respiration, and innovatively introduces a "baseline drift-based dynamic weight calibration" algorithm, effectively eliminating motion artifact interference and individual physiological baseline differences, resulting in more accurate assessments of emotional load. Simultaneously, the computationally intensive three-dimensional feature matrix derivation is extracted from the headband body and transferred to a smart device for operation, significantly reducing the headband's size, heat generation, and power consumption, and substantially improving the wearable device's battery life and wearing comfort. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the olfactory closed-loop feedback regulation method based on multimodal physiological signals according to an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of the process when a user begins mindfulness training, according to an embodiment of the present invention.

[0013] Figure 3 This is a schematic diagram illustrating the application of the olfactory closed-loop feedback control system based on multimodal physiological signals according to an embodiment of the present invention.

[0014] Figure 4 This is a schematic diagram of the process when a user begins meditation / psychological adjustment according to an embodiment of the present invention. Detailed Implementation

[0015] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] In this embodiment of the invention, directional indicators (such as up, down, left, right, front, back, etc.) are only used to explain the relative positional relationship and movement of each component in a specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.

[0017] Furthermore, in this invention, descriptions involving "first," "second," etc., are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features.

[0018] Please refer to Figure 1 The olfactory closed-loop feedback modulation method based on multimodal physiological signals in this embodiment of the invention includes steps 1 to 3.

[0019] Step 1: Acquire the user's multimodal physiological data, which includes at least full-band EEG signals and at least one non-EEG physiological auxiliary feature. The auxiliary feature includes heart rate, respiration, head movement, and electromyography; simultaneously, for the full-band EEG signals from 0-125H, short-time Fourier transform or wavelet transform is used to construct a time-frequency-space three-dimensional feature matrix containing time axis, frequency axis, and channel axis.

[0020] In specific implementation, the execution entities of this invention are the wearable body end (i.e., the microprocessor built into the headband, responsible for signal acquisition and transmission), the edge computing device end (such as a dedicated interactive feedback program running on an external smart terminal, responsible for edge computing and instruction generation, such as mobile phones, PC devices, etc.), and the multi-channel olfactory feedback execution terminal (responsible for the physical release of odors).

[0021] Synchronous acquisition and low-latency transmission of multimodal physiological signals: (1) Signal acquisition: The multimodal signal acquisition module built into the headband acquires the user's raw physiological electrical signals in real time.

[0022] (2) Preprocessing and transmission: The microprocessor built into the headband performs analog-to-digital conversion and data packetization on the physiological electrical signals, and transmits the raw physiological data with a high sampling rate (e.g., 250Hz) to the external smart terminal, which is an edge computing device, in real time with low latency through the built-in wireless communication module.

[0023] In practice, heart rate and respiratory signals can be acquired using photoplethysmography (PPG) sensors instead of skin-tight physiological electrodes; head movements and respiratory rate can also be assessed by capturing facial micro-expressions and chest rise and fall features through the front-facing camera of an external smart device.

[0024] Step 2: Based on the preset deep learning evaluation model, feature extraction and attention allocation are performed on the multimodal physiological data to output the original predicted score; at the same time, the physiological baseline of the user in the preset initial time period is extracted, and the deviation between the current multimodal physiological data and the physiological baseline is calculated to generate a dynamic correction coefficient; the dynamic correction coefficient and the original predicted score are weighted and calculated to obtain a continuous instantaneous state score.

[0025] The architecture of the deep learning evaluation model includes a multi-head attention mechanism, which automatically assigns dynamic attention scores to different frequency bands and auxiliary features in the range of 0-125Hz. After processing by a fully connected layer, the model outputs a raw prediction score that reflects the instantaneous physiological state.

[0026] The dedicated interactive feedback program on the smart terminal receives raw physiological data and executes the following multi-level evaluation logic: (1) Multimodal feature decoding and three-dimensional feature matrix construction: The program decodes non-EEG auxiliary features such as heart rate, respiration, head movement and electromyography (EMG) from the original signal; at the same time, for the full-band EEG signal of 0-125Hz, unlike traditional devices that only extract a single frequency band, this program uses short-time Fourier transform (STFT) or wavelet transform to construct a "time-frequency-space three-dimensional feature matrix" containing time axis, frequency band axis and channel axis, so as to completely preserve the phase synchronization information across frequency bands.

[0027] (2) Deep learning evaluation of multimodal feature fusion (raw score): The above three-dimensional feature matrix and auxiliary features are input into a preset deep learning evaluation architecture. This architecture includes a multi-head attention mechanism, which automatically assigns dynamic attention scores (i.e., dynamic weights) to different frequency bands in 0-125Hz (such as high gamma waves representing distraction, 100-125Hz) and auxiliary features. After processing by fully connected layers, the model outputs a raw predicted score that reflects the instantaneous physiological state.

[0028] (3) Dynamic weight calibration (corrected score) based on baseline drift: To eliminate individual baseline differences and environmental noise, the system introduces a baseline adaptive correction mechanism. The system collects multimodal data 30 seconds before each training session to lock the initial physiological baseline. In real-time evaluation, the deviation of the current multimodal features from the baseline is calculated and normalized to generate a dynamic correction coefficient N.

[0029] (4) State Index Output: The final continuous state score is calculated by multiplying the original predicted score by the dynamic correction coefficient N. This score is continuously mapped to the range of 0-100 and is used as the input threshold for the intensity of subsequent olfactory feedback, comprehensively reflecting the user's current cognitive state, emotional load and physical and mental relaxation level.

[0030] Step 3: Based on the instantaneous state score, generate and execute olfactory modulation instructions for the user, controlling an olfactory feedback execution terminal equipped with multiple independent odor chambers to dynamically adjust the release power and ratio of different odor bases. When the instantaneous state score reaches a preset target state threshold, trigger an olfactory anti-adaptation and adaptive extinction mechanism to stop the continuous high-concentration release of odors and switch to a pulsed intermittent release mode triggered by state decline or release an odorless neutralizing agent to overcome olfactory fatigue and avoid interrupting the user's immersion state.

[0031] As one implementation method, in step 3, the real-time breathing characteristics collected in step 1 are used as time axis anchor points. At the beginning of the user's inhalation phase, the olfactory feedback execution terminal is controlled to release mixed odor molecules synchronously, and the release stops at the exhalation phase, thereby greatly improving the utilization rate of odor molecules and the neural anchoring effect.

[0032] In practice, based on the final state score output in step 2, a dedicated interactive feedback program generates odor mixing and release instructions, which, in conjunction with the olfactory feedback execution terminal, provide the user with dynamic olfactory closed-loop feedback to guide the user into a target cognitive state (such as mindfulness focus or deep relaxation). (1) Brainwave-driven dynamic olfactory combination and adjustment mechanism: The olfactory feedback execution terminal has a built-in fragrance box containing multiple independent chambers (e.g., configured with basic odor matrices such as soothing base and activating base). The dedicated interactive feedback program uses the continuous state score of 0-100 and the energy ratio of specific EEG frequency bands as input variables, and dynamically maps them to the control signals (such as pulse width modulation PWM signals) of the heating elements or atomizing elements of each independent chamber. For example, when the system detects that the user is in a state of high cognitive load or anxiety, the system automatically increases the release power of the chamber corresponding to the soothing base; as the state score approaches the mindfulness target threshold, the system dynamically adjusts the odor release ratio of each chamber to achieve a seamless and gradual change in the combination of mixed odors.

[0033] (2) Olfactory Anti-adaptation and Adaptive Extinction Mechanism (Anti-olfactory Fatigue Design): To overcome the defect that human olfaction is prone to physiological adaptation, the system introduces an intermittent pulsed aroma release strategy based on EEG state verification. When the system determines that the user's EEG characteristics (such as continuous alpha and theta waves) indicate that they have entered a deep flow or mindfulness state, the system automatically triggers the sensory adaptive extinction mechanism. At this time, the system cuts off the continuous release of all aroma chambers and only releases the target odor in a short pulse form at the moment when the EEG state shows a downward trend; or during the transition period, it releases an extremely small amount of odorless neutralizer to clear the residual olfactory receptors in the nasal cavity, avoiding the disturbance of the flow state by the continuous high concentration of odor.

[0034] (3) Phase-synchronized release mechanism combined with respiratory rhythm: Effective odor inhalation is highly dependent on the user's respiratory physical actions. The system uses the "real-time respiratory rhythm" characteristics collected in step 1 as the time axis anchor point to control the start and stop of the directional fluid components (such as micro fans or piezoelectric elements) of the olfactory feedback execution terminal. The system limits the synchronous release of mixed odor molecules only at the beginning of the user's "inhalation" phase and stops releasing them at the "exhalation" phase, thereby greatly improving the utilization rate of odor molecules and the neural anchoring effect.

[0035] Simultaneously, the system continuously loops through steps 1-3 at a specific refresh rate, comparing the error value between the current EEG state and the target state in real time, dynamically fine-tuning the odor ratio and release parameters, and ultimately generating a quantified training report. The user's mindfulness training process using this invention is as follows: Figure 2 As shown.

[0036] In practice, if the headband or olfactory feedback execution terminal uses a higher-performance built-in SoC chip, the comprehensive evaluation calculation in "Step 2" can be completed entirely on the device itself, with the mobile phone serving only as a display screen for data status; or, the mobile phone can serve only as a data relay station (router), uploading the raw data to the cloud server, where the cloud-based large model performs multimodal fusion algorithm deduction, and then sends the olfactory matching ratio and start / stop control commands back to the local terminal.

[0037] This invention relates to an olfactory closed-loop feedback control system based on multimodal physiological signals, comprising a wearable device, an edge computing device, an edge data acquisition and transmission module, an adaptive correction and evaluation module, and a multi-channel dynamic olfactory feedback execution module. This invention collects the user's brainwaves and multimodal physiological characteristics such as respiration and heart rate in real time, uses a dedicated interactive program for comprehensive state assessment, and drives an olfactory execution module configured with multiple independent basic odor chambers to form a closed-loop olfactory stimulation feedback based on "physiological state-driven, dynamically proportioned fragrance release." In particular, it introduces a dynamic fragrance-mixing algorithm based on brainwave baseline offset and a phase-synchronized release mechanism combined with respiratory rhythm, overcoming the shortcomings of traditional one-way and static aromatherapy, and achieving real-time guidance, evaluation, and regulation of the user's cognitive state and relaxation level. Compared to other solutions, this invention emphasizes real-time feedback, dynamic proportioning, and closed-loop feedback control to prevent olfactory adaptation (olfactory fatigue). When the user uses this invention for meditation / psychological adjustment, the workflow is as follows: Figure 4 As shown.

[0038] In this system, when a user begins mindfulness training, the device collects EEG and respiration data in real time. The smart terminal processes the data and assesses the current state. Based on the assessment score, the system dynamically controls the heating or atomization ratio of the multiphase aroma chambers (such as soothing and activating bases) to release the most suitable auxiliary scents for the current state, guiding the user to correctly practice mindfulness. When the system determines that the user has entered the target mindfulness / flow state, it automatically triggers an olfactory adaptive extinction mechanism (such as switching to intermittent pulse release or stopping release) to avoid olfactory fatigue and overstimulation. When the user ends the mindfulness activity, the system records the mindfulness score and the mindfulness time, quantitatively evaluating the effectiveness of the user's mindfulness activity.

[0039] Data Acquisition and Transmission Module: Located on the wearable device, this module acquires the user's multimodal physiological data in real time and transmits it to the edge computing device. The multimodal physiological data acquired by this module includes at least full-band EEG signals and at least one non-EEG physiological auxiliary feature; the auxiliary feature includes heart rate, respiration, head movement, and electromyography (EMG). For the 0-125Hz full-band EEG signals, the module uses short-time Fourier transform or wavelet transform to construct a time-frequency-space three-dimensional feature matrix containing time, frequency, and channel axes. In specific implementations, the wearable device can be a headband body: using flexible wearable materials as the physical support carrier for the entire acquisition system. The data acquisition and transmission module consists of a multimodal physiological signal acquisition module and a main control and wireless transmission module. The multimodal physiological signal acquisition module: located inside the headband body, includes an EEG electrode assembly and multimodal sensors. The acquisition module is used to collect physiological signals such as EEG, heart rate, respiration, electromyography, and head movements in real time. The main control and wireless transmission module is built into the headband and electrically connected to the acquisition module. Due to the edge computing architecture, this module is primarily responsible for hardware-level analog-to-digital conversion (ADC), data packetization, and establishing wireless communication connections (such as Bluetooth Low Energy (BLE) or Wi-Fi), transmitting high-sampling-rate raw, multi-dimensional physiological data to external edge computing devices in real time.

[0040] Adaptive Correction Evaluation Module: Located on the edge computing device, this module outputs an initial predicted score based on a preset deep learning evaluation model. It then combines this score with the deviation from the user's initial physiological baseline to generate a dynamic correction coefficient, thereby calculating a continuous instantaneous state score. The deep learning evaluation model's architecture includes a multi-head attention mechanism, automatically assigning dynamic attention scores to different frequency bands and auxiliary features within the 0-125Hz range. After processing by a fully connected layer, the model outputs an initial predicted score reflecting the instantaneous physiological state. Edge Computing Device: This module connects the user's smartphone, tablet, or PC—external smart devices with high independent computing power—to the headband's main control and wireless transmission modules via a wireless communication network. Please refer to [link / reference needed]. Figure 3 Dedicated interactive feedback program: Installed and running within the edge computing terminal. This program is responsible for receiving raw physiological data, decoding the core algorithm, performing deep learning deduction and baseline calibration, mapping the evaluated state scores into ratio combinations and start / stop control commands for different odors, and sending the olfactory control signal back to the olfactory feedback execution terminal.

[0041] The multi-channel dynamic olfactory feedback execution module includes multiple internally isolated independent fragrance chambers, an independent controlled release array, and a directional fluid component. This module receives the instantaneous state score and executes dynamic odor combinations by independently adjusting the physical release power and ratio of different fragrance chambers. When the instantaneous state score reaches a preset target state threshold, the module triggers an olfactory anti-adaptation and adaptive extinction mechanism to dynamically weaken continuous odor stimulation and prevent olfactory adaptive physiological fatigue. The module uses real-time respiratory characteristics obtained from the data acquisition and transmission module as a time axis anchor point, synchronously releasing mixed odor molecules at the beginning of the user's inhalation phase and ceasing release at the end of the exhalation phase, thereby greatly improving the utilization rate of odor molecules and the neural anchoring effect.

[0042] The multi-channel dynamic olfactory feedback execution module establishes a communication connection with the edge computing terminal and is controlled by instructions sent by a dedicated interactive feedback program to perform physical odor mixing and release. The multi-channel dynamic olfactory feedback execution module can be used as a standalone external desktop device or integrated into the headband / neckband wearable structure. Its internal structure is electrically connected and includes the following sub-components: (1) Multiphase base fragrance box: The internal physical isolation consists of at least two independent fragrance storage liquid or solid carrier fragrance chambers, which encapsulate base scents of different phases (e.g., soothing base, activating base, etc.) as the physical source of scent release.

[0043] (2) Independent controlled release array: Each fragrance chamber of the fragrance cartridge is equipped with an independent thermal heating element or piezoelectric atomizing element. The array is controlled by pulse width modulation (PWM) commands issued by a dedicated interactive feedback program, and by changing the heating power or atomization frequency, the concentration of odor volatilization in different chambers can be precisely and independently controlled.

[0044] (3) Directional fluid assembly: Includes a miniature silent centrifugal fan and air guide structure, located at the front end of the evaporation path of the fragrance cartridge, and electrically connected to the main control circuit. This assembly receives breathing phase synchronization commands from a dedicated interactive feedback program, and only starts fluid delivery during the user's inhalation phase, accurately pushing the mixed odor molecules to the user's nasal breathing zone at a specific flow rate, and stops delivery during the exhalation phase, forming a physical isolation and odor retention prevention mechanism.

[0045] In practical implementation, the structure of the multi-channel dynamic olfactory feedback execution module can also be adopted in the following ways: (1) Replacement of physical mechanisms for odor emission: In addition to using a controlled thermosensitive heating array to accelerate the volatilization of solid or liquid fragrances, the odor release device of the independent fragrance chamber can be equivalently replaced by an ultrasonic atomization module, a microporous piezoelectric atomizing plate, or a high-precision air pump microfluidic valve. The above-mentioned alternative hardware can also receive PWM signals sent by the system to achieve precise electronic control of different basic odor release concentrations and ratios.

[0046] (2) Replacement of the physical form of the olfactory terminal: In addition to integrating the fragrance box and air guide component onto the EEG headband body, the multi-channel dynamic olfactory feedback execution module can also be equivalently split into an independent desktop-level intelligent digital aroma diffuser, a neck-mounted wearable diffuser, or integrated as an external module into the nose pad of the AR / VR head-mount device, all of which can achieve data linkage with the EEG acquisition terminal and precise odor delivery.

[0047] (3) Expansion of feedback mechanism: Based on the olfactory closed loop, the environmental smart home system (such as fresh air system or air purifier) ​​can be linked in time. When the "olfactory adaptive fading mechanism" is triggered, the fresh air system is activated to accelerate the metabolism and removal of residual odor molecules in the environment.

[0048] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A closed-loop feedback modulation method for olfaction based on multimodal physiological signals, characterized in that, include: Step 1: Acquire the user's multimodal physiological data, which includes at least full-band EEG signals and at least one non-EEG physiological auxiliary feature; Step 2: Based on the preset deep learning evaluation model, feature extraction and attention allocation are performed on the multimodal physiological data to output the original predicted score; at the same time, the physiological baseline of the user in the preset initial time period is extracted, and the deviation between the current multimodal physiological data and the physiological baseline is calculated to generate a dynamic correction coefficient; the dynamic correction coefficient and the original predicted score are weighted and calculated to obtain a continuous instantaneous state score. Step 3: Based on the instantaneous state score, generate and execute olfactory control instructions for the user, control the olfactory feedback execution terminal configured with multiple independent odor chambers, and dynamically adjust the release power and ratio of different odor bases.

2. The olfactory closed-loop feedback modulation method based on multimodal physiological signals as described in claim 1, characterized in that, In step 3, when the instantaneous state score reaches the preset target state threshold, the olfactory anti-adaptation and adaptive extinction mechanism is triggered to stop the continuous high-concentration release of odor and switch to a pulse intermittent release mode triggered by state fallback or release an odorless neutralizer to overcome olfactory fatigue and avoid interrupting the user's immersive state.

3. The olfactory closed-loop feedback modulation method based on multimodal physiological signals as described in claim 1, characterized in that, In step 1, the auxiliary features include heart rate, respiration, head movement, and electromyography; for the full-band EEG signal of 0-125H, short-time Fourier transform or wavelet transform is used to construct a time-frequency-space three-dimensional feature matrix containing time axis, frequency axis and channel axis.

4. The olfactory closed-loop feedback modulation method based on multimodal physiological signals as described in claim 3, characterized in that, In step 3, the real-time breathing characteristics collected in step 1 are used as time axis anchors. At the beginning of the user's inhalation phase, the olfactory feedback execution terminal is controlled to release mixed odor molecules synchronously, and the release stops at the exhalation phase, thereby greatly improving the utilization rate of odor molecules and the neural anchoring effect.

5. The olfactory closed-loop feedback modulation method based on multimodal physiological signals as described in claim 3, characterized in that, The architecture of the deep learning evaluation model includes a multi-head attention mechanism, which automatically assigns dynamic attention scores to different frequency bands and auxiliary features in the range of 0-125Hz. After processing by a fully connected layer, the model outputs a raw prediction score that reflects the instantaneous physiological state.

6. An olfactory closed-loop feedback control system based on multimodal physiological signals, comprising a wearable device and an edge computing device, characterized in that, Also includes: Data acquisition and transmission module: Located on the wearable device, it is used to acquire the user's multimodal physiological data in real time and transmit it to the edge computing device. Adaptive correction evaluation module: set on the edge computing device, used to output the original predicted score based on the preset deep learning evaluation model, and generate dynamic correction coefficients by combining the deviation of the user's initial physiological baseline, and then calculate the continuous instantaneous state score through the two. The multi-channel dynamic olfactory feedback execution module includes multiple internally isolated independent fragrance chambers, an independent controlled release array, and a directional fluid component. The multi-channel dynamic olfactory feedback execution module is used to receive the instantaneous state score and execute dynamic odor combinations by independently adjusting the physical release power and ratio of different fragrance chambers.

7. The olfactory closed-loop feedback control system based on multimodal physiological signals as described in claim 6, characterized in that, When the instantaneous state score reaches the preset target state threshold, the multi-channel dynamic olfactory feedback execution module triggers the olfactory anti-adaptation and adaptive extinction mechanism to dynamically weaken the continuous odor stimulation in order to prevent olfactory adaptive physiological fatigue.

8. The olfactory closed-loop feedback control system based on multimodal physiological signals as described in claim 6, characterized in that, The multimodal physiological data acquired by the data acquisition and transmission module includes at least full-band EEG signals and at least one non-EEG physiological auxiliary feature; the auxiliary feature includes heart rate, respiration, head movement and electromyography; the data acquisition and transmission module uses short-time Fourier transform or wavelet transform to construct a time-frequency-space three-dimensional feature matrix containing time axis, frequency axis and channel axis for the full-band EEG signals from 0-125H.

9. The olfactory closed-loop feedback control system based on multimodal physiological signals as described in claim 8, characterized in that, The multi-channel dynamic olfactory feedback execution module uses the real-time breathing characteristics obtained by the data acquisition and transmission module as the time axis anchor point. It synchronously releases mixed odor molecules at the beginning of the user's inhalation phase and stops releasing them at the end of the exhalation phase, thereby greatly improving the utilization rate of odor molecules and the neural anchoring effect.

10. The olfactory closed-loop feedback control system based on multimodal physiological signals as described in claim 8, characterized in that, The architecture of the deep learning evaluation model includes a multi-head attention mechanism, which automatically assigns dynamic attention scores to different frequency bands and auxiliary features in the range of 0-125Hz. After processing by a fully connected layer, the model outputs a raw prediction score that reflects the instantaneous physiological state.

Images