Individual identity recognition and intention perception system based on multi-frequency bioimpedance spectrum

Abstract

This invention discloses an individual identification and intent perception system based on multi-frequency bioimpedance spectrum. The system applies a broadband sweep signal to the human body through an excitation signal generation module, collects response signals via a core sensing module, and obtains complex impedance data through a signal processing module. The computational processing module uses a multi-frequency differential algorithm to extract unique physical fingerprint features characterizing the individual's deep tissue structure from the impedance spectrum, and introduces a psychological state perception algorithm based on a dynamic conductivity model. Through AI fusion decision-making, it simultaneously achieves high-precision identity authentication and real-time discrimination of abnormal physiological states, involuntary states such as sleep. Finally, the execution output module performs operations such as unlocking and silent alarm. This invention solves the shortcomings of traditional biometric technologies, such as ease of forgery and inability to perceive user intent, while possessing higher-level anti-counterfeiting capabilities, proactive security defense, and privacy protection.

Description

Technical Field

[0001] This invention relates to the field of individual identification and intent perception system based on multi-frequency bioimpedance spectrum. Background Technology

[0002] Current mainstream biometric technologies such as fingerprint, facial, and iris recognition all rely on the collection and comparison of static physical forms of the human body, which has inherent security limitations: First, static features are easily copied, for example, by using fingerprint films, high-definition masks, or 3D printed models to deceive sensors; second, they lack the ability to judge willingness, the system can only verify "whether it is a living body", but cannot identify whether the user is in a "conscious and voluntary" state, which may lead to the device being involuntarily unlocked when the user is in an abnormal physiological state, asleep, or unconscious; third, biometric features are irrevocable, and once leaked, they will lead to permanent privacy risks.

[0003] While existing bioelectrical impedance analysis (BIA) technology is continuously improving measurement accuracy—for example, patent CN202211511899.4 optimizes electrode contact impedance to enhance data accuracy—its fundamental goal remains limited to statistical body composition analysis in the medical and health field, failing to address the core needs of high-security identity authentication. This technology focuses solely on obtaining more accurate impedance parameters but fails to solve: how to extract uniquely individual features from impedance data, how to perceive a user's psychological state through dynamic impedance changes, and how to transform impedance technology into an active security authentication system. Summary of the Invention

[0004] The present invention aims to solve the above-mentioned technical problems by providing an individual identification and intent perception system based on multi-frequency bioimpedance spectrum. By applying a safe microcurrent to the human body and measuring its impedance spectrum response, the system extracts an individual-unique "impedance fingerprint" and a "dynamic intent feature" reflecting neural activity, thus simultaneously completing the dual determination of identity and status in a single contact.

[0005] To address the aforementioned technical problems, the present invention provides the following technical solution: an individual identification and intent perception system based on multi-frequency bioimpedance spectrum, comprising: The excitation signal generation module is used to generate multi-frequency sweep current signals in the range of 10Hz-1MHz; The core sensing module includes at least one pair of composite electrodes made of biocompatible conductive material, which are arranged on the surface of the device to form a closed loop with the human body to collect impedance signals. The signal processing module, connected in sequence to a differential amplifier, a bandpass filter, and an analog-to-digital converter, is used to obtain digitized complex impedance data. ; The computational processing module is used to: calculate impedance characteristic parameters at different frequencies based on the impedance data; generate feature vectors for identity recognition based on the impedance characteristic parameters; extract dynamic feature parameters based on the time change information of the impedance characteristic parameters; and includes a multi-frequency differential algorithm unit and an AI intent classifier for extracting features and recognizing states, wherein the AI ​​intent classifier includes a convolutional neural network and / or a recurrent neural network. The execution output module is used to output authentication results or warning signals; The determination module is used to: compare the feature vector with a pre-stored template to output the identity authentication result; and determine whether there is an abnormal physiological state based on the dynamic feature parameters. The output module is used to output corresponding control signals based on the identity authentication result and the abnormal physiological state determination result; The feature vector includes the impedance amplitude ratio and phase angle corresponding to at least two preset frequency points. The phase angle ; The computational processing module integrates a deep learning-based classifier for fusion analysis of the physical fingerprint features and the dynamic conductivity model.

[0006] Furthermore, the electrodes are arranged in a bipolar configuration on both sides of the device to form a closed loop that runs through the palm of the hand. The material is a biocompatible conductive material. The electrodes include discrete electrode sheets disposed on the surface of the housing or a transparent conductive layer integrated under the display panel.

[0007] Furthermore, the multi-frequency difference algorithm unit in the computation processing module performs the following steps: (a) Initiate a stepped frequency sweep excitation from 10 kHz to 500 kHz; (b) Obtain multiple sets of complex impedance data ; (c) Calculate the rate of change of impedance between different frequency points, i.e., the impedance slope. This is used to suppress linear noise caused by sweat. (d) Perform feature dimensionality reduction on the frequency response curve to extract the core resonance feature vector; (e) Input the extracted feature vector into the AI ​​intent classifier to classify and determine "normal unlocking", "sleep state" and "abnormal physiological state".

[0008] Furthermore, the algorithm unit calculates the high-frequency to low-frequency impedance ratio. This is to characterize individual differences in muscle and bone structure.

[0009] Furthermore, the AI ​​classifier employs a CNN-LSTM hybrid model to identify impedance micropulsation features and distinguish between "normal," "sleep," and "abnormal physiological states."

[0010] Furthermore, it also includes an environmental adaptive compensation module, which is used to compensate for changes in skin impedance based on temperature and humidity data.

[0011] Furthermore, it also includes an environmental adaptive compensation module, which compensates for changes in surface skin impedance based on real-time temperature and humidity data to ensure stable extraction of impedance characteristics of deep tissues.

[0012] Furthermore, the computational processing module is configured to monitor dynamic conductivity. The change in dynamic conductivity is used to identify stress states, wherein the dynamic conductivity is... It is modeled as consisting of a time-varying surface resistance component and a time-varying skin conductance component in parallel, wherein the time-varying skin conductance component is related to the intensity of sympathetic nerve activity. The system calculates the rate of change of the complex impedance Z at at least one selected frequency in the swept frequency signal over time and compares the rate of change with a preset threshold. If the rate of change exceeds the threshold, the system determines that the user is in a stress state and triggers an early warning.

[0013] Furthermore, the dynamic feature parameters include at least one of the following: Rate of change of impedance over time; Statistical characteristics of impedance signals; The spectral characteristics of the impedance signal within a preset frequency band.

[0014] An identity recognition method based on a multi-frequency bioimpedance spectrum individual identity recognition and intent perception system includes: When the user touches the device, the excitation signal generation module applies multi-frequency sweep excitation; Complex impedance data is acquired through the core sensing module. ; The signal processing module performs amplification, filtering, and A / D conversion. The computational processing module executes the multi-frequency difference algorithm and AI classification, and outputs the recognition results. The execution output module performs unlocking, locking, alarm, or data uploading operations based on the recognition results. The method supports cross-device frequency synchronization and encrypted near-field communication with the human body.

[0015] The advantages of this invention compared to the prior art are: This solution identifies the complex electrical properties of subcutaneous muscles, bones, and body fluids. These properties are uniquely determined by an individual's physiological structure and cannot be forged unless the human body is completely cloned. Only living organisms with normal cell activity and blood circulation can generate specific impedance spectra, thus fundamentally preventing attacks from non-living substances such as prostheses and silicone. This solution expands biometrics from "who is it" to "what state it is in," enabling the identification of abnormal physiological states, sleep, unconsciousness, and other involuntary scenarios. Through a multi-frequency differential algorithm, it effectively counteracts environmental interference such as hand sweat, temperature, and contact pressure. Verification can be completed during the user's natural contact with the device without any additional actions. Attached Figure Description

[0016] Figure 1 This is the overall system architecture diagram of the present invention.

[0017] Figure 2 This is a flowchart of the impedance feature extraction process of the present invention.

[0018] Figure 3 This is a diagram showing the position of the excitation electrode in Embodiment 1 of the present invention.

[0019] Figure 4 This is a diagram showing the position of the measuring electrodes in Embodiment 1 of the present invention.

[0020] Figure 5 This is a schematic diagram of sensor embedding in Embodiment 2 of the present invention.

[0021] Figure 6 This is a diagram showing the transmission path of the alarm signal for abnormal physiological state in Embodiment 2 of the present invention. Detailed Implementation

[0022] The present invention will be further described in detail below with reference to embodiments.

[0023] 1. Overall System Hardware Architecture

[0024] like Figure 1 As shown, the system of the present invention mainly includes the following modules: 1.1 Excitation Signal Generation Module: Generates a weak AC excitation signal (compliant with human safety standards) with an adjustable frequency (preferably 10Hz–1MHz) for penetrating the skin and stimulating deep tissues. This can be achieved using a DDS chip or a dedicated impedance analysis front-end (such as AD5933).

[0025] 1.2 Core Sensing Module: Employing at least one pair of excitation electrodes (TX) and measurement electrodes (RX), forming a four-electrode or simplified two-electrode system. The electrodes are made of biocompatible conductive materials, specifically ITO, stainless steel, or conductive silicone, to minimize contact resistance. In mobile phone applications, the electrodes are integrated into the middle section of both side bezels of the phone, forming a closed loop when the user holds the device. Excitation current flows in through the TX, passes through the hand and forearm tissue, and flows out through the RX, completing the measurement.

[0026] 1.3 Signal Processing Module: Connected to the measuring electrodes, this module processes weak response voltage signals. A typical circuit includes a differential amplifier (amplifying the differential voltage and suppressing common-mode interference), a bandpass filter (filtering out power frequency and high-frequency noise), and an ADC (converting analog signals to digital signals).

[0027] 1.4 Computation and Processing Module: Composed of MCU, DSP or SoC with NPU, with built-in core algorithm software, including multi-frequency differential algorithm unit (extracting stable impedance features) and AI intent classifier (classifying features based on trained model).

[0028] 1.5 Execution Output Module: Performs corresponding operations based on the judgment result, such as sending "unlock / lock" commands, triggering alarms, recording encrypted logs, or communicating with the cloud.

[0029] 2. Core Algorithm Flow and Principles

[0030] like Figure 2 As shown, the core processing flow of this invention is as follows: Step S201: Multi-frequency sweep excitation and data acquisition.

[0031] The system controls the excitation signal generation module to scan multiple frequency points within a preset frequency band (e.g., 10kHz to 500kHz) in a stepped or continuous manner. At each frequency point The signal processing module simultaneously measures the amplitude and phase of the voltage and current, and calculates the complex impedance value at that frequency. : (1) in, Let be the resistance (real part). For reactance (imaginary part). The unit is the imaginary unit. This yields an impedance spectrum curve characterizing an individual biological tissue.

[0032] Step S202: Individual feature extraction based on biophysical model.

[0033] The impedance characteristics of biological tissues can be described using the Fricke equivalent circuit model. Among them, Represents extracellular fluid resistance. Represents intracellular fluid resistance. This represents the cell membrane capacitance. The complex impedance of this model... With angular frequency The relationship is described by the following equivalent circuit equation: (2) Its standard expanded form is: (3) in, extracellular fluid resistance represents the path of current flowing through the extracellular fluid as it bypasses the cell; Intracellular fluid resistance represents the path of an electric current through the intracellular fluid after it penetrates the cell membrane. Cell membrane capacitance is a core parameter characterizing the uniqueness of individual biological tissues. Its value is determined by the thickness, dielectric constant, and surface area of ​​the cell membrane.

[0034] Because everyone's physiological structure is different, their equivalent circuit parameter combinations ( Each digit is unique, which constitutes the "physical fingerprint" of a biometric.

[0035] To robustly extract this deep feature from the measured spectrum, this invention employs a multi-frequency differential algorithm. The algorithm calculates the impedance slope between different frequency points. f, to suppress linear noise caused by factors such as sweat on the skin surface. More preferably, the ratio of impedance modulus in the high-frequency band to that in the low-frequency band is calculated. : (4) in, This refers to the impedance modulus at high frequencies (e.g., 500kHz), where the current primarily penetrates the cell membrane and interacts with... Strong correlation For low-frequency bands (such as 10kHz), the impedance modulus is such that the current mainly bypasses the cell membrane and... Strong correlation. Ratio It can effectively counteract fluctuations in epidermal resistance and is compatible with individual... and The ratios are highly correlated, thus identifying deep tissue features with high uniqueness.

[0036] Step S203: Dynamic intent perception and state classification.

[0037] This invention identifies a user's physiological and psychological state by monitoring microscopic dynamic changes in impedance. (Skin total conductivity) At any time It can be modeled as: (5) in, It is a time-varying surface resistance affected by contact pressure and temperature. It is a time-varying electrical conductance of sweat glands controlled by the sympathetic nervous system, and is a key physiological signal that characterizes emotional stress (such as fear and tension).

[0038] When a person is in a state of abnormal physiological stress, the sympathetic nervous system is excited, leading to... It increases sharply within a very short time (1-3 seconds). To detect this change, the system monitors the complex impedance in real time at a selected characteristic frequency. And calculate its rate of change. If its absolute value If the preset physiological threshold is exceeded within a short period of time, the user is determined to be in an "abnormal physiological state" or "high stress state".

[0039] Step S204: AI-based fusion decision-making.

[0040] The static "identity feature vector" extracted from step S202 and the dynamic "intent feature parameters" obtained from step S203 are input together into the AI ​​intent classifier. This classifier preferably employs a CNN-LSTM hybrid neural network model. The CNN part is used to learn spatial features from the impedance spectrum curve, and the LSTM part is used to understand the sequential patterns of impedance changes over time (such as respiratory rhythm and microtremors). This model is trained to distinguish at least three states: Status A (Normal Voluntary): Identity characteristics match, and dynamic characteristics are stable or within the normal physiological fluctuation range.

[0041] State B (Involuntary / Abnormal Physiological State): Identity characteristics may match, but dynamic characteristics show a strong stress response (e.g., (Exceeding the threshold).

[0042] State C (unconscious / sleep): Identity characteristics may match, but dynamic characteristics show extremely low variability (e.g., variance approaches zero) and lack autonomic neural activity signals.

[0043] Step S205: Decision and Output.

[0044] Based on the output of the AI ​​classifier, the output module executes the corresponding instructions: If it is state A, an authentication pass signal is output to control the device to unlock or authorize.

[0045] If it is status B, a silent alarm protocol is triggered while verifying identity: the system can simulate normal operation (such as displaying successful unlock), but the background will automatically start recording and locating, and send encrypted alarm information, real-time physiological characteristics and geographical location to the preset security platform or contact person.

[0046] If the status is C, output an authentication failure or "invalid attempt" signal, refuse to unlock, and may choose to emit a local prompt tone to wake up the user.

[0047] 3. Example 1: Intelligent door lock system based on the present invention

[0048] like Figures 3-4 As shown, the following uses a smart door lock as an example to illustrate the specific application of the present invention.

[0049] 3.1 Hardware Integration

[0050] A four-electrode composite electrode array, including an excitation electrode pair (TX) and a measurement electrode pair (RX), is integrated inside the door lock handle. The electrodes are insulated from the metal handle. The processing unit is built into the door lock body and is made of indium tin oxide or stainless steel.

[0051] 3.2 Workflow

[0052] The user naturally grasps the door lock handle, preparing to open the door.

[0053] Normal authentication: After the user grasps the handle, the system completes a 5kHz-1MHz frequency sweep within 50ms to collect the impedance spectrum. The real-time data (e.g., impedance magnitude 450Ω, phase angle -14.2° at 50kHz) is compared with a pre-stored template. If the Euclidean distance is less than a threshold (e.g., 0.03 < 0.05) and the AI ​​does not detect any abnormal dynamic features, it is determined as "voluntary opening" and unlocking is performed.

[0054] Abnormal Physiological State Recognition: After successful authentication, if the system detects a sharp drop in the impedance modulus at 200kHz within 1.2 seconds (rate of change exceeding the threshold), and FFT analysis reveals a significant increase in micro-vibration noise power in the 30-60Hz frequency band (e.g., from -50dB to -32dB), the AI ​​determines it as "abnormal physiological state door opening." The system executes a "silent alarm protocol": while the door lock is opened normally to ensure user safety, the backend sends an encrypted alarm containing time, location, and other information to the security center and emergency contacts via Wi-Fi.

[0055] Unconscious state defense: If the impedance signal is detected to be extremely stable in the time domain (standard deviation close to 0) and lacks physiological rhythm fluctuations, the AI ​​determines it as "unconscious contact", the system refuses to unlock and can trigger a local sound and light alarm.

[0056] Table 1: Simulated data for individual identification (full palm scan): This demonstrates the difference in resistance between User A and Imposter B under different grip strengths, caused by differences in skeletal structure and muscle ratio:

[0057] Even when User A increases grip strength, the frequency response characteristics (phase angle and high-frequency impedance ratio) of the deep tissues remain highly stable, and the system determines that it is the same person.

[0058] Table 2: Simulation Data for Special State Awareness (Defensive Core)

[0059] Electrophysiological responses of User A in normal state, hijacked state, and sleep state (such as sleepwalking or having fingers dragged).

[0060] The above data are only typical calculation values ​​for embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. In practical applications, the system will perform self-learning and adaptive adjustments based on the baseline values ​​of different individuals.

[0061] In conclusion, due to its large contact area, the smart door lock can stably acquire the attenuation characteristics of high-frequency signals above 500kHz from the palm bones. This characteristic is resistant to sweat interference and has strong identity recognition capabilities. When a user is extremely nervous, the skin conductivity can increase by 3-5 times within 1.5 seconds. The system can accurately identify abnormal physiological states by monitoring the slope of this change. The muscle resistance caused by being forced to open the door will trigger high-frequency micro-tremors. According to FFT analysis, its noise power can increase from below -50dB under normal conditions to above -20dB, which becomes a reliable indicator for identifying involuntary operations.

[0062] 4. Example 2: As Figures 5-6 The mobile secure payment system shown is based on the present invention.

[0063] 4.1 Continuous Identity Maintenance

[0064] After the phone is unlocked, the system performs a lightweight verification every 30 seconds; it immediately locks the phone if a change in impedance characteristics is detected (due to a change in the person holding it); this prevents the risk of the phone being stolen after unlocking.

[0065] 4.2 Payment Intent Verification

[0066] When a user initiates a large transfer: a high-intensity verification is performed immediately before payment; if "fear characteristics" (phase angle fluctuation > 2°) or "antagonistic characteristics" (increased noise in a specific frequency band) are detected; The system will implement any of the following protections: display a fake "system busy" screen; limit payment amount (e.g., a maximum of 1000 yuan); require two-factor authentication (password + impedance); and simultaneously notify the security center in the background.

[0067] 5. Example 3: Application of Smart Wearable Devices

[0068] This invention can be integrated into wearable devices such as smartwatches, bracelets, and rings to achieve continuous identity authentication and health monitoring.

[0069] 5.1 Equipment Structure and Integration

[0070] A flexible quadrupole electrode array is integrated on the inside of the strap, using silver fiber fabric or conductive silicone electrodes to ensure a comfortable fit. The device incorporates a miniature impedance analysis front-end (AFE) and a low-power Bluetooth module.

[0071] 5.2 Working Mode

[0072] Continuous authentication mode: The device remains connected to the paired phone. When the user raises their wrist to operate the phone, the system completes authentication seamlessly, ensuring that it is being used by the rightful owner.

[0073] Health and Safety Monitoring Mode: Continuously monitors the baseline impedance spectrum and impedance micro-motion related to heart rate variability, achieving the following functions: Identify prolonged abnormal stress states and promptly prompt users to relax; Detect impedance transients that match the characteristics of an "abnormal physiological state" and automatically trigger recording or send location information to emergency contacts. Monitor impedance signal flattening trends (such as pre-syncope symptoms) to trigger equipment vibration alarms or automatic emergency calls.

[0074] The devices involved in the embodiments of this application include, but are not limited to, smartwatches, wristbands, smartphones, tablets, laptops, body composition analyzers, smart home devices, car central control systems and related in-vehicle devices.

[0075] The present invention and its embodiments have been described above. This description is not restrictive. If a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. An individual identification and intent perception system based on multi-frequency bioimpedance spectrum, characterized in that: include: The excitation signal generation module is used to generate multi-frequency stepped sweep current signals in the range of 10Hz to 1MHz. The core sensing module includes at least one pair of composite electrodes made of biocompatible conductive material, which are arranged on the surface of the device and form a closed circuit with the human body. The signal processing module includes a differential amplifier, a bandpass filter, and an analog-to-digital converter connected in sequence, used to acquire complex impedance data. ; in, This represents the resistance of electrolyte solutions such as body fluids and blood to electric current (pure resistance characteristic). This represents the cell membrane acting as a capacitor, impeding alternating current (capacitive characteristics). The imaginary unit represents the phase offset; The computational processing module is used to: calculate impedance characteristic parameters at different frequencies based on the impedance data; generate feature vectors for identity recognition based on the impedance characteristic parameters; extract dynamic feature parameters based on the time change information of the impedance characteristic parameters; and includes a multi-frequency differential algorithm unit and an AI intent classifier for extracting features and recognizing states, wherein the AI ​​intent classifier includes a convolutional neural network and / or a recurrent neural network. The execution output module is used to output authentication results or warning signals; The determination module is used to: compare the feature vector with a pre-stored template to output the identity authentication result; and determine whether there is an abnormal physiological state based on the dynamic feature parameters. The output module is used to output corresponding control signals based on the identity authentication result and the abnormal physiological state determination result; The feature vector includes the impedance amplitude ratio and phase angle corresponding to at least two preset frequency points. The phase angle ; The computational processing module integrates a deep learning-based classifier for fusion analysis of the physical fingerprint features and the dynamic conductivity model.

2. The system according to claim 1, characterized in that, The electrodes are arranged in a bipolar configuration, located in the middle of both sides of the device, forming a closed loop that runs through the palm, and are made of a biocompatible conductive material. The electrodes include discrete electrode sheets disposed on the surface of the housing, or a transparent conductive layer integrated under the display panel.

3. The system according to claim 1, characterized in that, The multi-frequency difference algorithm unit in the computation processing module performs the following steps: (a) Initiate a stepped frequency sweep excitation from 10 kHz to 500 kHz; (b) Obtain multiple sets of complex impedance data ; (c) Calculate the rate of change of impedance between different frequency points, i.e., the impedance slope. This is used to suppress linear noise caused by sweat. (d) Perform feature dimensionality reduction on the frequency response curve to extract the core resonance feature vector; (e) Input the extracted feature vector into the AI ​​intent classifier to classify and determine "normal unlocking", "sleep state" and "abnormal physiological state".

4. The system according to claim 3, characterized in that, The multi-frequency differential algorithm unit calculates the high-frequency to low-frequency impedance ratio. It is used to characterize individual differences in muscle and bone structure.

5. The system according to claim 1, characterized in that, The AI ​​intent classifier uses a CNN-LSTM hybrid model to identify impedance pulsation features and distinguish between "normal", "sleep" and "abnormal physiological state" states.

6. The system according to claim 1, characterized in that, The execution output module triggers a silent alarm when it detects impedance characteristic distortion, including recording, encrypting and uploading data and location information, and allows users to disable the recognition.

7. The system according to claim 1, characterized in that, It also includes an environmental adaptive compensation module, which is used to compensate for changes in skin impedance based on temperature and humidity.

8. The system according to claim 1, characterized in that, The computational processing module monitors dynamic conductivity. Identify stress states, the It is composed of time-varying surface resistance and skin conductance components connected in parallel, and skin conductance is related to sympathetic nerve activity; The system calculates the rate of change of the complex impedance Z at at least one selected frequency in the swept frequency signal over time, compares it with a threshold, and if the rate exceeds the threshold, it determines that the system is in a stress state and triggers an early warning.

9. The system according to claim 1, characterized in that: The dynamic feature parameters include at least one of the following: The dynamic feature parameters include at least one of the following: Rate of change of impedance over time; Statistical characteristics of impedance signals; The spectral characteristics of the impedance signal within a preset frequency band.

10. An identity recognition method for an individual identity recognition and intent perception system based on multi-frequency bioimpedance spectrum, characterized in that, include: When the user touches the device, the excitation signal generation module applies multi-frequency sweep excitation; Complex impedance data is acquired through the core sensing module. ; The signal processing module performs amplification, filtering, and A / D conversion. The computational processing module executes the multi-frequency difference algorithm and AI classification, and outputs the recognition results. The execution output module performs unlocking, locking, alarm, or data uploading operations based on the recognition results. The method supports cross-device frequency synchronization and encrypted near-field communication with the human body.

Images