Smart glasses and watch collaborative identity verification system and method

Abstract

The application discloses an intelligent glasses and watch cooperative identity verification system and method, belongs to the technical field of intelligent wearable devices, and aims to solve the problems of insufficient identity verification security and reliability. Technical solution points are as follows: the glasses and the intelligent watch are connected through a Bluetooth high-speed communication protocol, the glasses collect user pupil images and extract pupil feature vectors V p , the intelligent watch collects user voiceprint information and extracts voiceprint feature vectors V S ; the watch calculates V S the similarity of the voiceprint feature vectors with pre-stored voiceprint feature vectors V s0 , V p the similarity of the pupil feature vectors with pre-stored pupil feature vectors V p0 , if the similarities are not lower than a voiceprint similarity threshold value T S and a pupil similarity threshold value T p respectively, a verification pass signal is generated, and then the signal is uploaded to a blockchain for consensus verification through a practical Byzantine fault tolerance algorithm. The application is mainly used for identity verification of intelligent wearable devices.

Description

Technical Field

[0001] This invention belongs to the field of smart wearable device technology, specifically relating to a collaborative authentication system and method for smart glasses and watches. Background Technology

[0002] Currently, smart glasses and smartwatches, as two common types of wearable smart devices, have achieved certain functions in their respective fields. However, in identity verification scenarios, they typically operate independently. While smart glasses possess image acquisition capabilities, their built-in processing units have limited computing power, making it difficult to independently complete complex biometric recognition calculations. Simultaneously, due to size and power consumption limitations, their communication module performance is also insufficient, restricting their ability to interact with remote servers in real time. On the other hand, although smartwatches have relatively stronger processing power and communication links, their security still has room for improvement if they rely solely on a single device for identity authentication, such as relying only on voiceprint or password verification. Once single-factor authentication is compromised, the entire system will be at risk.

[0003] While existing technologies have attempted to utilize multiple biometric features for identity verification, most focus on single-device implementation or fail to adequately consider collaborative mechanisms between heterogeneous devices. For example, after smart glasses capture images, the data needs to be transmitted to a mobile phone or cloud for processing. This process introduces additional device nodes, increasing system complexity and transmission latency, and also raises security risks associated with data transfer between multiple devices. To establish direct collaboration between glasses and watches, it is necessary to solve the problem of stable and efficient data transmission between them and design a reasonable task allocation strategy to prevent any single device from becoming a performance bottleneck.

[0004] Furthermore, existing authentication methods largely rely on centralized servers for final decisions, posing a single point of failure risk. If verification results are stored only on a local device or a single server, they can be tampered with or forged, lacking non-repudiation and traceability. While blockchain technology is considered a potential solution due to its decentralized and immutable characteristics, integrating lightweight wearable devices as trusted nodes into the blockchain network and participating in the consensus process without impacting user experience remains a challenge. This requires the verification process itself to be both highly efficient and lightweight, while also meeting the fundamental security requirements of the blockchain consensus mechanism.

[0005] Therefore, the existing technology faces the following challenges: how to effectively integrate the hardware advantages of smart glasses and smartwatches to build a collaborative identity verification scheme that can achieve high-security dual authentication, ensure the real-time performance and reliability of the verification process, and at the same time, the scheme needs to be able to resist internal and external attacks and provide an auditable verification traceability mechanism. Summary of the Invention

[0006] One object of the present invention is to solve at least the above-mentioned problems and to provide at least the advantages that will be described later.

[0007] Another objective of this invention is to provide an identity verification method based on the collaborative work of glasses and smartwatches. This method can comprehensively utilize the pupil image acquisition capability of glasses and the voiceprint acquisition and strong computing power advantages of smartwatches. Through dual biometric cross-verification and blockchain consensus mechanism, it can significantly improve the security, tamper resistance and anti-attack capability of the entire verification process while ensuring the accuracy of identity verification.

[0008] To achieve these objectives and other advantages of the present invention, an authentication method based on the collaborative operation of glasses and a smartwatch is provided, comprising the following steps:

[0009] The glasses and smartwatch establish a connection via Bluetooth high-speed communication protocol;

[0010] The glasses capture images of the user's pupils and extract pupil feature vectors. V p ;

[0011] Smartwatches collect users' voiceprint information and extract voiceprint feature vectors V S ;

[0012] Smartwatch calculates voiceprint feature vector and compares it with pre-stored voiceprint feature vector V s0 The similarity is determined by the similarity score; if the similarity score is greater than or equal to the voiceprint similarity threshold, the similarity score is determined by the voiceprint similarity score. T S Then proceed to the next step;

[0013] The smartwatch calculates the pupil feature vector and compares it with the pre-stored pupil feature vector. V p0 The similarity is determined by the pupil similarity threshold. T p If so, a verification pass signal will be generated;

[0014] The smartwatch uploads the verification signal and verification data to the blockchain network, where blockchain nodes perform consensus verification of the verification transaction based on the Practical Byzantine Fault Tolerance algorithm.

[0015] If consensus verification passes, identity verification is complete.

[0016] Among them, voiceprint similarity threshold T S The pupil similarity threshold is between 0.8 and 0.9. T p The value ranges from 0.75 to 0.85.

[0017] The similarity between the voiceprint feature vector and the pre-stored voiceprint feature vector is calculated as follows:

[0018] ;

[0019] The similarity between the pupil feature vector and the pre-stored pupil feature vector is calculated as follows:

[0020] .

[0021] Preferably, the glasses and the smartwatch establish a connection and transmit data via Bluetooth high-speed communication protocol. The transmitted data includes pupil image data collected by the glasses and to be sent to the smartwatch, and voiceprint data collected by the smartwatch and to be used for comparison. The bit error rate of the Bluetooth high-speed communication protocol when transmitting the pupil image data and voiceprint data is no greater than 10. -5 The data transmission rate of the Bluetooth high-speed communication protocol R Satisfy the following formula:

[0022] ;

[0023] Where B is the channel bandwidth and S / N is the signal-to-noise ratio;

[0024] The data transmission rate R The value must ensure that the pupil image data and voiceprint data can be transmitted within the time limit required by the identity verification process.

[0025] Preferably, the voiceprint feature vector is extracted using Mel frequency cepstral coefficients, and the pupil feature vector is obtained through an image feature extraction algorithm.

[0026] Preferably, the voiceprint feature vector is extracted using Mel frequency cepstral coefficients, specifically including the following steps:

[0027] The user's voice signal is collected through the smartwatch's microphone, and the analog voice signal is pre-emphasized to enhance the high-frequency components.

[0028] The pre-emphasized audio signal is divided into frames and windowed, with a frame length of 20 to 40 milliseconds and a frame shift of 10 to 20 milliseconds.

[0029] Perform a Fast Fourier Transform on each frame of signal to convert it to the frequency domain and obtain a linear spectrum;

[0030] The linear spectrum is passed through a set of Mel-scale filters that approximate a linear distribution below 1000 Hz and a logarithmic distribution above 1000 Hz to simulate the nonlinear perception of sound by the human ear.

[0031] Taking the logarithm of the filter bank's output yields the log-Mel spectrum;

[0032] Perform a discrete cosine transform on the logarithmic Mel spectrum and take the first 12 to 16 coefficients as the static Mel frequency cepstral coefficients;

[0033] The first-order and second-order difference coefficients of the static Mel frequency cepstral coefficients are calculated and together constitute the final voiceprint feature vector.

[0034] Preferably, the pupil feature vector is obtained through an image feature extraction algorithm, specifically including the following steps:

[0035] The glasses' camera captures infrared images of the user's eyes, and the images are preprocessed by grayscale conversion and histogram equalization.

[0036] The pupil region is located using a circular detection algorithm based on Hough transform. The parameter accumulator of the circular detection algorithm has a resolution of 1 to 2 pixels, and the minimum center distance is 1.5 to 2 times the expected radius of the pupil.

[0037] The effective iris texture area is segmented from the located pupil area, excluding the parts obscured by the eyelids and eyelashes;

[0038] The segmented annular iris region is geometrically normalized using a rubber membrane model and mapped to a rectangular iris image of fixed size, where the horizontal axis of the rectangle corresponds to the circumferential direction of the iris and the vertical axis corresponds to the radial direction of the iris.

[0039] The normalized iris region is divided into several sector regions, each with a radial resolution of 8 to 16 pixels and an angular resolution of 5 to 10 degrees.

[0040] A two-dimensional Gabor filter bank is applied to each sector region for texture feature extraction. The wavelength range of the Gabor filter bank is 3 to 15 pixels, the number of directions is 4 to 8, and the phase shift is 0 and π / 2.

[0041] Calculate the mean and standard deviation of the output response of each filter, which together constitute the local feature vector of the sector region;

[0042] The local feature vectors of all fan-shaped regions are concatenated, and principal component analysis is performed on the concatenated high-dimensional feature vector to reduce its dimensionality. Principal components with a cumulative contribution rate of 85% to 95% are retained to form the final pupil feature vector.

[0043] Preferably, the smartwatch verifies the blockchain network to which the verification signal and verification data are uploaded, forming a consortium blockchain containing the smartwatch and the glasses as consensus nodes; the blockchain network employs a practical Byzantine fault-tolerant algorithm to perform consensus verification on the identity verification transaction; and the total number of nodes N and the number of malicious nodes f in the blockchain network satisfy the following relationship: The smartwatch and the glasses, acting as two light nodes in the blockchain network, participate in the voting and verification process for the validity of identity verification transactions. Preferably, the smartwatch also receives and stores identity verification-related message data from the glasses, including identity verification trigger instructions, verification process logs, or pupil image feature vectors from the current verification. If the amount of identity verification-related message data exceeds the remaining storage space of the smartwatch, the identity verification-related message data is compressed using Huffman coding before storage. Preferably, the step of compressing the identity verification-related message data using Huffman coding includes:

[0044] Perform probability statistics on the symbols in the message data;

[0045] Construct a Huffman tree based on the probability of symbol occurrence;

[0046] Each symbol is assigned a variable-length code, where symbols with high occurrence probability are assigned short codes and symbols with low occurrence probability are assigned long codes; the average code length L of the Huffman code satisfies:

[0047] ;

[0048] in, p i The probability of the symbol appearing. l i For the corresponding encoding length;

[0049] The average coding length L This value is used to evaluate the compression effect of the authentication-related message data; the smaller the value, the higher the compression efficiency.

[0050] Preferably, during the identity verification process, the collaborative allocation of computing power between the glasses and the smartwatch satisfies the following relationship: The smartwatch, with its higher computing power resources, undertakes the core computing tasks of voiceprint feature extraction, dual biometric similarity calculation, and blockchain transaction construction and uploading; the glasses undertake the computing load of pupil image acquisition and preprocessing; the total effective computing power of the system... C total The value used to complete the entire identity verification process is determined by the following formula:

[0051] ;

[0052] in, C g For computing power of glasses, C w The computing power of the smartwatch is represented by α, which is the computing power utilization coefficient, ranging from 0.8 to 1.0.

[0053] The computing power utilization coefficient α is dynamically adjusted according to the real-time requirements of the identity verification task and the current network status to ensure the smooth execution of the identity verification process.

[0054] A smart wearable system based on the collaborative operation of glasses and a smartwatch is used to perform the aforementioned authentication method, the system comprising smart glasses and a smartwatch;

[0055] The smart glasses include:

[0056] A camera module configured to capture images of the user's eyes;

[0057] The first processing module is configured to preprocess the eye image and extract the pupil feature vector;

[0058] A first Bluetooth communication module is configured to establish a connection with the smartwatch and transmit data.

[0059] The smartwatch includes:

[0060] Microphone module, configured to capture the user's voice signal;

[0061] The second processing module is configured to process the sound signal and extract the voiceprint feature vector, and calculate the similarity between the voiceprint feature vector and the pre-stored voiceprint feature vector, as well as the similarity between the pupil feature vector and the pre-stored pupil feature vector.

[0062] The second Bluetooth communication module is configured to establish a connection with the smart glasses and transmit data.

[0063] A blockchain communication module, configured to upload verification pass signals and verification data to the blockchain network;

[0064] The storage module is configured to store pre-stored voiceprint feature vectors, pre-stored pupil feature vectors, and message data from the smart glasses.

[0065] The first Bluetooth communication module and the second Bluetooth communication module establish a connection via the Bluetooth High Speed ​​Communication Protocol, wherein the bit error rate of the Bluetooth High Speed ​​Communication Protocol during data transmission is no greater than 10. -5 ;

[0066] The smart glasses and the smartwatch work together to complete the identity verification process.

[0067] The present invention has at least the following beneficial effects:

[0068] 1. This invention collects pupil and voiceprint features through glasses and a smartwatch respectively, performs dual comparison, and finally verifies them through blockchain consensus. This solves the problem of insufficient security of single-device verification and significantly improves the tamper-proof and attack-resistant capabilities of the verification process.

[0069] 2. This invention solves the problems of real-time performance and integrity of biometric data transmission between devices by limiting the upper limit of the bit error rate of Bluetooth transmission and ensuring the transmission rate based on the channel capacity model, thus ensuring the reliability of the verification process.

[0070] 3. This invention uses Mel frequency cepstral coefficients and image feature extraction algorithms to process voiceprint and pupil images respectively, solving the technical problem of standardization of biometric feature extraction and providing a basic guarantee for high-precision feature comparison.

[0071] 4. This invention solves the problem of poor stability of voiceprint features in noisy environments by using a complete MFCC feature extraction process, including pre-emphasis, framing, and frequency domain transformation, thereby improving feature discrimination and anti-interference ability.

[0072] 5. This invention solves the problem of unstable pupil feature extraction in non-cooperative environments by processing steps such as image preprocessing, pupil localization, iris segmentation, and Gabor filtering, thereby improving the accuracy of feature recognition.

[0073] 6. This invention solves the problem of malicious nodes tampering with data in a distributed verification environment by incorporating devices as light nodes into the consortium blockchain and following the PBFT consensus mechanism, thus ensuring the immutability of verified transactions.

[0074] 7. This invention solves the problem of limited storage space in mobile devices by receiving and storing verification data on the watch and compressing it when needed, thus enabling long-term preservation of verification process data.

[0075] 8. This invention solves the problem of low data storage efficiency for verification data by using a Huffman coding compression scheme, which significantly reduces storage space usage while ensuring data integrity.

[0076] 9. This invention solves the problem of unreasonable allocation of computing tasks between two devices by using a dynamic computing power collaborative allocation model, optimizes the utilization of system resources, and ensures the real-time performance of the verification process.

[0077] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Detailed Implementation

[0078] The present invention will now be described in further detail so that those skilled in the art can implement it based on the description.

[0079] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0080] It should be noted that, unless otherwise specified, the experimental methods described in the following implementation plan are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified.

[0081] Example 1

[0082] An authentication method based on the collaborative operation of glasses and a smartwatch includes the following steps:

[0083] The glasses and smartwatch establish a connection via Bluetooth high-speed communication protocol;

[0084] The glasses capture images of the user's pupils and extract the pupil feature vector V. p ;

[0085] Smartwatches collect users' voiceprint information and extract voiceprint feature vector V S ;

[0086] The smartwatch calculates the voiceprint feature vector and compares it with the pre-stored voiceprint feature vector V. s0 The similarity is determined by the similarity between the voiceprint and the voiceprint. If the similarity is greater than or equal to the voiceprint similarity threshold T, then... S Then proceed to the next step;

[0087] The smartwatch calculates the pupil feature vector and compares it with the pre-stored pupil feature vector V. p0 The similarity is determined by the pupil similarity threshold T. p If so, a verification pass signal will be generated;

[0088] The smartwatch uploads the verification signal and verification data to the blockchain network, where blockchain nodes perform consensus verification of the verification transaction based on the Practical Byzantine Fault Tolerance algorithm.

[0089] If consensus verification passes, identity verification is complete.

[0090] Among them, the voiceprint similarity threshold T S The pupil similarity threshold T is between 0.8 and 0.9. p The value ranges from 0.75 to 0.85.

[0091] The similarity between the voiceprint feature vector and the pre-stored voiceprint feature vector is calculated as follows:

[0092] ;

[0093] The similarity between the pupil feature vector and the pre-stored pupil feature vector is calculated as follows:

[0094] .

[0095] In one specific embodiment of the invention, a user wears smart glasses and a smartwatch, which automatically pair and maintain a connection via Bluetooth. When authentication is required, the smart glasses' built-in camera captures an image of the user's pupils and performs preliminary processing locally to generate a pupil feature vector. Simultaneously, the smartwatch records a segment of the user's voice through its microphone and extracts a voiceprint feature vector. The smartwatch then calculates the similarity between the received pupil feature vector and the locally extracted voiceprint feature vector, comparing them to pre-registered templates. If the voiceprint comparison passes first, pupil comparison continues; if both reach a preset similarity threshold, the smartwatch generates a verification pass signal. This signal, along with key data for this verification, is uploaded to a blockchain network composed of smartwatches, smart glasses, and other optional device nodes. These nodes use a practical Byzantine fault-tolerant algorithm to reach consensus on the verification transaction. Once consensus is reached, the authentication is finally completed.

[0096] This differs significantly from the closest existing technologies. Existing smart glasses or smartwatches often operate independently; for example, a smartwatch may only perform voiceprint verification, or smart glasses may attempt single-factor pupil recognition. This single-factor verification method compromises system security if counterfeited or attacked. Furthermore, existing technologies often rely on local device storage or transmission to a central server for judgment, posing a single point of failure and the risk of data tampering. This invention achieves cross-verification of dual biometrics through dual-device collaboration and utilizes the decentralized consensus mechanism of blockchain to confirm the verification results, thereby significantly improving the security and reliability of the identity verification process without requiring complex centralized infrastructure.

[0097] Furthermore, the glasses and the smartwatch establish a connection and transmit data via Bluetooth high-speed communication protocol. The transmitted data includes pupil image data collected by the glasses and to be sent to the smartwatch, and voiceprint data collected by the smartwatch and to be used for comparison. The bit error rate of the Bluetooth high-speed communication protocol when transmitting the pupil image data and voiceprint data is no greater than 10. -5 The data transmission rate R of the Bluetooth high-speed communication protocol satisfies the following formula:

[0098] ;

[0099] Where B is the channel bandwidth and S / N is the signal-to-noise ratio;

[0100] The value of the data transmission rate R must ensure that the pupil image data and voiceprint data can be transmitted within the time limit required by the identity verification process.

[0101] In this embodiment, once the smart glasses and smartwatch establish a Bluetooth connection, they immediately begin transmitting the biometric data required for authentication. The smart glasses send the pupil image feature vectors they have collected and processed to the smartwatch via the Bluetooth link. Simultaneously, the smartwatch processes its own raw voiceprint data locally. The entire transmission process has strict requirements for communication quality. The Bluetooth protocol is configured to prioritize the transmission of this critical data, ensuring an extremely low bit error rate even in complex wireless environments, and that the transmission rate is sufficient to meet the real-time requirements of the authentication process, allowing for smooth dual-feature comparison without user waiting.

[0102] This contrasts with common existing technologies. In most current solutions, data transmission between smart devices is typically based on standard, unoptimized Bluetooth protocols, whose transmission stability is susceptible to environmental interference, potentially resulting in packet loss or latency fluctuations. When transmitting larger amounts of data, such as pupil images, this uncertainty can cause the verification process to stall or be interrupted, impacting user experience, and in extreme cases, even leading to verification failure due to incomplete data. This invention effectively overcomes these reliability challenges by optimizing the transmission assurance mechanism for the specific scenario of identity verification, providing a stable and reliable data path for collaborative verification between two devices.

[0103] Furthermore, the voiceprint feature vector is extracted using Mel frequency cepstral coefficients, and the pupil feature vector is obtained through an image feature extraction algorithm.

[0104] In existing technologies, biometric extraction schemes for wearable devices are often simple or non-standardized, such as directly using raw voiceprint spectra or simple image contours as features. The feature vectors extracted by this method are easily affected by environmental noise, changes in lighting, or slight changes in user posture, resulting in poor feature stability and making it difficult to guarantee the accuracy and reliability of subsequent comparisons. This invention employs a widely recognized and thoroughly validated feature extraction algorithm, providing a stable, consistent, and repeatable computational foundation for dual biometrics. This ensures the consistency of feature generation across different devices and verification processes, thus laying a solid technical foundation for high-precision identity comparison.

[0105] In this embodiment, after receiving a voice signal, the smartwatch extracts voiceprint features using a Mel frequency cepstral coefficient process. This process includes a series of standardization operations such as pre-emphasis, framing, frequency domain transformation, filtering with a Mel filter bank, logarithmic calculation, and discrete cosine transform, ultimately generating a voiceprint feature vector with good discriminative power. Simultaneously, the smart glasses process the acquired eye images, locating the pupil using image algorithms, segmenting the effective iris region, performing geometric normalization, and extracting iris texture features using a Gabor filter bank. After dimensionality reduction, a stable pupil feature vector is formed.

[0106] Furthermore, the extraction of the voiceprint feature vector using Mel frequency cepstral coefficients specifically includes the following steps:

[0107] The user's voice signal is collected through the smartwatch's microphone, and the analog voice signal is pre-emphasized to enhance the high-frequency components.

[0108] The pre-emphasized audio signal is divided into frames and windowed, with a frame length of 20 to 40 milliseconds and a frame shift of 10 to 20 milliseconds.

[0109] Perform a Fast Fourier Transform on each frame of signal to convert it to the frequency domain and obtain a linear spectrum;

[0110] The linear spectrum is passed through a set of Mel-scale filters that approximate a linear distribution below 1000 Hz and a logarithmic distribution above 1000 Hz to simulate the nonlinear perception of sound by the human ear.

[0111] Taking the logarithm of the filter bank's output yields the log-Mel spectrum;

[0112] Perform a discrete cosine transform on the logarithmic Mel spectrum and take the first 12 to 16 coefficients as the static Mel frequency cepstral coefficients;

[0113] The first-order and second-order difference coefficients of the static Mel frequency cepstral coefficients are calculated and together constitute the final voiceprint feature vector.

[0114] In this embodiment, when the smartwatch performs voiceprint feature extraction, it first pre-emphasizes the acquired speech signal to enhance high-frequency components. Then, the signal is divided into short frames and windowed. Each frame is converted to the frequency domain via a Fast Fourier Transform (FFT) and then mapped to a more perceptual Mel-scale using a set of Mel-scale filters that simulate human hearing. The logarithm of the filter output is then taken, followed by a Discrete Cosine Transform (DCT) to obtain the static Mel-frequency cepstral coefficients. To fully capture the dynamic characteristics of the sound, the system further calculates the first and second differences of these coefficients, ultimately combining these dynamic and static features to form a complete feature vector characterizing the user's voice.

[0115] The processing method of this invention contrasts sharply with common existing technologies. Existing technologies often stop at static coefficients or simply calculate first-order differences. When a user's voice changes in the short term due to a cold, emotional fluctuations, or ambient noise, the extracted features are prone to significant drift, directly affecting the accuracy of the comparison. This invention, by introducing a complete dynamic feature extraction process, enhances the ability to describe short-term changes in sound, making the extracted feature vectors more resistant to various interferences in real-world application scenarios, and improving the robustness and reliability of the system in practical environments.

[0116] Furthermore, the pupil feature vector is obtained through an image feature extraction algorithm, specifically including the following steps:

[0117] The glasses' camera captures infrared images of the user's eyes, and the images are preprocessed by grayscale conversion and histogram equalization.

[0118] The pupil region is located using a circular detection algorithm based on Hough transform. The parameter accumulator of the circular detection algorithm has a resolution of 1 to 2 pixels, and the minimum center distance is 1.5 to 2 times the expected radius of the pupil.

[0119] The effective iris texture area is segmented from the located pupil area, excluding the parts obscured by the eyelids and eyelashes;

[0120] The segmented annular iris region is geometrically normalized using a rubber membrane model and mapped to a rectangular iris image of fixed size, where the horizontal axis of the rectangle corresponds to the circumferential direction of the iris and the vertical axis corresponds to the radial direction of the iris.

[0121] The normalized iris region is divided into several sector regions, each with a radial resolution of 8 to 16 pixels and an angular resolution of 5 to 10 degrees.

[0122] A two-dimensional Gabor filter bank is applied to each sector region for texture feature extraction. The wavelength range of the Gabor filter bank is 3 to 15 pixels, the number of directions is 4 to 8, and the phase shift is 0 and π / 2.

[0123] Calculate the mean and standard deviation of the output response of each filter, which together constitute the local feature vector of the sector region;

[0124] The local feature vectors of all fan-shaped regions are concatenated, and principal component analysis is performed on the concatenated high-dimensional feature vector to reduce its dimensionality. Principal components with a cumulative contribution rate of 85% to 95% are retained to form the final pupil feature vector.

[0125] In this embodiment, the smart glasses' camera first captures an infrared image of the user's eye, then performs grayscale conversion and histogram equalization to enhance contrast. Next, the system uses a circular detection algorithm based on Hough transform to accurately locate the pupil region and segments the effective iris texture region, unaffected by eyelids and eyelashes, using image processing techniques. This circular iris region is then geometrically normalized using a rubber membrane model, transforming it into a fixed-size rectangular image to eliminate scale and deformation differences caused by wearing distance or angle. On this rectangular image, the system further divides it into multiple sector regions and applies a set of multi-directional, multi-scale two-dimensional Gabor filters to each sector region to extract detailed texture features. Finally, by calculating the statistical values ​​of each filter response and concatenating all local features, and then performing dimensionality reduction through principal component analysis, a compact and highly discriminative final pupil feature vector is obtained.

[0126] Traditional methods often extract global features directly from the entire eye image or perform only simple pupil contour recognition. These methods are highly sensitive to changes in lighting, slight head movements, and partial occlusion, resulting in poor feature repeatability and large fluctuations in recognition rate. This invention, through a series of standardized and adaptive processing steps, significantly improves the stability and anti-interference ability of pupil features extracted in real, non-cooperative environments, providing a reliable guarantee for subsequent high-precision identity verification.

[0127] Furthermore, the smartwatch verifies that the blockchain network to which the signal and verification data are uploaded is a consortium blockchain containing the smartwatch and the glasses as consensus nodes; the blockchain network uses a practical Byzantine fault-tolerant algorithm to perform consensus verification on the identity verification transaction; the total number of nodes N and the number of malicious nodes f in the blockchain network satisfy the following relationship: In this embodiment, the smartwatch and the glasses act as two light nodes in the blockchain network, participating in the voting and verification process for the validity of the identity verification transaction. After the smartwatch completes the dual biometric comparison and generates a verification pass signal, this signal and related verification data are uploaded to a consortium blockchain network specifically built for identity verification. This blockchain network is unique in that its consensus nodes not only include traditional server nodes but also incorporate the participating smartwatch and smart glasses as light nodes. These device nodes collectively employ a practical Byzantine fault-tolerant algorithm to perform consensus verification of the identity verification transaction. If a few malicious nodes in the network attempt to tamper with or deny the verification result, as long as the number of honest nodes meets the protocol requirements, the system can still reach a consensus on the validity of the transaction and ultimately confirm successful verification.

[0128] This is fundamentally different from the closest existing technologies. Traditional identity verification schemes often rely on a single central server to receive and store verification results. Once this server is attacked or malfunctions, the authenticity and non-repudiation of the entire verification process cannot be guaranteed. Even if some schemes attempt to introduce distributed storage, most of them are only used for data backup and lack consensus mechanisms for data consistency and transaction finality. This invention, by directly incorporating the wearable device initiating verification into the blockchain consensus system, makes the device itself part of the trusted decision-making process, fundamentally eliminating the risks that may be caused by the breach or malicious operation of a single central node, and achieving highly reliable identity verification in a decentralized environment.

[0129] Furthermore, the smartwatch is also used to receive and store authentication-related message data from the glasses. The authentication-related message data includes authentication trigger instructions, authentication process logs, or pupil image feature vectors for this authentication. If the amount of the authentication-related message data is greater than the remaining storage space of the smartwatch, the authentication-related message data is compressed using Huffman coding before storage.

[0130] In this embodiment, the smartwatch, while completing the core identity verification process, is also responsible for receiving and storing various verification process data sent by the smart glasses, such as verification trigger commands, detailed operation logs, and pupil feature vectors used in this verification. This data is crucial for subsequent auditing, troubleshooting, and model optimization. Considering the limitations of the smartwatch's local storage space, if the amount of data to be stored at once exceeds the remaining capacity, the system will automatically initiate a lossless compression process, using Huffman coding to compress the data before storing it in the storage module, thereby retaining historical verification records for a longer period within a limited space.

[0131] In similar scenarios, existing smart wearable devices typically employ two simple strategies: first, storing only the minimum result data and discarding rich intermediate process information, leading to a lack of basis for subsequent analysis; second, uploading data to the cloud in real time, which not only relies on a stable network environment but also poses a risk of data privacy leakage. This invention introduces an intelligent compression storage mechanism on the device side, achieving long-term and complete retention of key process data without external dependencies, effectively supporting the traceability and continuous optimization capabilities of the identity verification system. Furthermore, the steps for Huffman coding compression of identity verification-related message data include:

[0132] Perform probability statistics on the symbols in the message data;

[0133] Construct a Huffman tree based on the probability of symbol occurrence;

[0134] Each symbol is assigned a variable-length code, where symbols with high occurrence probability are assigned short codes and symbols with low occurrence probability are assigned long codes; the average code length L of the Huffman code satisfies:

[0135] ;

[0136] Where, p i l represents the probability of the symbol appearing. i For the corresponding encoding length;

[0137] The average encoding length L is used to evaluate the compression effect of the authentication-related message data; the smaller the value, the higher the compression efficiency.

[0138] In this implementation, when the smartwatch needs to compress the received verification process data, it initiates a Huffman coding process. The system first analyzes the frequency of occurrence of various symbols in the message data to be compressed and constructs an optimal Huffman tree accordingly. Then, based on this tree, a unique variable-length code is assigned to each symbol; frequently occurring symbols receive shorter codes, while less frequently occurring symbols receive longer codes. In this way, data that originally required fixed-length storage is transformed into a compact bit sequence composed of codewords of varying lengths, thereby significantly reducing the data size without losing any information.

[0139] In existing technologies, mobile devices often employ simple fixed-length truncation or lossy compression methods to save storage space. These methods either fail to effectively reduce data volume or compromise data integrity and parsability, posing challenges for subsequent auditing and analysis. The adaptive Huffman coding technique employed in this invention dynamically generates the optimal coding table based on the actual statistical characteristics of each verification data point, achieving efficient and completely lossless compression and ensuring that all critical verification details are preserved in their original form.

[0140] Furthermore, during the identity verification process, the collaborative allocation of computing power between the glasses and the smartwatch satisfies the following relationship: The smartwatch, with its higher computing power resources, undertakes the core computing tasks of voiceprint feature extraction, dual biometric similarity calculation, and blockchain transaction construction and uploading; the glasses undertake the computing load of pupil image acquisition and preprocessing; the total effective computing power Ctotal of the system is used to complete the entire identity verification process, and its value is determined by the following formula:

[0141] ;

[0142] Among them, C g For computing power of glasses, C w The computing power of the smartwatch is represented by α, which is the computing power utilization coefficient, ranging from 0.8 to 1.0.

[0143] The computing power utilization coefficient α is dynamically adjusted according to the real-time requirements of the identity verification task and the current network status to ensure the smooth execution of the identity verification process.

[0144] In this embodiment, the computational tasks of the entire identity verification process are intelligently distributed between the smart glasses and the smartwatch. Leveraging its image acquisition capabilities, the smart glasses are primarily responsible for real-time capture and preliminary processing of pupil images to form feature vectors. Tasks with higher computational demands, including voiceprint feature extraction, precise comparison of the two biometric features, and the construction of blockchain transactions, are handled by the smartwatch, which boasts greater computing power. The system can dynamically assess the urgency of the current task and network conditions, adjusting the allocation of computational resources between the two in real time to ensure a highly efficient and smooth verification process.

[0145] Existing multi-device collaboration technologies often employ static or preset task allocation strategies, which cannot adapt to real-time changes in device status and network environment. This can easily lead to one device being overloaded while another remains idle, resulting in verification delays or resource waste. This invention introduces a dynamic computing power collaboration mechanism, enabling dual devices to flexibly adjust their computing workload based on real-time needs, fully leveraging their respective hardware advantages. This ensures that the authentication system maintains high responsiveness and stability even in complex application scenarios.

[0146] Example 2

[0147] A smart wearable system based on the collaborative operation of glasses and a smartwatch is used to perform the aforementioned authentication method, the system comprising smart glasses and a smartwatch;

[0148] The smart glasses include:

[0149] A camera module configured to capture images of the user's eyes;

[0150] The first processing module is configured to preprocess the eye image and extract the pupil feature vector;

[0151] A first Bluetooth communication module is configured to establish a connection with the smartwatch and transmit data.

[0152] The smartwatch includes:

[0153] Microphone module, configured to capture the user's voice signal;

[0154] The second processing module is configured to process the sound signal and extract the voiceprint feature vector, and calculate the similarity between the voiceprint feature vector and the pre-stored voiceprint feature vector, as well as the similarity between the pupil feature vector and the pre-stored pupil feature vector.

[0155] The second Bluetooth communication module is configured to establish a connection with the smart glasses and transmit data.

[0156] A blockchain communication module, configured to upload verification pass signals and verification data to the blockchain network;

[0157] The storage module is configured to store pre-stored voiceprint feature vectors, pre-stored pupil feature vectors, and message data from the smart glasses.

[0158] The first Bluetooth communication module and the second Bluetooth communication module establish a connection via the Bluetooth High Speed ​​Communication Protocol, wherein the bit error rate of the Bluetooth High Speed ​​Communication Protocol during data transmission is no greater than 10. -5 ;

[0159] The smart glasses and the smartwatch work together to complete the identity verification process.

[0160] Application example: The application of a smart glasses and smartwatch collaborative identity verification system in mobile payment scenarios.

[0161] A user wearing smart glasses and a smartwatch has established a stable connection via Bluetooth high-speed communication. The user attempts to withdraw cash from an ATM using facial recognition, triggering a high-security authentication process.

[0162] Step 1: Device Connection and Biometric Data Collection

[0163] The smart glasses activate an infrared camera to capture images of the user's eyes. After preprocessing with grayscale conversion and histogram equalization, a circular detection algorithm based on Hough transform (accumulator resolution 1.5 pixels, minimum center distance 1.8 times the expected pupil radius) is used to accurately locate the pupil region. The effective iris texture region is segmented and geometrically normalized into a fixed-size rectangular image using a rubber membrane model, divided into 12 sector regions (radial resolution 10 pixels, angular resolution 8°). Each region is subjected to 4-direction, 6-scale Gabor filtering (wavelength range 5~12 pixels, phase offset 0 and π / 2) to extract texture features. The mean and standard deviation of the response values ​​constitute a local feature vector, which is then subjected to PCA dimensionality reduction (retaining principal components with a cumulative contribution rate of 90%) to generate the final pupil feature vector. V p .

[0164] Simultaneously, the smartwatch captures the user's voice signal as they read dynamically generated random phrases via its microphone, and performs MFCC feature extraction: after pre-emphasis processing, the signal is divided into frames (25ms frame length, 12ms frame shift), windowed, and each frame signal is converted to the frequency domain via FFT. Then, it passes through a 40-Melbourne filter bank (linear-logarithmic transition frequency 1000Hz), and after taking the logarithm, it undergoes DCT transformation to extract the first 14 static coefficients. First-order and second-order difference coefficients are calculated to collectively form the voiceprint feature vector. V S .

[0165] Step 2: Feature Transmission and Local Comparison

[0166] The smart glasses will extract the pupil feature vector V p Data is transmitted to the smartwatch via Bluetooth high-speed communication protocol. This protocol guarantees a transmission error rate of less than 10%. -5 Transmission rate R Satisfying Shannon's formula The measured speed is 2.1 Mbps, ensuring that the feature data is delivered completely within 150ms and ensuring real-time and reliable data transmission.

[0167] Smartwatch receiver V p Then, calculate its relationship with the pre-stored pupil feature vector. V p0 similarity S p ,as well as V S With pre-stored voiceprint feature vector V s0 similarity S s Similarity is calculated using the cosine similarity formula:

[0168] ;

[0169] Set voiceprint similarity threshold T S = 0.85, pupil similarity threshold T p = 0.8. If S S ≥T S and S p ≥T p If so, a verification pass signal is generated. In this example... S S ≈ 0.92, S p≈ The values ​​are all 0.98, exceeding the threshold, and local verification is successful.

[0170] Step 3: Blockchain Consensus Verification

[0171] The smartwatch constructs and verifies a transaction data packet (including a timestamp, feature hash value, and device ID), and uploads it to the consortium blockchain network (containing the smartwatch, the watch itself, and three bank server nodes, for a total of N=5 nodes). The network uses the PBFT consensus mechanism, tolerating a maximum of [missing information - likely a specific number of nodes]. (A maximum of f = 1 malicious node can be tolerated). Each node votes on the validity of a transaction, and once a consensus is reached, the transaction is written to the blockchain. The verification result cannot be tampered with.

[0172] Step 4: Data storage and compression during the verification process:

[0173] The smartwatch receives and stores the verification process data (including raw feature vectors and operation logs) sent by the glasses, totaling 15MB. With only 12MB of remaining storage space, Huffman compression is triggered: a Huffman tree is constructed after statistically probabilities of the symbols, assigning short codes to high-frequency symbols, resulting in an average code length of... , L =2.4 bits / symbol, the compressed data size was reduced to 9.8MB, and the data was successfully stored.

[0174] Step 5: Dynamic Computing Power Collaboration and Task Allocation

[0175] The system dynamically allocates computing power based on real-time load: smart glasses ( C g = 50MIPS) is responsible for image acquisition and preprocessing; smartwatch (computing power) C w = 200 MIPS) is responsible for voiceprint extraction, dual-feature comparison, and blockchain transaction construction (computing power utilization coefficient α = 0.95). Total effective computing power of the system: ;Right now C total =50 + 0.95 * 200 = 240 MIPS. The entire computation task is completed within 1.2 seconds, meeting the real-time requirements.

[0176] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details.

Claims

1. An authentication method based on the collaborative operation of glasses and a smartwatch, characterized in that, Includes the following steps: The glasses and smartwatch establish a connection via Bluetooth high-speed communication protocol; The glasses capture images of the user's pupils and extract pupil feature vectors. V p ; Smartwatches collect users' voiceprint information and extract voiceprint feature vectors V S ; Smartwatch calculates voiceprint feature vector and compares it with pre-stored voiceprint feature vector V s0 The similarity is determined by the similarity score; if the similarity score is greater than or equal to the voiceprint similarity threshold, the similarity score is determined by the voiceprint similarity score. T S Then proceed to the next step; The smartwatch calculates the pupil feature vector and compares it with the pre-stored pupil feature vector. V p0 The similarity is determined by the pupil similarity threshold. T p Then a verification pass signal is generated; The smartwatch uploads the verification signal and verification data to the blockchain network, where blockchain nodes perform consensus verification of the verification transaction based on the Practical Byzantine Fault Tolerance algorithm. If consensus verification passes, identity verification is complete. Among them, voiceprint similarity threshold T S The pupil similarity threshold is between 0.8 and 0.

9. T p The value ranges from 0.75 to 0.

85. The similarity between the voiceprint feature vector and the pre-stored voiceprint feature vector is calculated as follows: ; The similarity between the pupil feature vector and the pre-stored pupil feature vector is calculated as follows: ； The glasses and the smartwatch establish a connection and transmit data via Bluetooth high-speed communication protocol. The transmitted data includes pupil image data collected by the glasses and to be sent to the smartwatch, and voiceprint data collected by the smartwatch and to be used for comparison. The bit error rate of the Bluetooth high-speed communication protocol when transmitting the pupil image data and voiceprint data is no greater than 10. -5 The data transmission rate of the Bluetooth high-speed communication protocol R Satisfy the following formula: ； Where B is the channel bandwidth and S / N is the signal-to-noise ratio; The data transmission rate R The value must ensure that the pupil image data and voiceprint data can be transmitted within the time limit required by the identity verification process; During the authentication process, the collaborative allocation of computing power between the glasses and the smartwatch satisfies the following relationship: The smartwatch undertakes the core computing tasks of voiceprint feature extraction, dual biometric feature similarity calculation, and blockchain transaction construction and uploading; The glasses bear the computational load of pupil image acquisition and preprocessing; Total effective computing power of the system C total The value used to complete the entire authentication process is determined by the following formula: ； in, C g For computing power of glasses, C w The computing power of the smartwatch is represented by α, which is the computing power utilization coefficient, ranging from 0.8 to 1.

0. The computing power utilization coefficient α is dynamically adjusted according to the real-time requirements of the identity verification task and the current network status to ensure the smooth execution of the identity verification process.

2. The authentication method based on the collaborative operation of glasses and smartwatches as described in claim 1, characterized in that, The voiceprint feature vector is extracted using Mel frequency cepstral coefficients, and the pupil feature vector is obtained through an image feature extraction algorithm.

3. The authentication method based on the collaborative operation of glasses and smartwatches as described in claim 1, characterized in that, The voiceprint feature vector is extracted using Mel-frequency cepstral coefficients, specifically including the following steps: The smartwatch's microphone collects the user's voice signal and pre-emphasizes the signal to boost high-frequency components. The pre-emphasized audio signal is divided into frames and windowed, with a frame length of 20 to 40 milliseconds and a frame shift of 10 to 20 milliseconds. Perform a Fast Fourier Transform on each frame of signal to convert it to the frequency domain and obtain a linear spectrum; The linear spectrum is passed through a set of Mel-scale filters that approximate a linear distribution below 1000 Hz and a logarithmic distribution above 1000 Hz to simulate the nonlinear perception of sound by the human ear. Taking the logarithm of the filter bank's output yields the log-Mel spectrum; Perform a discrete cosine transform on the logarithmic Mel spectrum and take the first 12 to 16 coefficients as the static Mel frequency cepstral coefficients; The first-order and second-order difference coefficients of the static Mel frequency cepstral coefficients are calculated and together constitute the final voiceprint feature vector.

4. The authentication method as described in claim 1, characterized in that, The pupil feature vector is obtained through an image feature extraction algorithm, specifically including the following steps: The glasses' camera captures infrared images of the user's eyes, and the images are preprocessed by grayscale conversion and histogram equalization. The pupil region is located using a circular detection algorithm based on Hough transform. The parameter accumulator of the circular detection algorithm has a resolution of 1 to 2 pixels, and the minimum center distance is 1.5 to 2 times the expected radius of the pupil. The effective iris texture area is segmented from the located pupil area, excluding the parts obscured by the eyelids and eyelashes; The segmented annular iris region is geometrically normalized using a rubber membrane model and mapped to a rectangular iris image of fixed size, where the horizontal axis of the rectangle corresponds to the circumferential direction of the iris and the vertical axis corresponds to the radial direction of the iris. The normalized iris region is divided into several sector regions, each with a radial resolution of 8 to 16 pixels and an angular resolution of 5 to 10 degrees. A two-dimensional Gabor filter bank is applied to each sector region for texture feature extraction. The wavelength range of the Gabor filter bank is 3 to 15 pixels, the number of directions is 4 to 8, and the phase shift is 0 and π / 2. Calculate the mean and standard deviation of the output response of each filter, which together constitute the local feature vector of the sector region; The local feature vectors of all fan-shaped regions are concatenated, and principal component analysis is performed on the concatenated high-dimensional feature vector to reduce its dimensionality. Principal components with a cumulative contribution rate of 85% to 95% are retained to form the final pupil feature vector.

5. The authentication method based on the collaborative operation of glasses and smartwatches as described in claim 1, characterized in that, The smartwatch verifies the blockchain network to which the signal and verification data are uploaded, forming a consortium blockchain containing both the smartwatch and the glasses as consensus nodes; the blockchain network employs a practical Byzantine fault-tolerant algorithm to perform consensus verification on the identity verification transaction; the total number of nodes N and the number of malicious nodes f in the blockchain network satisfy the following relationship: The smartwatch and the glasses, as two light nodes in the blockchain network, participate in the voting and verification process for the validity of identity verification transactions.

6. The authentication method based on the collaborative operation of glasses and a smartwatch as described in claim 1, characterized in that, The smartwatch is also used to receive and store authentication-related message data from the glasses. The authentication-related message data includes authentication trigger instructions, authentication process logs, or pupil image feature vectors for this authentication. If the amount of the authentication-related message data is greater than the remaining storage space of the smartwatch, the authentication-related message data is compressed using Huffman coding before storage.

7. The authentication method as described in claim 6, characterized in that, The steps for Huffman coding compression of authentication-related message data include: Perform probability statistics on the symbols in the message data; Construct a Huffman tree based on the probability of symbol occurrence; Each symbol is assigned a variable-length code, where symbols with high occurrence probability are assigned short codes and symbols with low occurrence probability are assigned long codes; the average code length L of the Huffman code satisfies: ； in, p i The probability of the symbol appearing. l i For the corresponding encoding length; The average coding length L This value is used to evaluate the compression effect of the authentication-related message data; the smaller the value, the higher the compression efficiency.

8. A smart wearable system based on the collaborative operation of glasses and a smartwatch, used to perform the authentication method as described in any one of claims 1 to 7, characterized in that, The system includes smart glasses and a smartwatch; The smart glasses include: A camera module configured to capture images of the user's eyes; The first processing module is configured to preprocess the eye image and extract the pupil feature vector; A first Bluetooth communication module is configured to establish a connection with the smartwatch and transmit data. The smartwatch includes: Microphone module, configured to capture the user's voice signal; The second processing module is configured to process the sound signal and extract the voiceprint feature vector, and calculate the similarity between the voiceprint feature vector and the pre-stored voiceprint feature vector, as well as the similarity between the pupil feature vector and the pre-stored pupil feature vector. The second Bluetooth communication module is configured to establish a connection with the smart glasses and transmit data. A blockchain communication module, configured to upload verification pass signals and verification data to the blockchain network; The storage module is configured to store pre-stored voiceprint feature vectors, pre-stored pupil feature vectors, and message data from the smart glasses. The first Bluetooth communication module and the second Bluetooth communication module establish a connection via the Bluetooth High Speed ​​Communication Protocol, wherein the bit error rate of the Bluetooth High Speed ​​Communication Protocol during data transmission is no greater than 10. -5 ; The smart glasses and the smartwatch work together to complete the identity verification process.

Images