Virtual reality user identity authentication method based on pupil light reaction

Abstract

The invention belongs to the technical field of virtual reality terminal security access control, and discloses a system for performing identity authentication on a user accessing equipment in a virtual reality terminal. In the identity authentication process, eye movement data flow information is utilized to extract pupil light reaction related information of the user, and whether the identity of the access user is matched with the claimed account of the access user or not can be efficiently judged. According to the invention, a rapid and lightweight virtual reality user identity authentication mechanism can be realized, and the requirements of the industry are met. According to the virtual reality user identity authentication method based on pupil light reaction, the pupil reaction biological characteristics of the user under light stimulation of different wavelengths are fully utilized, and high accuracy is achieved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of virtual reality (VR) terminal security access control, and specifically relates to a user identity authentication method and system based on biometric recognition. The system performs identity authentication on an access user in a virtual reality terminal. By utilizing eye movement data stream information, pupil light reaction related information of the user is extracted, and whether the identity of the access user matches the claimed account can be efficiently determined. The present application can realize a fast and lightweight virtual reality user identity authentication mechanism, and meets the needs of the industry. BACKGROUND

[0002] Virtual reality (VR) technology has greatly changed the digital experience of users by virtue of the immersive virtual environment constructed by the technology, and thus has been widely welcomed and applied in an increasingly rich range of scenarios. According to statistics from eMarkerter and Ciklum, the number of VR users in the United States has risen to 77 million in 2024. With the continued warming of user interest, it is predicted that the global VR market size will exceed 520 billion US dollars by 2031.

[0003] Users enjoy immersive interactive experiences in VR applications through head-mounted displays (HMDs), handheld controllers and other devices. However, this process also harbors risks such as data leakage, and the security risks are particularly prominent in sensitive scenarios such as financial transactions, private meetings or military simulations. Once a VR device is accessed by an unauthorized user, a series of problems such as privacy leakage, virtual asset theft and business operation interruption may be caused, which poses a serious threat to personal life and social security. Therefore, it is crucial to design a secure VR device user identity authentication mechanism.

[0004] However, the mainstream identity authentication method currently widely used by VR devices, i.e., the authentication method based on password input, has obvious security risks and is vulnerable to side-channel attacks. Moreover, the process of inputting a password using a virtual keyboard is quite cumbersome and adds an additional memory burden to the user. In addition to the traditional password input method, biometric technology is gradually attracting people's attention. With the help of advanced tracking sensors, people can obtain rich biometric data, such as eye movement behavior data captured by infrared cameras. Studies in the fields of biology and medicine have confirmed that the eye movement characteristics of different individuals differ significantly.

[0005] However, implementing virtual reality user identity authentication based on eye movement biometrics faces the following challenges:

[0006] (1) How to select appropriate eye movement behaviors and specific tasks or visual stimuli to induce such eye movement behaviors so that the user authentication process is fast, convenient and safe, and can also reflect the differences between individuals? Based on the fact that users may have to exert a lot of effort when performing body movements, and exposed body movements are easily attacked by external observers.

[0007] (2) How to extract effective features from noisy eye-tracking behavior data to represent the user's identity information? At present, most research on eye-tracking biometrics is based on professional eye-tracking equipment, while the eye-tracking sensors built into VR devices are not as accurate and have a lower sampling rate than professional equipment, making the obtained eye-tracking behavior data less accurate and stable.

[0008] (3) How to implement a user authentication scheme in a real-world virtual reality terminal and demonstrate its effectiveness and robustness?

[0009] Therefore, designing a virtual reality user authentication system and method based on eye-tracking biometrics has great potential. Summary of the Invention

[0010] To address the three challenges mentioned above, this invention proposes a practical virtual reality user authentication method: by collecting eye-tracking biometric data under specific conditions, extracting user-identity-related feature information, and thereby verifying whether the user is a legitimate user. Authentication scenarios include, but are not limited to: collecting user eye-tracking biometric features during the user registration phase and verifying user identity during the user login phase.

[0011] First, a novel biometric feature is introduced: the pupillary light response characteristics induced by different wavelengths of light. Under stimulation from different wavelengths of light, the diameter of the human pupil exhibits varying degrees of dilation and contraction. Second, considering the insufficient accuracy of eye tracking in VR devices, this invention designs an effective data preprocessing and biometric feature extraction process. This biometric feature extraction process extracts features from three aspects: statistical, temporal, and frequency domains. It combines the differences in pupil size between the left and right eyes to enhance the feature set and achieves feature enhancement through methods such as sample reuse, thereby fully acquiring the effective information from the original data.

[0012] To achieve real-world VR end-user authentication, this invention utilizes Unity, the largest VR application development platform, to acquire abundant pupil diameter response data via an eye-tracking sensor interface. After data preprocessing and feature extraction, an effective classifier can be trained for user authentication.

[0013] The technical solution of the present invention is as follows:

[0014] On the one hand, this invention provides a virtual reality user authentication method based on pupillary light response, applied to virtual reality terminals, characterized in that the method includes:

[0015] S1. Using the eye tracker built into the virtual reality terminal, collect data on the change in pupil diameter of a specific user under different wavelength light stimulation to form a raw dataset;

[0016] S2. Preprocess the raw data collected in step S1, including blink removal, low-pass filtering and resampling, to obtain preprocessed pupil diameter data;

[0017] S3. Analyze the preprocessed pupil diameter data obtained in step S2, select the optimal wavelength combination for a specific user as the exclusive authentication light stimulation for that user, and construct a specific light stimulation range;

[0018] S4. Within the specific light stimulation range obtained in step S3, extract basic features from the preprocessed pupil diameter data obtained in step S2, including statistical features, temporal morphological features, frequency domain energy features, and enhancement features, and construct pupil diameter response features;

[0019] S5. Using the pupil diameter response features obtained in step S4, train a binary classifier for each user using the extreme random forest algorithm;

[0020] S6. When a user logs in, provide the corresponding light stimulation according to the declared account, collect real-time pupil data and extract features, and input them into the classifier to complete authentication.

[0021] Specifically:

[0022] S1. Data Acquisition: For users during the registration and login phases, relevant biometric information is collected. The virtual reality terminal device sequentially changes the wavelength of the light source in the virtual scene, while the built-in sensor acquires pupil diameter response data, resulting in a pupil diameter data stream. The data stream obtained during the registration phase is denoted as... The data stream obtained during the login phase is denoted as

[0023] S2. Data Preprocessing: Preprocessing the pupil diameter data stream or Interpolation was used to remove spontaneous blinking, followed by low-pass filtering and resampling to obtain the preprocessed pupil diameter data stream. or

[0024] S3. Illumination Stimulus Selection: Preprocessing the pupil diameter data stream obtained during the registration phase... For a specific user, the KS test is used to assess the difference between their sample and other user samples, and to assess the consistency among their own samples. Finally, N, with the largest difference between the difference and the consistency, is selected. w A specific wavelength will be used as the user's customized light stimulus. When the user claims to log in, only this N wavelength will be provided during phase S1. w Each wavelength is used as a light stimulus.

[0025] S4. Feature Extraction: Extracting features from the preprocessed pupil diameter data stream. or In the selected N w At each wavelength, statistical, temporal, and frequency domain features are extracted from the data streams of the left eye, right eye, and both eyes. Statistical and temporal domain features are extracted from the data streams showing differences between the left and right eyes, and feature enhancement is performed.

[0026] S5. Classifier Construction: During the training phase, using pupil response feature data extracted from the pupil diameter data stream obtained from the registration phase, an extreme random forest algorithm is used to construct a binary classifier for each user.

[0027] S6. Identity Authentication: The pupil response feature data extracted from the pupil diameter data stream obtained during the login phase is processed by the classifier trained in step S5 to output an identity label. This completes the user identity authentication process.

[0028] The user authentication scheme proposed in this invention utilizes pupillary light response. Users only need to receive visual stimulation from the light source in the virtual scene, without requiring any active task execution or memory. It is fully covered by virtual reality terminal devices, and eye images are not accessible to external observers. By randomly changing the order of light source wavelength changes, replay attacks can be effectively resisted. Currently, most commercial VR devices are equipped with eye-tracking functionality, supporting the widespread application of the scheme proposed in this invention.

[0029] On the other hand, the present invention also provides a virtual reality user authentication system based on pupillary light response, characterized in that it includes:

[0030] The data acquisition module collects data on changes in pupil diameter when a user receives light of different colors in a virtual reality environment.

[0031] The preprocessing module preprocesses the collected pupil data to improve data quality before sending it to the next module.

[0032] The light stimulation selection module selects the most suitable light source wavelength for the user as the light stimulation during authentication based on the pre-processed data on the change in the diameter of the pupils of both eyes.

[0033] The feature extraction module extracts biometric features from the preprocessed data on changes in the diameter of the pupils of both eyes for final identity authentication.

[0034] Finally, in the identity authentication module, after obtaining pupil change features from the feature extraction module, a classifier is generated for each user. The classifier is then used to determine the identity of the logged-in user.

[0035] The data acquisition module relies on a VR device with eye-tracking functionality. During the registration phase, multiple rounds of user sampling are required, while during the login phase, only one customized light stimulus needs to be played for the user to collect the corresponding pupil diameter data.

[0036] To ensure high data availability, the acquired data streams undergo preprocessing, including spontaneous blink removal, noise reduction, and resampling, to make each data segment smooth and of uniform length.

[0037] The light stimulation module needs to evaluate the differences between user samples and the consistency of user samples based on the KS test, and select the light source wavelength that best reflects the differences and consistency of pupil response for each user as the user's customized light stimulation.

[0038] This invention utilizes data streams from the left eye, right eye, and both pupils to extract statistical and time-frequency domain features, ensuring rich biometric information. Simultaneously, it leverages data streams from the differences between the left and right pupils to extract statistical and time-domain features, enhancing the feature set. To obtain sufficient training samples from the limited sampling rounds during the registration phase, a random concatenation method is used to select and combine corresponding samples from different wavelengths. Then, sample reuse based on wavelength sub-combinations further expands the dataset, thereby ensuring the performance of the classification model.

[0039] The classification model described above needs to output the authentication result from the features extracted by the feature extraction module. The classifier first uses training samples to construct a classification model using an extreme random forest, and then uses Bayesian optimization to adjust the model parameters. The extreme random forest algorithm is used because it differs from traditional random forests in its random attribute selection and segmentation criteria, making it more robust to noise and able to more accurately determine whether a user accessing the device is a legitimate user.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] (1) Novel eye-tracking authentication method. The user authentication scheme proposed in this invention is an identity authentication method that utilizes the reflective biometric feature of pupillary light response in VR scenarios. This scheme does not require users to actively perform additional tasks, nor does it require users to memorize any text or pattern passwords.

[0042] (2) Novel Eye-Motion Authentication Biometrics. This invention designs a novel eye-motion authentication biometric, utilizing the change in pupil diameter under different wavelength light sources as a biometric associated with the user's identity, enabling rapid and lightweight secure authentication. Furthermore, this invention proposes incorporating the difference between the left and right pupils as one of the feature sets, enhancing the robustness of the solution.

[0043] (3) In practical applications, the present invention can still guarantee the effectiveness and security of user authentication in multiple scenarios (e.g., replay attacks, outsider attacks). Specifically, the feature extraction scheme based on this method has advantages over existing methods. Attached Figure Description

[0044] Figure 1 Flowchart of the present invention

[0045] Figure 2 Virtual Reality Environment User Identity Authentication Diagram Detailed Implementation

[0046] The following is in conjunction with the appendix Figure 2 The technical solution provided by the present invention will be further described in detail below: This embodiment is implemented based on the technical solution of the present invention, and provides detailed implementation methods and specific operation processes. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

[0047] like Figure 2 As shown, this invention provides a virtual reality user authentication scheme based on pupillary light response, which can be deployed and used in virtual reality terminals. The detailed steps are as follows:

[0048] Data Acquisition: First, an eye-tracking dataset of the user is constructed. Only by acquiring rich biometric data can the accuracy of the final authentication result be guaranteed. To construct the eye-tracking dataset, during the registration phase, several rounds of N-mode data are played sequentially on the virtual terminal screen. k Different wavelengths of light (N) k =10), and simultaneously, eye-tracking sensors collect data on changes in pupil diameter in both eyes to construct a user eye-tracking dataset. Each element in the dataset represents a segment of pupil diameter records for the user at a specific wavelength. Where u is the user identity label (u∈[1,N) u ]), k is the wavelength label (k∈[1,N) k ]), where i is the record number (i∈[1,N) s ]), representing the i-th record at that wavelength.

[0049] Data preprocessing: Recording the acquired pupil diameters To eliminate abnormal pupil readings (less than 0) caused by spontaneous blinking, linear interpolation was used for blink removal. To eliminate background noise from ambient light and electromagnetic signals, Fourier transform revealed that the noise was concentrated in the high-frequency region; therefore, low-pass filtering was employed to obtain a smooth data signal. Finally, to ensure that each record has the same number of frames, interpolation was used for resampling, with a uniform sampling rate of f. s =60Hz. The preprocessed pupil diameter is recorded as... The length is L = f s ×T, where T is the sampling time for each record segment.

[0050] Illumination stimulus selection: Selecting the most suitable light source wavelength for each user requires using the KS test to assess inter-user sample variability and user-specific sample consistency. First, for user u... i Preprocessed pupil diameter records at a given wavelength k Take the average using the following formula:

[0051]

[0052] Let user u i with u j KS test value between samples For user u i At a given wavelength k, the average KS test value with all other users is As a standard for the difference between user samples.

[0053] For user u i For preprocessed pupil diameter records at a given wavelength k, the KS test value is calculated between every two records, and the average is taken to obtain the consistency standard for the user's own sample. The N values ​​with the largest difference between the discrepancy and the consistency are selected. w A wavelength is used as the light stimulus for user authentication, denoted as w. J For example, N w When = 4, w J =[w1,w2,w3,w4]. During feature extraction, only these N... w Features are extracted from the data stream corresponding to each wavelength and used for training.

[0054] Feature extraction: For pupil diameter records Extract its statistical, time-domain, and frequency-domain features. Statistical features F s Including mean, variance, median, skewness, and kurtosis. Time-domain characteristics F. tThe main description focuses on the morphological characteristics, including the normalized sequence C[n] during the period of most significant signal fluctuation, the area under the signal curve (AUC), peak and trough values, the contraction and expansion rates (the ratio of amplitude difference to time interval) between adjacent peaks and troughs, and polynomial simulation coefficients. To ensure uniform feature vector length, the number of peaks, troughs, contraction rates, and expansion rates must be consistent. For samples exceeding this number, they are truncated in descending or ascending order; for samples with insufficient numbers, existing data is reused.

[0055] To construct frequency domain features, a short-time Fourier transform (SFT) is required. The SFT sampling rate is determined to be f. fft =60Hz, number of sampling points is N fft =512, Hanning window size is 32, sliding step is 16. Perform short-time Fourier transform and truncate the cutoff frequency f. AF The spectrum at 5Hz is Spec. The frequency domain characteristics are obtained by averaging the time domain of Spec.

[0056] For the pupil diameter records of the user's left eye, right eye, and both eyes collected by the virtual reality terminal, after preprocessing, statistical and time-frequency domain features are extracted according to the above steps to obtain F. left ,F right With F cross The difference in pupil diameter between the left and right eyes after preprocessing is calculated. Statistical and temporal domain data are extracted based on the steps described above to obtain F. diff By concatenating the above four feature sequences, we obtain...

[0057] To further enhance the feature set, for each user u, the wavelength combination is w. J =[w1,w2,w3,w4], and obtain the feature sample set for each wavelength according to the above steps. The elements in the feature sample set are: k is w J Let i be a sample number representing the i-th sample in the sample set for a given wavelength. Randomly selecting a feature sample from the sample set for each wavelength allows us to combine them to obtain new feature samples. p, q, r, and s are the sample indices in the corresponding sample sets. Based on this, the number of feature samples can be expanded.

[0058] Obtain new feature samples F C After that, from N w By selecting 3 wavelengths from a given set of wavelengths, we can obtain 4 wavelength sub-combinations. The i-th sub-combination includes wavelengths w. J [i mod4],w J [i+1mod4] and wJ [i+2mod4], whose corresponding feature sample is F C (w J [imod4]), F C (w J [i+1mod4]) and F C (w J [i+2mod4]). The average and standard deviation of the feature samples in each sub-combination are calculated using the following formula:

[0059] F avg [i] = avg(F C (w J [imod4]), F C (w J [i+1mod4]),F C (w J [i

[0060] +2mod4]))

[0061] F std [i] = std(F C (w J [imod4]), F C (w J [i+1mod4]),F C (w J [i

[0062] +2mod4]))

[0063] The final pupil diameter response feature sample F was obtained. U =[F U [i]]=[F avg [i],F std [i]。

[0064] Classifier construction: During the training phase, based on the pupil diameter response feature samples F U For user u, a binary classifier Ψ is constructed using the Extreme Random Forest algorithm. u .

[0065] Identity Authentication: During the login phase, users first undergo a data collection process. Unlike the registration phase, at this time the VR device only plays the N selected during the light stimulus selection process once. w Each wavelength is used to collect the user's pupil diameter data for this instance. w i Indicates wavelength, J represents the user-claimed login account identity. Pupil diameter data. Following the data preprocessing and feature extraction procedures described above, unlike the registration phase, since there is only one pupil diameter record, the sample size is not expanded through random combination. The extracted features are then input into the constructed classifier Ψ. J It will output the final authentication result, that is, whether the user is the identity J they claim.

[0066] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A virtual reality user authentication method based on pupillary light response, characterized in that, The method includes the following steps: S1. Using the eye tracker built into the virtual reality terminal, collect data on the change in pupil diameter of a specific user under different wavelength light stimulation to form a raw dataset; S2. Preprocess the raw data collected in step S1, including blink removal, low-pass filtering and resampling, to obtain preprocessed pupil diameter data; S3. Analyze the preprocessed pupil diameter data obtained in step S2, select the optimal wavelength combination for a specific user as the exclusive authentication light stimulation for that user, and construct a specific light stimulation range; S4. Within the specific light stimulation range obtained in step S3, extract basic features from the preprocessed pupil diameter data obtained in step S2, including statistical features, temporal morphological features, frequency domain energy features, and enhancement features, and construct pupil diameter response features; S5. Using the pupil diameter response features obtained in step S4, train a binary classifier for each user using the extreme random forest algorithm; S6. When a user logs in, provide the corresponding light stimulation according to the declared account, collect real-time pupil data and extract features, and input them into the classifier to complete authentication.

2. The virtual reality user authentication method based on pupillary light response according to claim 1, characterized in that, Step S2, obtaining the preprocessed pupil diameter data, specifically includes: S2.1 For the user registration process, collect N data displayed on the virtual terminal screen when the user sees it. k Data on pupil diameter changes under various wavelengths of light were used to construct a user training dataset. Each element in this dataset represents a segment of pupil diameter records for that user at a specific wavelength. Where u is the user identity label (u∈[1,N) u ]), k is the wavelength label (k∈[1,N) k ]), where i is the record number (i∈[1,N) s ]), representing the i-th record at that wavelength; S2.2 For the user login process, collect the N displayed on the virtual terminal screen when the user sees it. w Data on pupil diameter changes under illumination of various wavelengths were used to construct a user test dataset. The elements in this dataset represent the user's pupil diameter at a specific wavelength (w). i The next section of pupil diameter record J represents the login account identity declared by the user. S2.3 Record the given pupil diameter or Perform blink removal to eliminate abnormal pupil readings (less than 0) caused by spontaneous blinking by the user; S2.4 Recording of pupil diameter or Perform low-pass filtering to eliminate background noise; S2.5 for pupil diameter recording or Perform data resampling with a uniform sampling rate of f s =60Hz, ensuring each record has the same number of frames; finally, based on Obtain preprocessed pupil diameter records The length is L = f s ×T, where T is the sampling time for each record segment; based on Obtain preprocessed pupil diameter records 3. The virtual reality user authentication method based on pupillary light response according to claim 1, characterized in that, Step S3, constructing a specific light stimulus, specifically includes: S3.1 Calculate the KS test value between user samples and other user samples: First, for user u i Preprocessed pupil diameter records at a given wavelength k Take the average using the following formula: Let user u i with u j KS test value between samples For user u i At a given wavelength k, the average KS test value with all other users is As a standard for differences in user samples; S3.2 Calculate the KS test value between user samples: For user u i For each pair of preprocessed pupil diameter records at a given wavelength k, the KS test value is calculated and averaged to obtain the consistency standard of the user's own samples. S3.3 For user u i Given preprocessed pupil diameter records at a given wavelength k, calculate the difference between the inter-user variability obtained in S3.1 and the user-specific consistency obtained in S3.

2. Select the N value with the largest difference. w A wavelength is used as the light stimulus for user authentication, denoted as w. J .

4. The virtual reality user authentication method based on pupillary light response according to claim 1, characterized in that, Step S4, constructing pupil diameter response features, specifically includes: S4.1 Constructing statistical features: For preprocessed pupil diameter records or (Hereinafter referred to as) ), calculate its mean, variance, median, skewness, and kurtosis, as statistical characteristics F. s . S4.2 Constructing Temporal Features: For the preprocessed pupil diameter records Based on its temporal morphology, the normalized sequence C[n] of its most significant fluctuation period is obtained; the area under the curve (AUC) is obtained; its peak and trough values ​​are obtained; the contraction and expansion rates (the ratio of amplitude difference to time interval) between the peak and trough values ​​are obtained; and the polynomial coefficients are obtained through polynomial simulation. These are used as the temporal feature F. t ; S4.3 Constructing Frequency Domain Features: For the preprocessed pupil diameter records Perform a short-time Fourier transform. The sampling rate of the short-time Fourier transform is f. fft =60Hz, number of sampling points is N fft =512, Hanning window size is 32, sliding step is 16, short-time Fourier transform is performed and cutoff frequency f is selected. AF The spectrum at 5Hz is Spec. The frequency domain characteristics are obtained by averaging the time domain of Spec. S4.4 Feature Enhancement Based on Left and Right Eye Differences: After preprocessing in step S2, the user's left and right eye pupil diameter records collected by the virtual reality terminal are processed according to steps S4.1 to S4.3 to obtain F. left ,F right With F cross Calculate the difference in pupil diameter between the user's left and right eyes after preprocessing, and obtain F based on the features extracted in steps S4.1 and S4.

2. diff By concatenating the above four feature sequences, we obtain S4.5 Feature enhancement based on wavelength sub-combination: For each user u, according to step S3, obtain the most suitable N for that user. w N wavelengths are used as light stimuli (default N) w =4), the feature extraction in step S4. only occurs in this N w The pupil diameter is recorded for each wavelength; for the training sample, assuming the wavelength combination is w J =[w1,w2,w3,w4], based on the feature sequence corresponding to wavelength k obtained in step S4.

4. From the set, a feature sample is randomly selected from each wavelength, and combined to obtain a new feature sample. For test samples, record the original pupil diameter. The features are obtained by combining the sample numbers in sequence. With feature sample F C For example, its corresponding N w Grouping wavelengths into sets of three yields four sub-combinations. The i-th sub-combination includes wavelengths w. J [i mod 4],w J [i+1 mod 4] and w J [i+2 mod 4], whose corresponding sample is F C (w J [i mod 4]), F C (w J [i+1 mod 4]) and F C (w J [i+2 mod 4]). The average and standard deviation of the samples in each sub-combination are taken to obtain the final pupil diameter response characteristic sample F. U =[F U [i]]=[F avg [i],F std [i]]. The same applies to the test samples.

5. The virtual reality user authentication method based on pupillary light response according to claim 1, characterized in that, include: The S5 identity estimation includes the following steps: S5.1 Classifier Construction Based on Pupil Diameter Response Features: Based on the pupil diameter response feature training samples obtained in step S4, for user u, positive samples are the pupil diameter response feature samples of user u, and negative samples are an equal number of samples randomly selected from the feature samples of other users. Using the extreme random forest algorithm, a binary classifier Ψ is constructed. u ; S5.2 Identity Authentication: Based on the pupil diameter reaction feature test sample obtained in step S4, use Ψ u Generate a probability value that classifies a user as a legitimate user. For each sub-group of samples, obtain a probability value, denoted as the score. i Since the sub-combinations originate from the same pupil diameter response during user authentication, the probability values ​​of all sub-combinations are accumulated, resulting in score = ∏score. i Samples with a score not lower than the threshold t are marked as valid, thus completing the user authentication process.

6. A virtual reality user authentication system based on pupillary light response, characterized in that, include: The data acquisition module is configured to collect the pupil diameter response of the target user when receiving light of different colors in a virtual reality environment through the built-in eye-tracking sensor interface, and save it as data on the change in pupil diameter of both eyes. A preprocessing module is configured to perform blink removal, noise reduction, and resampling on the pupil diameter change data of both eyes, and save it as preprocessed pupil diameter change data of both eyes; A light stimulation selection module is configured to select the most suitable light source wavelength as the light stimulation for authentication for the target user based on the preprocessed pupil diameter change data. A feature extraction module is configured to extract eye movement features from the preprocessed binocular pupil diameter change data within the specific illumination stimulus range, and construct pupil diameter change features that characterize user identity information. The identity authentication module uses a classifier based on pupil diameter variation characteristics to authenticate whether the logged-in user's identity is legitimate.

Images