Method for retrieving password through intelligent voiceprint
Through intelligent voiceprint recognition technology, the voice control module and voiceprint comparison module are used to perform voiceprint comparison, which solves the problems of dependence and lack of convenience of existing password retrieval methods, and achieves higher security and convenience. .
Patent Information
- Application Number
- CN202411430308.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-14
- Publication Date
- 2025-01-21
AI Technical Summary
The existing password recovery methods have the disadvantages of dependency, insufficient convenience, accessibility issues, abuse risks, privacy leakage risks and vulnerability.
Through intelligent sound pattern recognition technology, the sound control module and sound pattern comparison module generate and compare the new old sound pattern to achieve user identity verification, thereby retrieve the password.
It improves the security and convenience of password retrieval, reduces the risk of sensitive information leakage, and avoids the dependence and vulnerability problems in traditional methods.
Smart Images
Figure CN119341799A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for retrieving a password through intelligent voiceprint recognition. Background Art
[0002] The account and password system is a widely used authentication mechanism for controlling user access to application systems or services. By logging into the system with an account and password, it can be ensured that only authorized users can access sensitive information and resources. The account and password system and login function play an important role in modern digital services, providing users and enterprises with a safe, personalized and efficient service experience. With the advent of the digital age, personal information security has become particularly important. As the first line of defense for protecting personal information, the importance of passwords is self-evident. However, the characteristics of human memory make forgetting passwords a common phenomenon. In order to solve this problem, the password retrieval function came into being, which allows users to regain access to their accounts when they forget their passwords.
[0003] The traditional methods of retrieving passwords mainly include:
[0004] 1. [Security Questions]: A series of questions and answers set by the user when registering an account. When you forget your password, you can verify your identity by answering these questions;
[0005] 2. [Email Verification]: The email address provided by the user during registration will receive a link to reset the password;
[0006] 3. [SMS Verification Code]: The user's bound mobile phone number receives a text message containing a verification code, and enters the verification code to reset the password;
[0007] 4. [Alternative email address or mobile phone number]: The alternative contact information provided by the user during registration. The email address or mobile phone number is used to receive the link or verification code to reset the password.
[0008] Although these methods solve the problem of forgetting passwords to a certain extent, they still have the following disadvantages:
[0009] 1) Security questions: Security questions can be guessed by malicious users or the answers can be obtained through social engineering methods, thereby bypassing identity authentication;
[0010] 2) Dependency: Both email verification and SMS verification codes rely on the user's email and mobile phone. If there is a problem with these services, the user may not receive the link or verification code to reset the password;
[0011] 3) Convenience: To retrieve your password by email or SMS, you need to log in with a valid email and mobile phone number to check the verification code, which is not convenient enough;
[0012] 4) Accessibility issues: The backup email address or mobile phone number may become invalid because the user has not used it for a long time, or the user may change the contact information but not update it, resulting in the inability to use the backup contact information;
[0013] 5) Abuse: If someone maliciously uses the password retrieval function to try to guess or crack other people's passwords, it may threaten the user's account security;
[0014] 6) Privacy issues: In order to retrieve the password, the user may need to provide some personal information, which may increase the risk of personal information leakage;
[0015] 7) Vulnerability: Email and SMS services may be subject to security threats such as phishing attacks, malware or man-in-the-middle attacks, affecting the security of user accounts. Summary of the invention
[0016] The purpose of the present invention is to generate and compare new and old voiceprints through a voice control module and a voiceprint comparison module to help users retrieve passwords and thus regain access rights to application systems or services.
[0017] In order to achieve the above object, the technical solution of the present invention provides a method for retrieving a password through an intelligent voiceprint, which is characterized by comprising the following steps:
[0018] Step 1: Build a database for storing voiceprint features of different users. After obtaining all types of voiceprint features of the current user through the sound control module, store the obtained voiceprint features and the recording time of the obtained voiceprint features in the database;
[0019] Step 2: If the current user needs to retrieve the password, the voiceprint duration is calculated based on the current time and the entry time stored in the database: if the voiceprint duration is greater than the preset voiceprint validity threshold, the voiceprint feature in the database is updated and then step 3 is entered; otherwise, step 3 is entered directly;
[0020] Step 3: After obtaining any N types of voiceprint features from the database, the voice control module randomly generates M sample sentences containing all N types of voiceprint features, where N≥2 and M≥1.
[0021] Step 4: After the current user gives a voice signal according to the text corresponding to the sample sentence, the real-time voiceprint feature of the voice signal is obtained through the sound control module;
[0022] Step 5: Compare the real-time voiceprint features with the voiceprint features obtained from the database and calculate the voiceprint similarity: if the voiceprint similarity is not less than the preset similarity threshold, the current user is allowed to retrieve the password; otherwise, the current user is prohibited from retrieving the password.
[0023] Preferably, in step 1, the database is located on the server side.
[0024] Preferably, in step 2, the N types of voiceprint features of the current user obtained by the sound control module from the database are encrypted data.
[0025] Preferably, the voiceprint feature is obtained through the sound control module using the following method:
[0026] Step 101: The sound control module obtains a voice signal input by a user;
[0027] Step 102: The sound control module determines whether the noise value of the voice signal exceeds a preset noise threshold. If so, the process returns to step 101 and requires the user to re-enter the voice signal. If not, the process proceeds to the next step.
[0028] Step 103: After preprocessing the speech signal, the speech signal is divided into overlapping frames, and a window function is applied to each frame to perform a windowing operation;
[0029] Step 104: converting the windowed speech signal into a frequency domain signal by fast Fourier transform;
[0030] Step 105, extracting features from the frequency domain signal, including:
[0031] Through Mel filter bank and logarithmic compression, the result is then transformed through discrete cosine to obtain Mel frequency cepstrum coefficients, and then the linear prediction coefficients of the audio signal are estimated through autocorrelation analysis or least squares method;
[0032] Step 106: Calculate the first-order and second-order differential features of the features obtained in the previous step, normalize the differential features, perform feature vectorization, and use the obtained feature vector as the voiceprint feature.
[0033] Preferably, in step 103, there is a 50% overlap between frames.
[0034] Preferably, in step 5, the voiceprint similarity is Euclidean distance or cosine similarity.
[0035] Compared with the prior art solutions, the invention is innovative in that:
[0036] 1. The sound control module in the present invention follows a series of precise sound processing steps to input and output test samples. The present invention accurately captures the intrinsic characteristics of the sound through preprocessing, frame windowing, fast Fourier transform (FFT), feature extraction, dimensionality reduction, dynamic feature calculation and normalization, and converts it into feature vectors. These feature vectors are safely stored to build a unique voiceprint feature library for each user.
[0037] 2. The present invention discloses a dynamic update mechanism for voiceprints, which can not only update voiceprint data regularly, but also update it during critical periods when the voice may change significantly. The dynamic update mechanism described in the present invention ensures the real-time and accuracy of voiceprint data, while improving the security and reliability of the system.
[0038] 3. By using voice features, the present invention can randomly generate test samples and enable the password retrieval function by using voiceprint storage and comparison. This method reduces the predictability of samples and enhances security. Voiceprint data is invisible to users, ensuring privacy protection.
[0039] Based on the above innovations, the technical solution disclosed in the present invention has the following advantages compared with the existing technical solution:
[0040] Beneficial effects:
[0041] 1. By using intelligent voiceprint technology to retrieve passwords, users can enjoy a fast and easy experience. Account access rights can be quickly restored simply through voice recognition and comparison. The whole process is efficient and intuitive.
[0042] 2. Compared with traditional password retrieval methods, the intelligent voiceprint password retrieval method provides a higher level of security. It reduces the reliance on fragile static passwords, avoids potential delay issues, and greatly reduces the risk of sensitive information leakage, ensuring the privacy and security of user data. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] Figure 1 is a flow chart of the present invention;
[0044] Figure 2 This is a flowchart for comparing new and old voiceprints. DETAILED DESCRIPTION
[0045] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.
[0046] The method for retrieving a password through an intelligent voiceprint disclosed in an embodiment of the present invention is based on the following concepts:
[0047] 1. Voiceprint: refers to the collection of unique sound features that a person exhibits when speaking. These features can be used for individual identification, that is, to determine the identity of the speaker by comparing sound samples. Voiceprint recognition is a biometric technology that uses a person's voice to verify personal identity, similar to fingerprint or facial recognition technology; voiceprints mainly include pitch, sound intensity, timbre, formant, speaking speed, intonation, pronunciation method and other components.
[0048] 2. Voice control module: This module is a voice input and output control system. This module can output sample sentences that can be recorded by users according to some types of voice, so that users can record their own voice samples. This module is also responsible for saving the voice recorded by users through samples, extracting voiceprint features, and saving the voiceprint features into a feature vector in the database.
[0049] 3. Mel-frequency cepstral coefficients (MFCCs): It is a speech feature extraction method that converts audio signals into a series of cepstral coefficients by simulating the human ear's perception of sound. These coefficients capture key information of speech signals, such as pitch and tone, and are widely used in speech recognition and voiceprint recognition.
[0050] 4. Linear Predictive Coding (LPC): It is a speech signal processing technology based on the speech generation model. It estimates the linear prediction coefficient of the speech signal by analyzing the autocorrelation of the signal, thereby capturing the dynamic and spectral characteristics of the speech signal. It is often used in speech synthesis and speech coding.
[0051] 5. FFT: The abbreviation of "Fast Fourier Transform". It is a mathematical algorithm used to efficiently calculate the Fourier Transform. The Fourier transform is a method of converting a signal or function from the time domain (time) to the frequency domain (frequency). The main advantage of the fast Fourier transform is its computational efficiency. The traditional Fourier transform has a high computational complexity, while the fast Fourier transform reduces the computational time from On2n2 to Onlognnlogn by using a divide-and-conquer strategy and butterfly operations, where n is the length of the input data. This makes FFT widely used in digital signal processing, image processing, audio processing and other fields.
[0052] Based on the above concepts, a method for retrieving passwords through intelligent voiceprints disclosed in an embodiment of the present invention generates and compares new and old voiceprints through a sound control module and a voiceprint comparison module, and determines whether the user is unlocked and whether the password can be successfully retrieved based on the result of the voice comparison.
[0053] like Figure 1As shown, when the user forgets the password of the system or application and needs to retrieve it, the intelligent voiceprint password retrieval method will be triggered. At this time, the sound control module intervenes and randomly generates three sample sentences containing several voiceprint features, wherein the voiceprint features include but are not limited to basic tone, speaking speed, voice rhythm, pitch, timbre, intonation, sound intensity, rhythm, formant, continuity of sound, accuracy of pronunciation, continuity of sound, emotionality of sound, etc. Take a sample sentence as an example, it contains four voiceprint features: (A) pitch, (B) timbre, (C) emotionality of sound and D (formant). These voiceprint features are randomly generated, and the user does not know which voiceprint features the sample sentence contains. The voiceprint feature only exists on the server side, and the encryption process has been done when the voiceprint feature is sent down (the encryption method is AES, RSA, ECC, etc., and the encryption method is not limited in the present invention). When the user needs to retrieve the password, the voiceprint features recorded by the server are used to generate a sample sentence containing the voiceprint features of the first entry. Due to the randomness and encryption of the voiceprint features contained in the sample sentence, the security and privacy of the password retrieval are improved to a certain extent, and some disadvantages of retrieving passwords through traditional methods such as email, mobile phone verification codes, and security questions (such as dependence, vulnerability to attack, abuse, etc.) are effectively avoided. The sample length of each sample sentence is not less than the preset threshold. The user enters the corresponding sentence according to the text of the sample sentence generated by the sound control module. After the entry is completed, the sound control module extracts real-time voiceprint features for the collected real-time sentences.
[0054] like Figure 2 As shown: First, the sound control module will determine whether the sound noise value entered by the user exceeds the preset threshold of the voiceprint required to be extracted by the system. If it is higher than the defined threshold, the user will be asked to re-enter the voice until the sound noise value is less than or equal to the threshold that the system can handle before proceeding to the next step.
[0055] Next, the sound control module will preprocess the audio signal obtained in real time. Preprocessing first performs denoising, using filtering algorithms to remove background noise, such as Gaussian noise, white noise, etc., to improve the accuracy of feature extraction. After preprocessing, framing and windowing operations are performed to divide the audio signal into overlapping frames, usually 25ms to 30ms per frame, with 50% overlap between frames, and apply a window function, such as a Hamming window, to each frame to reduce frame boundary effects.
[0056] In the next step, the windowed audio signal is subjected to FFT (Fast Fourier Transform) to convert the time domain signal into a frequency domain signal. After this step is completed, voiceprint feature extraction can begin. The Mel-frequency cepstral coefficients are obtained through the Mel filter bank and logarithmic compression, and then the result is subjected to discrete cosine transform (DCT). The linear prediction coefficients of the audio signal are estimated through autocorrelation analysis or least squares method to further obtain the reflection coefficient or other related features. If it is for Mel-frequency cepstral coefficients (MFCCs), dimensionality reduction operations are required to retain only the most important coefficients. For linear prediction coefficients (LPC), specific subsets need to be extracted, such as prediction error energy.
[0057] The next step is to calculate dynamic features, and calculate the first-order and second-order differential features of the features obtained in the previous step. These differential features describe the dynamic changes of the speech signal. After the differential features are normalized, the feature vectors are standardized to have a uniform scale, which helps to improve the generalization ability of the model. The last step is to vectorize the features and save the obtained feature vectors to a file. The format can be text (such as csv) or binary (such as a binary file). After the feature vectorization is completed, it can be saved to the database of the sound control module.
[0058] When the user needs to retrieve the password, the voice control module is called again, and based on the voiceprint features recorded by the user, three sentences with voiceprint features are randomly generated again (the voiceprint features are consistent with the features of the previous samples) for the user to re-enter. The entry process is the same as Figure 2 . After the input is completed, the newly generated voiceprint features are compared with the voiceprint features saved in the database. The specific comparison method is voiceprint similarity calculation, including but not limited to: Euclidean distance (calculating the Euclidean distance between two sets of feature vectors, the smaller the distance, the higher the similarity), cosine similarity (calculating the cosine similarity of feature vectors, that is, the cosine value of the angle between them, the closer the cosine value is to 1, the higher the similarity). When the voiceprint detection similarity reaches the set threshold, the user can perform the password recovery operation and directly set a new password to regain the use permission of the system or application. If the voiceprint comparison is lower than the threshold set by the system, the password cannot be recovered.
[0059] For the preservation process of voiceprint features, that is, after the user records the voiceprint features for the first time and does not trigger the voiceprint password retrieval behavior, the feature vector quantization of the voiceprint has a certain validity period. The regular validity period is 3 years, and the special validity period may require re-recording every year, such as during adolescence (girls are usually between 12 and 16 years old; boys are usually between 10 and 14 years old). Adolescence is one of the periods when voiceprint changes most significantly. During this period, due to hormonal changes, the vocal cords of adolescents will undergo rapid growth and development, resulting in significant changes in the pitch, timbre and volume of the voice. Boys' voices tend to become deeper, while girls' voices may become higher or thinner. When the re-entry validity threshold is reached, the system will give a re-entry prompt, and the user can choose whether to re-enter. If you do not agree to re-enter, you will be informed of the risks associated with not entering the sample, such as: the password retrieval function may fail. Due to the validity period reactivation function of voiceprint extraction, the situation where the password cannot be retrieved when retrieving the password through mobile phone number, email, etc. due to the deactivation of the mobile phone number or email address is avoided. The availability problem of mobile phone number and email address is avoided, which greatly improves the convenience of password retrieval. At the same time, using voiceprint method to retrieve password is interesting, which improves the user experience.
Claims
1. A method for retrieving a password through an intelligent voiceprint, characterized in that: The following steps are involved: Step 1: Build a database for storing voiceprint features of different users. After obtaining all types of voiceprint features of the current user through the sound control module, store the obtained voiceprint features and the recording time of the obtained voiceprint features in the database; Step 2: If the current user needs to retrieve the password, the voiceprint duration is calculated based on the current time and the entry time stored in the database: if the voiceprint duration is greater than the preset voiceprint validity threshold, the voiceprint feature in the database is updated and then step 3 is entered; otherwise, step 3 is entered directly; Step 3: After obtaining any N types of voiceprint features from the database, the voice control module randomly generates M sample sentences containing all N types of voiceprint features, where N≥2 and M≥1. Step 4: After the current user gives a voice signal according to the text corresponding to the sample sentence, the real-time voiceprint feature of the voice signal is obtained through the sound control module; Step 5: Compare the real-time voiceprint features with the voiceprint features obtained from the database and calculate the voiceprint similarity: if the voiceprint similarity is not less than the preset similarity threshold, the current user is allowed to retrieve the password; otherwise, the current user is prohibited from retrieving the password.
2. A method for retrieving a password through an intelligent voiceprint as claimed in claim 1, characterized in that: In step 1, the database is located on the server side.
3. The method for retrieving a password through an intelligent voiceprint as claimed in claim 1, characterized in that: In step 2, the N types of voiceprint features of the current user obtained by the sound control module from the database are encrypted data.
4. The method for retrieving a password through an intelligent voiceprint as claimed in claim 1, characterized in that: The voiceprint features are obtained through the sound control module using the following method: Step 101: The sound control module obtains a voice signal input by a user; Step 102: The sound control module determines whether the noise value of the voice signal exceeds a preset noise threshold. If so, the module returns to step 101 and requires the user to re-enter the voice signal. If not, proceed to the next step; Step 103: After preprocessing the speech signal, the speech signal is divided into overlapping frames, and a window function is applied to each frame to perform a windowing operation; Step 104: converting the windowed speech signal into a frequency domain signal by fast Fourier transform; Step 105, extracting features from the frequency domain signal, including: Through Mel filter bank and logarithmic compression, the result is then transformed through discrete cosine to obtain Mel frequency cepstrum coefficients, and then the linear prediction coefficients of the audio signal are estimated through autocorrelation analysis or least squares method; Step 106: Calculate the first-order and second-order differential features of the features obtained in the previous step, normalize the differential features, perform feature vectorization, and use the obtained feature vector as the voiceprint feature.
5. The method for retrieving a password through an intelligent voiceprint as claimed in claim 4, characterized in that: In step 103, there is a 50% overlap between frames.
6. The method for retrieving a password through an intelligent voiceprint as claimed in claim 1, characterized in that: In step 5, the voiceprint similarity is Euclidean distance or cosine similarity.