CN114707132A - Brain wave encryption and decryption method and system based on emotional voice - Google Patents

Abstract

The invention discloses a brainwave encryption and decryption method based on emotional speech, which comprises the following steps: connecting a brainwave sensor to a user's brain, a central processing unit displays a piece of content on a display interface; The user's brainwave input is collected on the sensor; after the central processor performs voice recognition of the user's reading through the microphone, the recognized text is the displayed content, so as to confirm that the user is reading the content, and then perform the brainwave processing operation, otherwise it will not be performed. Brain wave processing operation; the present invention distinguishes different users based on the difference of the brain wave signals of the emotional speech emitted by the user when reading, realizes the distinction of different users based on different biological characteristics of different users, and realizes user identity verification.

Description

Brainwave encryption and decryption method and system based on emotional speech

technical field

The invention relates to the technical field of brainwave signal encryption, in particular to a brainwave encryption and decryption method and system based on emotional speech.

Background technique

In recent years, related authentication schemes based on biometrics have been increasing. In the early days, external biometrics were used as the basis, but the risk of being counterfeited was higher than that of internal biometrics. Therefore, the direction of most solutions has shifted from external biometrics to Internal biometrics. As a kind of internal biological characteristics, brain waves are difficult to obtain and unique, which can effectively avoid the problem of counterfeiting, and its continuity can be continuously verified by users. In particular, different people present different brainwaves when they read aloud. Currently, there is no encryption scheme using brainwaves based on emotional speech.

The existing patent proposes a "Brain-Computer Switching Technology-Based Combination Lock and Its Encryption and Decryption Method", the patent application number is "201410101482.X". The main idea is to use brainwaves to decrypt the password in the brainwaves, so as to perform encryption and decryption. The final encryption and decryption judgment is still the password, and different people can decrypt it as long as they know the password. In this way, brainwaves are the medium for obtaining passwords. The brainwaves of different people do not play a role in distinguishing, and it is impossible to distinguish the biological characteristics of different people. The prior art has the following disadvantages: it is impossible to distinguish between different people, and it can be decrypted as long as the password is known.

SUMMARY OF THE INVENTION

To this end, it is necessary to provide a brainwave encryption and decryption method and system based on emotional speech, to solve the problem that the existing technology cannot distinguish between different people and can be decrypted only by knowing the password.

In order to achieve the above object, the present invention provides a brainwave encryption and decryption method based on emotional speech, which is used in a brainwave encryption and decryption system. The method includes the following steps: connecting the brain wave sensor to the user's brain, the central processing unit displays a piece of content on the display interface; the user reads the content aloud, and the central processing unit collects the user's brain wave input from the brain wave sensor ; After the central processing unit performs the voice recognition of the user's reading through the microphone, the recognized text is the displayed content, so as to confirm that the user is reading the content, and then perform the brain wave processing operation, otherwise the brain wave processing operation is not performed; brain wave processing The operation includes taking the continuous brain wave data of the stage, taking the time period of each hyphenation, finding out all hyphenation in the content that matches the top ten occurrence frequencies and having more than two words, and recording each hyphenation. The time segment of the guided reading, and find out the brainwave data recorded in the same time segment, store the brainwave data as the verification data source, and perform the encryption lock operation at the same time, and the encryption lock operation is associated with the verification data source; when decrypting , enter the authentication stage; the central processing unit displays another piece of content on the display interface, and the central processing unit collects the user's brainwave input from the brainwave sensor; the central processing unit collects the user's reading voice through the microphone and performs speech recognition, and recognizes The following text is verified with the displayed other content. When the verification is consistent, the brainwave decryption processing operation is performed, otherwise, the brainwave decryption processing operation is not performed; For all hyphenated words with ten names and two or more words, record the time period of each hyphenation guide in another piece of content and find out the brainwave data recorded in the same time period; input the brainwave data into the classification verification model , compare with the verification data source, and output the comparison result; if the comparison result is the same user, the decryption succeeds, otherwise the decryption fails, and the decryption lock state is maintained.

Further, the construction of the classification verification model includes the following methods: performing a feature value extraction stage step, the feature value extraction stage is used to process the data, and the collected brainwave data is subjected to the 3-Gram method and positive and negative three standard deviations. After the outliers are removed, the training and test data required for cross-validation with a preset number of repetitions are generated respectively; in the steps of the classifier construction stage, according to the training and test data, the overall learning method Bagging is used to generate a plurality of trainers Set, use OC-SVM to independently train each training subset to get its own training model, use the same segmentation test data for testing, and use the majority decision method of the merging rule for the classification results to obtain the final classification and verification model results.

Further, the feature value extraction stage comprises the following steps: analyzing the specified Chinese articles, and then using the pre-set available hyphenation screening conditions, selecting the available hyphenation of this article, and making it Chinese. Decode and store it; collect the brainwave data of the user reading the Chinese article, match it with the word segmentation in the Chinese article through voice recognition, and obtain the start and end times corresponding to each word segmentation; The brainwave data of the time segment corresponding to the time is obtained at the start and end times; the brainwave data acquisition process of the preset number of repetitions is performed to obtain training and test data.

Further, if the brainwave sensor and the central processing unit are connected through Bluetooth, the obtaining of the brainwave data of the time section corresponding to the time according to the corresponding start and end times of each segment includes the steps of: according to the start and end times corresponding to each segment. After adding a delay time to the end time, the brainwave data of the time segment corresponding to the time are obtained.

Further, the preset number of repetitions is 5 times.

Further, the characteristic values of the brain wave data include concentration, relaxation, stress and fatigue.

At the same time, the present invention provides a brainwave encryption and decryption system based on emotional speech, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, any one of the embodiments of the present invention is implemented. steps of the method described.

The implementation of the technical solution of the present invention is different from the prior art. The above-mentioned technical solution distinguishes different users based on the difference in the brain wave signals of the emotional speech emitted by the user during reading, and realizes different users based on different biological characteristics of different users. User authentication.

Description of drawings

Fig. 1 is the system structure diagram described in the specific embodiment;

2 is a flow chart of a method for generating a classification verification model according to the specific embodiment;

Fig. 3 is the word segmentation analysis diagram of the experimental article described in the specific embodiment;

4 is a flow chart of the guide reading data processing program described in the specific embodiment;

FIG. 5 is a flowchart of an electroencephalogram data processing program according to a specific embodiment;

6 is a schematic diagram of brainwave data processing of each segmented word according to the specific embodiment;

Fig. 7 is the brain wave data diagram of the classification algorithm described in the specific embodiment;

8 is a five-fold cross-validation diagram of legal verification according to the specific embodiment;

Fig. 9 is the five-fold cross-validation diagram of the illegal verification described in the specific embodiment;

FIG. 10 is a flow chart of the construction stage of the classifier according to the specific embodiment.

Detailed ways

In order to describe the technical content, structural features, achieved objects and effects of the technical solution in detail, the following detailed description is given in conjunction with the specific embodiments and the accompanying drawings.

Referring to FIG. 1 to FIG. 10 , this embodiment provides a method and system for brainwave encryption and decryption based on emotional speech. First of all, the user's pre-input must be performed first: the brainwave sensor of the system is first connected to the user's brain, that is, the user's head is equipped with a brainwave collector, and then the system now displays a piece of content (Tang poetry, five-character quatrains, seven-character quatrains) on the display interface. and Buddhist scriptures, etc.), the user reads the content, and the system collects the user's input from the brainwave sensor. Speech recognition through the microphone can confirm that the user is indeed reading the content. When the system processes brainwave data, it takes continuous brainwave data, and the basis is the occurrence time segment of each word segmentation (or word segmentation). The purpose is to select the brainwave segment to be analyzed more accurately. Unlike previous studies that used all brainwave data. Find out all the hyphenated words that match the top ten frequency and have more than two words in a piece of content, record the time period of each hyphenation guide, and find out the brainwave data recorded in the same time period. The brainwave data is used as a verification data source, and encryption operations are performed at the same time.

During authentication, a piece of content is also displayed, and the system collects the user's input from the brainwave sensor. Find out all the hyphenated words that match the top ten frequency and have more than two words in a piece of content, record the time period of each hyphenation guide, and find out the brainwave data recorded in the same time period. Input into the classification verification model, compare with the verification data source, and output the comparison result. If the comparison is the same user, the decryption succeeds, otherwise the decryption fails.

The classification verification model generation process, as shown in Figure 2, is a flow chart of the classification verification model generation process method. It is mainly divided into two stages, namely the feature value extraction stage and the classifier construction stage. In the feature value extraction stage, it is mainly the part of data processing. First, the collected brainwave data is subjected to 3-Gram method and positive and negative three standard deviations After the outliers are removed, the training and test data required for five-fold cross-validation are generated respectively. In the classifier construction stage, the overall learning method Bagging is used, and the base classifier is OC-SVM, and then the majority decision in the merger rule is used to obtain the final classification result. The details of these two stages will be introduced below.

In the feature value extraction stage, data processing must be performed first, as shown in Figure 3. In the first step, word segmentation analysis is performed on the specified Chinese article, and then the pre-set available word segmentation filter conditions are used to select the article. The available word segmentation of the article, decode it into Chinese and store it in the system. Word segmentation can be manually marked and then stored or call the existing word segmentation interface (such as Baidu word segmentation API) for word segmentation. Since the present invention does not consider the relationship between the brainwaves generated by the reading guide and word segmentation, and from the perspective of machine learning, the higher the frequency of word segmentation, the more available data, so the analysis of word segmentation The condition is set as the number of occurrences of word segmentation in the article, that is, the focus of the present invention is to use the word segmentation part as the judgment basis, rather than the entire sentence.

Table 1 shows the top ten hyphenated words after analysis, and then according to the conditions of available hyphenation: the number of words is greater than or equal to 2 and the frequency of occurrence is in the top ten, after screening, six available hyphenation words are obtained, which are where, I don’t know, smile, Wanli , thousands of miles and today. The six segmented words will be decoded in Chinese and stored as corresponding Unicode code combinations, as shown in Table 2, which includes the number of times each segmented word appears in the experimental article.

Table 1. Frequency statistics of the most frequent two-character strings in "Full Tang Poems"

string frequency string frequency string frequency string frequency string frequency where 166 unmanned 881 Castle Peak 662 running water 550 sunset 498 I don't know 146 wind blows 834 juvenile 634 look back 544 better 497 Miles 145 melancholy 780 meet 629 Pitiful 539 go home 496 Thousands of miles 130 deceased 778 life 597 in this way 526 sunset 496 today 116 autumn wind 749 every year 593 gray hair 520 cannot 481 not see 115 long 740 lonely 592 Owner 517 leave 481 not possible 114 Acacia 733 gold 589 now 516 when 478 spring breeze 112 Chang'an 722 Son of Heaven 588 moonlight 515 at this time 477 Baiyun 110 day 697 people do not 587 from now on 509 Luoyang 476 must not 947 how 687 world 586 sun and moon 508 world 472 bright moon 896 ten years 678 what's the matter 579 pedestrian 507 herb 472 world 890 who 663 Egami 553 general 499 return 471

Table 2. Unicode code combinations of available word segmentation

When the subject performs the task of "Guide to read the specified article", enter the second step, as shown in Figure 4, collect the subject's reading guide and brainwave data for each round at the same time, and use the Unicode that can be segmented stored in the first step. Code combination, compare the corresponding Unicode code combination in the guide data, and find out the corresponding start and end time of each segmented word, so as to facilitate the subsequent comparison with the brainwave data and the acquisition of characteristic data, that is, to display the guide article materials, Acquire the voice information read by the user and the corresponding time, and convert the voice information into the Unicode code of the text to obtain the corresponding time of the Unicode code, thereby obtaining the brainwave data of the corresponding time period.

When the test subject guides reading and can use word segmentation, the present invention will record the time and UNICODE code of the guide reading. Table 3 is the guide reading data format when the test subject guides reading "Where", and the UNICODE code combination is {(4f55,8655} , after comparing it with the "where" UNICODE code combination {4f55,8655} stored in the first step, it was found that between the time 16:27:4:609 and 16:27:5:308, the subject read "Where" "Location" segmentation. Then record the corresponding start time (16:27:4) and end time (16:27:5) of the guide reading. Since the minimum unit of each brainwave data is seconds, record the guide reading time segment The smallest unit is also seconds.

Table 3. Guide data format, taking "Where" as an example

time UNICODE code 16:27:4:60 4f 16:27:4:72 55 16:27:5:10 86 16:27:5:30 55

The third step is shown in Figure 5, using the start and end times of each segment recorded in the second step to compare the brainwave data to obtain the required brainwave data segment, and these brainwave data reports contain 8 In addition to Delta, Theta, Low Alpha, High Alpha, Low Beta, High Beta, Low Gamma, and HighGamma, it also includes the EEG202 brainwave detector developed by BeneGear, which has the functions of Attention, Relaxation There are twelve eigenvalues in total: Meditation, Pressure and Fatigue. Among them, the EEG202 brain wave detector uses the EEG signal collected by a single electrode to transmit to the EEG202 chip, the chip will perform noise reduction processing, and then obtain the EEG202 through a self-opening algorithm to digitally measure the current mental state of the person. .

In the specific implementation process, the brainwave instrument communicates with the computer software through the Bluetooth connection, and there is a delay time between them. Therefore, the present invention takes the brainwave data for 3 seconds after the end time of the recording, as shown in FIG. 6 . The time point of the "where" record is 16:27:4 to 16:27:5, and the brainwave time period to be taken is 16:27:4 to 16:27:8 to ensure that the Corresponding brain wave segment.

The purpose of taking the next 3 seconds is to obtain the corresponding brain wave segment. However, due to the limitation of the experimental equipment, the main brain wave data time cannot be determined. Therefore, a classification algorithm is used to add the possible brain wave data together. If the data covers the main brain wave data time, as shown in Figure 7, the verification accuracy will be improved. The part of the eigenvalues has also gone through the classification algorithm, so when entering the classifier construction stage, there are a total of 36 eigenvalues in the brainwave data.

In the classifier construction stage, each subject will repeat the experiment five times. Each experiment will go through the data processing step of the feature value extraction stage. Each word segmentation will generate brain wave data numbered A to E. Generate the required training data and test data. Since the verification method adopts the five-fold cross-validation method, as shown in Figure 8, the training data and test data need to be generated according to the requirements of five-fold cross-validation. Then the remaining four data will be merged into training data, which is legal user authentication.

During the abnormal identity verification of illegal users, as shown in Figure 9, the A to E brainwave data generated by all subjects except subject No. 1 will be merged according to the same number, and then the data will be merged according to the same number. Subject No. 1 has the same number of test data, and randomly select the same number of data from the combined data as the illegal verification test data of subject No. 1.

After generating the required training data and test data, start the overall learning method, as shown in Figure 10, first use Bagging to generate multiple training subsets of the training data of each segment, and use OC-SVM to combine each training subgroup. The sets independently train their own training models, and then use the same test data for each word segmentation for testing, and use the majority decision method of the merging rule for the classification results to obtain the final classification and verification model results.

In addition to the comparison between the experimental situation and the presence or absence of word segmentation, there are many existing merging rules, which are mainly divided into three majority decision methods, simple average method and weighted average method. The present invention adopts the simplest majority decision method. Whether it is legal or not is determined according to the output value of the classification verification model, and the relevant standards have been explained in detail in the above-mentioned classifier construction stage. The other two merging rules are added here to compare which merging rules work better. For the simple average method, if the final average value is ≥ 50%, it means the classification is correct, and if the value is less than 50%, it means the classification is wrong.

If the weighted average method is used, it is mentioned that the setting of the weight is usually based on the positive ratio in each classification verification model, but the classification verification model used in this method is a single category, which cannot be determined according to the general setting method Therefore, the present invention considers the number of occurrences of each segmented word, and the number may have different degrees of influence on the classification result, which can be used as the basis for setting the weight value. Therefore, taking the ratio of the occurrences of each segment to the total number of occurrences as the weight value of each segment can achieve better results.

The present invention also provides a brainwave encryption and decryption system based on emotional speech, including a storage medium, wherein the storage medium stores a computer program, and the computer program implements the steps of the above method when the computer program is executed by the processor. The storage medium in this embodiment may be a storage medium set in an electronic device, and the electronic device can read the content of the storage medium and achieve the effects of the present invention. The storage medium can also be a separate storage medium, and the storage medium is connected to the electronic device, and the electronic device can read the content in the storage medium and implement the method steps of the present invention.

It should be noted that, although the above embodiments have been described herein, it does not limit the scope of the patent protection of the present invention. Therefore, based on the innovative concept of the present invention, changes and modifications to the embodiments described herein, or equivalent structures or equivalent process transformations made by using the contents of the description and drawings of the present invention, directly or indirectly apply the above technical solutions In other related technical fields, all are included within the scope of patent protection of the present invention.

Claims (7)

1. A brain wave encryption and decryption method based on emotional voice is characterized by being used for a brain wave encryption and decryption system, wherein the brain wave encryption and decryption system comprises a brain wave sensor, a central processing unit, a display interface and a microphone which are sequentially connected, and the method comprises the following steps:

connecting the brain wave sensor with the brain of a user, and displaying a section of content on a display interface by a central processing unit;

reading the content by the user, and collecting the brain wave input of the user from the brain wave sensor by the central processing unit;

after the central processing unit performs voice recognition read aloud by the user through the microphone, the recognized characters are displayed contents, so that the user is confirmed to read the contents, and then brain wave processing operation is performed, otherwise, brain wave processing operation is not performed;

the brain wave processing operation comprises the steps of taking brain wave data in continuous stages, finding out all broken words which are consistent with the first ten times of the occurrence frequency and have more than two words in a section of content according to the occurrence time section of each broken word, recording the time section of each broken word for guiding reading, finding out the brain wave data recorded in the same time section, storing the brain wave data as a verification data source, and simultaneously carrying out encryption locking operation which is related to the verification data source;

entering into an identity verification stage when decrypting; the central processing unit displays another section of content on the display interface, and the central processing unit collects brain wave input of the user from the brain wave sensor; the central processing unit collects the voice read by the user through the microphone and carries out voice recognition, verifies the recognized characters and the displayed other section of content, carries out brain wave decryption processing operation when the characters are verified to be consistent with the displayed other section of content, or does not carry out brain wave decryption processing operation;

the brain wave decryption processing operation comprises finding out all word breaks which are consistent with the first ten words of the occurrence frequency and have more than two words in the other section of content, recording the time section of reading guide of each word break in the other section of content and finding out the brain wave data recorded in the same time section; inputting brain wave data into a classification verification model, comparing the brain wave data with a verification data source, and outputting a comparison result;

if the comparison result is the same user, the decryption is successful, otherwise, the decryption is failed, and the decryption locking state is maintained.

2. The electroencephalogram encryption and decryption method based on emotional speech according to claim 1, wherein: the construction of the classification verification model comprises the following steps:

performing a characteristic value extraction stage step, wherein the characteristic value extraction stage is used for processing data, and respectively generating training and test data required by cross validation for preset times after performing 3-Gram method and outlier removal processing outside positive and negative standard deviations on the acquired brain wave data;

and a classifier construction stage step, namely generating a plurality of training subsets by using a whole learning method Bagging according to the training and testing data, independently training each training subset into a respective training model by using an OC-SVM, testing by using the same word-breaking testing data, and then using a majority decision method of a merging rule for a classification result to obtain a final classification verification model result.

3. The electroencephalogram encryption and decryption method based on emotional speech according to claim 1, wherein: the eigenvalue extraction stage comprises the steps of:

performing word segmentation analysis on a designated Chinese article, selecting the available word segmentation of the article by using a preset available word segmentation screening condition, and performing Chinese decoding and storage on the selected available word segmentation;

collecting brain wave data of the Chinese article read by the user, matching with word breaks in the Chinese article after voice recognition, and acquiring starting time and ending time corresponding to each word break;

acquiring brainwave data of a time section corresponding to time according to the starting time and the ending time corresponding to each word segmentation;

and performing a brain wave data acquisition process with preset repetition times to obtain training and testing data.

4. The electroencephalogram encryption and decryption method based on emotional speech according to claim 3, wherein: the brain wave sensor is connected with the central processing unit through Bluetooth, and the step of acquiring brain wave data of a time section corresponding to time according to the starting time and the ending time corresponding to each word-breaking comprises the following steps:

and acquiring brainwave data of a time section of corresponding time after adding a delay time to the start time and the end time corresponding to each word segmentation.

5. The electroencephalogram encryption and decryption method based on emotional speech according to claim 3, wherein: the preset repetition number is 5.

6. The electroencephalogram encryption and decryption method based on emotional speech according to claim 1, wherein: the characteristic values of the brain wave data include concentration, relaxation, stress and fatigue.

7. An electroencephalogram encryption and decryption system based on emotional speech is characterized in that: comprising a memory, a processor, said memory having stored thereon a computer program which, when being executed by the processor, carries out the steps of the method according to any one of claims 1 to 6.

Images