CN113486648B - A method and system for de-information masking evaluation based on closed-concentration Chinese short sentence test - Google Patents

Abstract

The invention discloses a method and system for de-information masking evaluation based on closed-concentration Chinese short sentence testing. The first test sound is obtained based on the first masking sound information and the test phrase information, wherein the first masking sound information is energy masking. Or information masking; based on the first test tone, obtain the first test result; based on the first test result and the first test tone, obtain the first test adjustment tone; obtain the second test result based on the first test adjustment tone; based on the first test tone, the first test adjustment tone, and obtain the second test adjustment tone; obtain the third test result; filter the first, second, and third test results according to the first set signal-to-noise ratio; according to the filtered test results, Get the first evaluation result. It achieves the technical effect of separating energy masking and information masking by classifying and setting masking sounds and continuously dynamically adjusting noise, providing a more scientific and reliable auditory perception evaluation method for the field of information masking evaluation.

Description

A method and system for de-information masking evaluation based on closed-concentration Chinese short sentence test

Technical field

The invention relates to the technical field of data analysis, and in particular to a method and system for de-information masking evaluation based on closed-concentration Chinese short sentence testing.

Background technique

Children's speech perception ability is developing all the time, and the process of continuous improvement of speech recognition ability under noise with age is very complicated. The evaluation of the speech perception ability of children and children with cochlear implants has always been the focus of scholars. Since there is a big gap between children's speech recognition ability in noise and that of adults, we cannot simply use adult testing tools to evaluate children. In our country, many domestic scholars have developed a series of test materials for children's speech recognition ability in noise based on foreign research on children's speech recognition ability and based on the characteristics of Mandarin-speaking children. Research after introducing multi-person conversation noise and speech spectrum noise into the Mandarin Children's Word Adjacency Test designed and developed specifically for children shows that the masking effect of multi-person conversation noise is more obvious than that of speech spectrum noise. However, there is currently no quantitative or qualitative evaluation of the auditory masking effect of children using cochlear implants in the context of multi-person conversations, that is, multi-person competition.

Contents of the invention

The purpose of the present invention is to solve at least one of the above technical deficiencies by providing a de-information masking evaluation method and system based on closed-concentration short sentence testing to solve the main problem of evaluation of energy masking effect in the prior art. Information masking effect factors lack strict control, and there are technical problems that affect the evaluation results.

To this end, the first purpose of the present invention is to propose a de-information masking evaluation method based on closed-concentration Chinese short sentence testing. The method includes: obtaining test sentence information; obtaining a set of masking sound information types, and the masking sound information The category set includes steady-state background noise, dynamic noise, spectrogram noise, and multi-person speaking noise; according to the masking sound information category set, the first masking sound information is obtained; according to the first masking sound information, the test phrase Information, obtain a first test tone, wherein the first masking sound information in the first test tone is energy masking or information masking; based on the first test tone, obtain a first test result; according to the first test As a result, the first test tone is used to obtain a first test adjustment tone; based on the first test adjustment tone, a second test result is obtained; based on the first test tone and the first test adjustment tone, a second test result is obtained Test the adjustment tone; obtain a third test result based on the second test adjustment tone; filter the first test result, the second test result, and the third test result according to the first set signal-to-noise ratio; according to the filtered The test results obtained the first evaluation result.

The second object of the present invention is to provide a de-information masking evaluation system based on closed-concentration Chinese short sentence testing. The system includes:

A first obtaining unit, the first obtaining unit is used to obtain test sentence information;

a second obtaining unit, the second obtaining unit is used to obtain a masking sound information type set, the masking sound information type set includes steady background noise, dynamic noise, spectrogram noise, and multi-person speaking noise;

a third obtaining unit, the third obtaining unit is configured to obtain the first masking sound information according to the masking sound information type set;

A fourth obtaining unit, the fourth obtaining unit is configured to obtain a first test sound according to the first masking sound information and the test phrase information, wherein the first masking sound information in the first test sound For energy masking or information masking;

A fifth obtaining unit, the fifth obtaining unit is used to obtain a first test result based on the first test tone;

A sixth obtaining unit, the sixth obtaining unit is used to obtain a first test adjustment tone according to the first test result and the first test tone;

A seventh obtaining unit, the seventh obtaining unit is used to obtain a second test result based on the first test adjustment tone;

An eighth obtaining unit, the eighth obtaining unit is used to obtain a second test adjustment tone according to the first test tone and the first test adjustment tone;

A ninth obtaining unit, the ninth obtaining unit is used to obtain a third test result based on the second test adjustment tone;

A first screening unit, the first screening unit is used to screen the first test results, the second test results, and the third test results according to the first set signal-to-noise ratio;

The tenth obtaining unit is used to obtain the first evaluation result according to the screened test results.

The third object of the present invention is to provide a computer device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above method is implemented.

One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

Embodiments of the present invention provide a method and system for de-information masking evaluation based on closed-concentration Chinese phrase testing. By obtaining test phrase information, a masking sound information type set is obtained. The masking sound information type set includes steady-state background noise. , dynamic noise, spectrogram noise, multi-person speaking noise; according to the set of masking sound information types, the first masking sound information is obtained; according to the first masking sound information and the test phrase information, the first test sound is obtained , wherein the first masking sound information in the first test tone is energy masking or information masking; based on the first test tone, a first test result is obtained; according to the first test result, the first test tone, obtain a first test adjustment tone; obtain a second test result based on the first test adjustment tone; obtain a second test adjustment tone based on the first test tone and the first test adjustment tone; based on the The second test adjusts the tone to obtain the third test result; the first test result, the second test result, and the third test result are screened according to the first set signal-to-noise ratio; and the first test result is obtained according to the screened test results. evaluation result. By classifying the masking sounds and continuously adjusting the noise dynamically, it is possible to accurately evaluate the information masking effect of cochlear implants in children's multi-person competition context, and to separate energy masking from information masking to provide information masking. The assessment field provides the technical effect of a more scientific and reliable auditory perception assessment method, thus solving the existing technical problems that mainly focus on the assessment of energy masking effects. There is a lack of strict control over the factors that affect the information masking effect, and there are technical problems that affect the assessment results.

The above description is only an overview of the technical solutions of the present application. In order to have a clearer understanding of the technical means of the present application, they can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and understandable. , the specific implementation methods of the present application are specifically listed below.

Description of drawings

Figure 1 is a schematic flow chart of an information masking evaluation method based on closed-concentration Chinese short sentence testing according to an embodiment of the present application;

Figure 2 is a schematic structural diagram of an information masking evaluation system based on closed-concentration Chinese short sentence testing according to an embodiment of the present application;

FIG. 3 is a schematic structural diagram of an exemplary electronic device according to an embodiment of the present application.

Explanation of reference numerals: first obtaining unit 11, second obtaining unit 12, third obtaining unit 13, fourth obtaining unit 14, fifth obtaining unit 15, sixth obtaining unit 16, seventh obtaining unit 17, eighth obtaining unit Unit 18, ninth obtaining unit 19, first screening unit 20, tenth obtaining unit 21, bus 300, receiver 301, processor 302, transmitter 303, memory 304, bus interface 305.

Detailed ways

The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present invention and cannot be understood as limiting the present invention. On the contrary, embodiments of the invention include all changes, modifications and equivalents falling within the spirit and scope of the appended claims.

In the description of the present invention, it should be noted that, unless otherwise clearly stated and limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection. Ground connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood on a case-by-case basis. Furthermore, in the description of the present invention, unless otherwise specified, "plurality" means two or more.

Any process or method descriptions in flowcharts or otherwise described herein may be understood to represent modules, segments, or portions of code that include one or more executable instructions for implementing the specified logical functions or steps of the process. , and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed out of the order shown or discussed, including in a substantially simultaneous manner or in the reverse order, depending on the functionality involved, which shall It should be understood by those skilled in the art to which embodiments of the present invention belong.

The following describes a de-information masking evaluation method based on closed-concentration Chinese short sentence testing according to an embodiment of the present invention with reference to the accompanying drawings.

The technical solution of this application is: obtain test sentence information; obtain a masking sound information type set, which includes steady background noise, dynamic noise, spectrogram noise, and multi-person speaking noise; according to the masking sound information type set Information type set, obtain the first masking sound information; obtain the first test sound according to the first masking sound information and the test phrase information, wherein the first masking sound information in the first test sound is energy Masking or information masking; based on the first test tone, obtain a first test result; based on the first test result and the first test tone, obtain a first test adjustment tone; based on the first test adjustment tone, Obtain a second test result; obtain a second test adjustment tone based on the first test tone and the first test adjustment tone; obtain a third test result based on the second test adjustment tone; and obtain a third test result based on the first setting information. The first test result, the second test result, and the third test result are screened by the noise ratio; and the first evaluation result is obtained based on the screened test results. It solves the problem that the existing technology mainly evaluates the energy masking effect, lacks strict control over the factors that affect the information masking effect, and has technical problems that affect the evaluation results. By classifying the masking sounds and continuously adjusting the noise dynamically, it is possible to accurately evaluate the information masking effect of cochlear implants in children's multi-person competition context, and to separate energy masking from information masking to provide information masking. The assessment field provides the technical effect of a more scientific and reliable auditory perception assessment method.

Embodiment 1

As shown in Figure 1, the embodiment of the present application provides an information removal masking evaluation method based on closed-concentration Chinese short sentence testing. The method includes:

Step S100 obtains test sentence information;

Further, obtaining the test sentence information includes: obtaining a preset test word library, the preset test word library includes preset type word information; obtaining a preset sentence structure, the preset sentence structure and words The preset type of information corresponds to the preset type; words are extracted from the preset test word library according to the preset type; the extracted word information and the preset sentence structure are input into the sentence construction model to obtain the test Short message.

Specifically, different types of test words are set to build a preset test word library, which includes words with multiple parts of speech, such as nouns, verbs, adverbs, adjectives, etc., and the corresponding settings are made according to the test requirements and application environment. For example, according to the age, gender, or hobbies corresponding to the test, if it is used in a special field, the words can also be set based on the characteristics of the field. For example, for children aged 3-5, the number of word choices they face is limited and the content is limited. To suit their specific situation, for children aged 10-15, the test content can be complicated. In order to improve the versatility of use, it can also be set at a low level, which can expand the age range of the test. In addition, the number of preset test vocabulary libraries should have certain setting requirements to facilitate the management of the vocabulary library. However, in order to improve the efficiency of the test, the number of settings should not be too many. In the embodiment of this application, the preferred number is 50-100. Word types are divided into 5-10 categories, and 5-10 words are selected for each category. When setting test sentences, a fixed sentence structure is set according to the word type, and one word library is randomly selected from each category. The words are combined to form a complete sentence. According to the different category settings, the type and number of words selected are also different. The difficulty of the test sentences composed in this way will be different. In order to provide component testing To improve the efficiency and accuracy of short sentences, the embodiments of this application utilize artificial intelligence technology to construct a neural network model, and use multiple sets of word information and preset sentence structures as training data to train the neural network model to obtain the corresponding sentence construction model. , in this way, inputting the corresponding input information into the sentence construction model will automatically output the constructed sentence results, effectively providing computational efficiency and accuracy.

Step S200 obtains a masking sound information type set, which includes steady background noise, dynamic noise, spectrogram noise, and multi-person speaking noise;

Specifically, the embodiment of the present application evaluates multi-person competitive context de-information masking based on the closed-set Chinese sentence test. Since fixed words are used to construct test sentences, closed-set noise evaluation materials are selected, which can preliminarily eliminate the subject. Limitations caused by the age, cognition, education level, memory and other factors of the tester. By using different types of masking sounds, such as steady-state background noise, dynamic noise, spectral noise, multi-person speaking noise, etc., and by constructing masking effect evaluations in different noises, energy masking and information masking can be separated.

Step S300: Obtain first masking sound information according to the masking sound information type set;

Specifically, the corresponding settings are made according to the various types of noise in the masking sound information type, and the masking evaluation of different noises is carried out. When conducting the evaluation, the appropriate volume is first determined, and different noise evaluation contents are selected, including steady noise and dynamic noise. , male voice reverse, female voice reverse, English reverse, etc. The first masking sound information is a test item corresponding to one of the noise information selected according to the type of masking sound information. The user can set and select the test items. .

Step S400 obtains a first test sound according to the first masking sound information and the test phrase information, wherein the first masking sound information in the first test sound is energy masking or information masking;

Specifically, the target test sound is generated according to the test sentence information and then the first masking sound information corresponding to the test item is superimposed to generate the corresponding first test sound. Depending on the selected test item, the corresponding one can be energy masking or information masking. A kind of method that enables separate evaluation of energy masking and information masking.

Step S500 obtains a first test result based on the first test tone;

Specifically, the test user will give the result according to the first test tone he heard. The test user can choose to give the result himself according to the difficulty of the test, and compare the answer given with the test sentence information. , to obtain the first test result, the selection content can also be given on the display interface, and the test user selects the closest content according to what he heard, and compares the content according to the selected content and the test sentence information, and performs according to the correct number The proportion is calculated to determine the first test result.

Step S600: Obtain a first test adjustment tone according to the first test result and the first test tone;

Step S700 obtains a second test result based on the first test adjustment tone;

Specifically, the masking sound is dynamically changed with the test results, such as increasing the frequency, decreasing the frequency, increasing or decreasing the volume, etc. to adjust the masking sound and continue testing. Multiple test results can be selected for each test item. average to ensure the reliability of test results.

Step S800 obtains a second test adjustment tone based on the first test tone and the first test adjustment tone;

Step S900 obtains a third test result based on the second test adjustment tone;

Step S1000 filters the first test result, the second test result, and the third test result according to the first set signal-to-noise ratio;

Step S1100 obtains a first evaluation result based on the screened test results.

Specifically, each test item is preferably tested 3-5 times to improve accuracy, and it can be more on the same day. This embodiment is not limited. Finally, the test results are selected to have a signal-to-noise ratio that satisfies a certain recognition rate as reference information. For example, the signal-to-noise ratio of the recognition rate is 50%, 70%, etc., and then the selected test results are weighted to obtain the final evaluation result. For example, the selected test results are the first test result, the second test result, and the third test result. The calculation formula for test results is (first test result + second test result + third test result)/3. By classifying the masking sounds and continuously adjusting the noise dynamically, it is possible to accurately evaluate the information masking effect of cochlear implants in children's multi-person competition context, and to separate energy masking from information masking to provide information masking. The evaluation field provides the technical effect of a more scientific and reliable auditory perception evaluation method, and solves the existing technical problems that mainly focus on the evaluation of energy masking effects. There is a lack of strict control over the factors that affect the information masking effect, and there are technical problems that affect the evaluation results.

Further, the method includes: obtaining the target keyword quantity requirement; filtering from the preset test word library according to the target keyword quantity requirement to obtain the first target keyword; and according to the masking sound information type set, Obtain an interference assessment content set; input the interference assessment content set and the first target keyword into a masking analysis model. The masking analysis model is obtained through training convergence through multiple sets of training data. Each of the multiple sets of training data Each set of training data includes interference evaluation content, target keywords and identification information indicating speech intelligibility; a first output result of the masking analysis model is obtained, and the first output result includes speech intelligibility information, based on the obtained All speech intelligibility information constructs a speech intelligibility ranking table, and each speech intelligibility information in the speech intelligibility ranking table is sorted from easy to difficult; according to the speech intelligibility ranking table, the first Target keywords are used to obtain test playback information; speech test results are obtained sequentially based on the test playback information; and second evaluation results are obtained based on all speech test results.

Specifically, by setting the required number of keywords, quantitative analysis is achieved by controlling the number of target keywords and the intelligibility of the masked sounds, and the corresponding number is selected from the preset test word library according to the set number. Keywords are used as target keywords, and the interference evaluation content set is set correspondingly to the masking sounds concentrated according to the types of masking sound information, such as noise target keywords, dynamic keywords, male voice forward, girl forward, boy reverse, and girl. Reverse, English forward, English reverse, etc., conduct understandability evaluation based on the effect of superimposing the interference evaluation content and the target keywords, and sort the evaluation results in order from easy to difficult. During the test It is suitable to choose forward playback or reverse playback to quantitatively evaluate the information masking effect results under different influencing factors. Test the test result threshold under different noise interference. Also during the test, the test user gives the corresponding answer according to the requirements. Compare the answers with the target keywords to obtain the evaluation results. In order to improve the accuracy of the evaluation results, multiple groups are still selected for each type of test. If the test is not completed once, the noise is dynamically adjusted according to the test results, and the test is continued. , to obtain the final test results, for example, select 10 groups, and when performing the final evaluation, perform a weighted average calculation on the selected test results to obtain the evaluation results. In order to improve the accuracy and computing efficiency of speech intelligibility assessment, embodiments of the present application add a neural network model. The masking analysis model is the trained neural network model. Masking analysis is implemented through training convergence to obtain the intelligibility of the corresponding speech.

Further, the method includes: obtaining the test target sound according to the first target keyword; obtaining a first incident angle and a second incident angle, where the first incident angle is different from the second incident angle; according to the The masking sound information type set is used to obtain multiple groups of test masking sounds, wherein the number of masking sounds in each group of the multiple groups of test masking sounds is different; the test target sound is transmitted according to the first incident angle, and the test target sound is transmitted according to the second incident angle. Multiple sets of test masking sounds are sent in sequence, and multiple sets of test results are obtained in sequence; and a third evaluation result is obtained based on the multiple sets of test results.

Specifically, since the propagation position and angle of sound will affect the reception effect, by adjusting the incident angle of the test target sound corresponding to the target keyword and the noise interference sound, that is, the masking sound, and at the same time superimposing interference according to the number of different masking sounds, It achieves masking evaluation under dual conditions of quantitative and relative position changes, and realizes multi-directional information masking evaluation. For example, the target sound is sent to the left ear of a boy, and the other two male masking sounds are sent to the left ear and the other to the right ear, or two to the left ear and one to the right ear. At the same time, the masking sounds can also be changed. There are boys and girls. Changes in the number of masking sounds and changes in incident angle can realize the evaluation of the masking effect under the dual conditions of quantification and spatial position of the masking sound, further realizing the classification evaluation of the masking sound through the control and change of the masking sound, improving the comprehensiveness of the evaluation. It has the technical effect of comprehensively evaluating each masking effect from multiple angles, enriching the angles of evaluation, improving the reliability of the evaluation results, and thus solving the problem of the existing technology that mainly evaluates the energy masking effect and lacks strict control over the factors that affect the information masking effect. , there are technical problems that affect the evaluation results.

Further, the method includes: setting the test target sound according to a first intensity; setting the test masking sound according to a second intensity, wherein the second intensity is different from the first intensity; and obtaining a binaural listening record. information; according to the binaural listening record information, obtain the first simulation parameter; according to the first simulation parameter, obtain the binaural listening time difference; according to the binaural listening time difference, sequentially send the test target sound, the test Masking sound and obtaining binaural test results.

Specifically, the binaural listening effect is evaluated and analyzed through historical listening data, the frequency and time relationship during binaural listening are obtained, and the first simulation parameter is obtained. The first simulation parameter is obtained through the analysis of historical listening records. Obtain the difference in frequency and ear entry time between the two ears during listening, control the transmission time of the target sound and the masking sound according to the binaural listening time difference, and simulate the binaural listening effect in daily life, thereby achieving the effect of quantitative binaural information masking assessment. , multiple types of masking sounds can be achieved by comprehensively adjusting the sound type, sending position, and sound size, and different types of tests can be completed under different masking sounds. If the target is to send a female voice to both ears, set two boys to interfere in one ear, and set one boy to interfere in one ear. The interference sounds in both ears can be equally loud, or you can set the right ear to be a few points louder than the left ear. It is also true that the left ear is larger than the right ear. Various factors are integrated and adjusted to achieve a multi-mode test environment and achieve the effect of binaural information masking evaluation. It further enriches the direction and angle of information masking and improves the comprehensiveness and reliability of the evaluation results. It further solves the problem that the existing technology mainly evaluates the energy masking effect, lacks strict control over the factors that affect the information masking effect, and has technical problems that affect the evaluation results.

Further, inputting the extracted word information and the preset sentence structure into a sentence construction model to obtain the test sentence information includes: inputting the word information and the preset sentence structure as input information The sentence construction model is obtained by training and convergence through multiple sets of training data. The multiple sets of training data all include the word information, the preset sentence structure and an identifier identifying the short sentence information. Information; obtain a second output result of the sentence construction model, where the second output result is a test short sentence information set.

Specifically, the preset sentence structure is a neural network model in machine learning, which can continuously learn and adjust and is a highly complex nonlinear dynamic learning system. Simply put, it is a mathematical model. Through training with a large amount of training data, a large amount of word information and the preset sentence structure are used as training training data for model training, the preset sentence structure is constructed, and the word information and the preset sentence structure are input into the neural network model, the test sentence information is output. Furthermore, the training process is essentially a supervised learning process. Each set of supervision data includes the word information, the preset sentence structure and identification information used to identify the test sentence information. The word information is The word information and the preset sentence structure are input into the neural network model. According to the identification information used to identify the test sentence information, the neural network model continuously corrects and adjusts itself until the obtained output result is consistent with the required If the above identification information is consistent, the supervised learning of this group of data will be ended and the supervised learning of the next group of data will be carried out; when the output information of the neural network model reaches the predetermined accuracy rate/reaches the convergence state, the supervised learning process ends. Through the supervised learning of the neural network model, the neural network model processes the input information more accurately, thereby obtaining more accurate and suitable test sentence information, and thereby achieving the results according to the selected words and corresponding settings. The short sentence structure is used to intelligently analyze and construct short sentences. The addition of the neural network model improves the efficiency and accuracy of data processing results, laying a solid foundation for providing more accurate and reliable masking information assessment.

Further, the method further includes: obtaining test user information; obtaining test user interference factors according to the test user information; inputting the test user interference factors into the masking analysis model to obtain predicted speech understandable information; The predicted speech intelligibility information and the speech intelligibility information are subjected to loss analysis to obtain the first loss data; the first loss data is brought into the masking analysis model for incremental learning to obtain an incremental masking analysis model. , the incremental masking analysis model is a new model obtained by incrementally learning the masking analysis model.

Specifically, when conducting speech intelligibility analysis, in addition to interfering with the difficulty of evaluating content and keywords, it is also related to the personal characteristics of the tested users. In order to improve the application scope of the masking analysis model, further expand The comprehensiveness of the information masking assessment is based on the analysis of personal interference factors based on the test user's personal information, such as age characteristics, living environment, personal interests and hobbies, vocabulary exposure, sensitivity to words, etc., and based on the personal interference factors The masking analysis model performs incremental learning, inputs new influencing factors into the model for calculation to obtain the corresponding predicted intelligible information, and calculates the loss function between the original intelligibility information and the intelligible information corresponding to the added factors to obtain the corresponding Loss data, use loss data to train the model, and improve the inclusiveness of the mask analysis model. Incremental learning means that a learning system can continuously learn new knowledge from new samples and retain most of the previously learned knowledge. Incremental learning is very similar to humans' own learning model. Because when people grow up, they learn and accept new things every day. Learning is carried out step by step, and humans generally will not forget the knowledge they have learned. It has the technical effect of reducing the training process and expanding the calculation effect of the model, making the speech intelligibility analysis more suitable for the user's characteristics, thereby ensuring the reliability of the masking evaluation results.

Further, the method further includes: obtaining the first evaluation result, the second evaluation result and the Nth evaluation result, generating a first verification code according to the first evaluation result, the first verification code and the first evaluation result One-to-one correspondence, N is a natural number greater than 1;

Step S1420: Generate a second verification code based on the second evaluation result and the first verification code, and so on, generate the Nth verification code based on the Nth evaluation result and the N-1th verification code;

Step S1430: Copy and store all evaluation results and verification codes on M electronic devices, where M is a natural number greater than 1.

Specifically, in order to ensure the security of the evaluation results, each evaluation result or the test results corresponding to each test function are stored through the blockchain to provide guarantee for later analysis and evaluation. Blockchain technology, also known as distributed ledger technology, is an emerging technology in which several computing devices jointly participate in "bookkeeping" and jointly maintain a complete distributed database. Because blockchain technology is decentralized, open and transparent, each computing device can participate in database records, and data can be synchronized quickly between computing devices, blockchain technology has been widely used in many fields. Apply. A first verification code is generated according to the first evaluation result, and the first verification code corresponds to the first evaluation result; a second verification code is generated according to the second evaluation result and the first verification code, and the second verification code corresponds one-to-one with the second evaluation result; and by analogy, the Nth verification code is generated according to the Nth evaluation result and the N-1th verification code, where N is a natural number greater than 1, and all evaluation results and verification codes are The copy is stored on M devices, wherein the first evaluation result and the first verification code are stored on one device as a first storage unit, and the second evaluation result and the second verification code are stored as a first storage unit. Two storage units are stored on one device. The Nth evaluation result and the Nth verification code are stored on one device as the Nth storage unit. When the evaluation result needs to be called, each subsequent node receives the previous one. After the data is stored in one node, it is verified and saved through the "consensus mechanism". Each storage unit is connected in series through the hash function, making the evaluation results less likely to be lost and destroyed. The logic of the blockchain is used to verify the data. The evaluation results are encrypted to ensure the security of the evaluation results, thereby ensuring the user's information security and ensuring the reliability of the information masking evaluation.

Embodiment 2

Based on the same inventive concept as the information-de-masking evaluation method based on closed-concentrated Chinese short sentence testing in the aforementioned embodiment, the present invention also provides a de-information-masking evaluation system based on closed-concentrated Chinese short sentence testing, as shown in Figure 2 As shown, the system includes:

The first obtaining unit 11, the first obtaining unit 11 is used to obtain test phrase information;

a second obtaining unit 12. The second obtaining unit 12 is used to obtain a masking sound information type set. The masking sound information type set includes steady background noise, dynamic noise, spectrogram noise, and multi-person speaking noise;

The third obtaining unit 13 is configured to obtain the first masking sound information according to the masking sound information type set;

The fourth obtaining unit 14 is configured to obtain a first test sound according to the first masking sound information and the test phrase information, wherein the first masking sound in the first test sound Acoustic information is energy masking or information masking;

A fifth obtaining unit 15, the fifth obtaining unit 15 is used to obtain a first test result based on the first test tone;

A sixth obtaining unit 16, the sixth obtaining unit 16 is used to obtain a first test adjustment tone according to the first test result and the first test tone;

A seventh obtaining unit 17, the seventh obtaining unit 17 is used to obtain a second test result based on the first test adjustment tone;

An eighth obtaining unit 18, the eighth obtaining unit 18 is used to obtain a second test adjustment tone according to the first test tone and the first test adjustment tone;

A ninth obtaining unit 19, the ninth obtaining unit 19 is used to obtain a third test result based on the second test adjustment tone;

A first screening unit 20, the first screening unit 20 is used to screen the first test results, the second test results, and the third test results according to the first set signal-to-noise ratio;

The tenth obtaining unit 21 is used to obtain the first evaluation result according to the screened test results.

Furthermore, the system also includes:

An eleventh obtaining unit, the eleventh obtaining unit is used to obtain a preset test word library, where the preset test word library includes preset type word information;

A twelfth obtaining unit, the twelfth obtaining unit is used to obtain a preset sentence structure, where the preset sentence structure corresponds to the preset type of word information;

A first execution unit, the first execution unit is used to extract words from the preset test word library according to the preset type;

A thirteenth obtaining unit, the thirteenth obtaining unit is used to input the extracted word information and the preset sentence structure into the sentence construction model to obtain the test short sentence information.

Further, the method includes:

A fourteenth acquisition unit, the fourteenth acquisition unit is used to obtain the target keyword quantity requirement;

A fifteenth obtaining unit, the fifteenth obtaining unit is used to filter from the preset test word library according to the target keyword quantity requirement to obtain the first target keyword;

A sixteenth obtaining unit, the sixteenth obtaining unit is configured to obtain an interference assessment content set according to the masking sound information type set;

A first input unit, the first input unit is used to input the interference assessment content set and the first target keyword into a masking analysis model. The masking analysis model is obtained through training convergence through multiple sets of training data, so Each of the multiple sets of training data includes interference assessment content, target keywords and identification information indicating speech intelligibility;

The seventeenth obtaining unit is used to obtain the first output result of the masking analysis model. The first output result includes speech intelligibility information and is constructed based on all obtained speech intelligibility information. A speech intelligibility sorting table, in which the speech intelligibility information in the speech intelligibility ranking table is sorted from easy to difficult;

An eighteenth obtaining unit, the eighteenth obtaining unit is used to obtain test playback information according to the speech intelligibility ranking table and the first target keyword;

A nineteenth obtaining unit, the nineteenth obtaining unit is used to obtain speech test results in sequence according to the test playback information;

A twentieth obtaining unit, the twentieth obtaining unit is used to obtain a second evaluation result based on all speech test results.

Furthermore, the system also includes:

The twenty-first obtaining unit, the twenty-first obtaining unit is used to obtain the test target sound according to the first target keyword;

A twenty-second obtaining unit, the twenty-second obtaining unit is used to obtain a first incident angle and a second incident angle, where the first incident angle is different from the second incident angle;

The twenty-third obtaining unit is configured to obtain multiple groups of test masking sounds according to the masking sound information type set, wherein the number of masking sounds in each group of the multiple groups of test masking sounds is different. ;

a second execution unit, the second execution unit is configured to transmit the test target sound according to the first incident angle, sequentially send multiple sets of test masking sounds according to the second incident angle, and sequentially obtain multiple sets of test results;

A twenty-fourth obtaining unit, the twenty-fourth obtaining unit is configured to obtain a third evaluation result according to the plurality of sets of test results.

Furthermore, the system also includes:

A third execution unit, the third execution unit is used to set the test target sound according to the first intensity;

A fourth execution unit, the fourth execution unit is used to set the test masking sound according to a second intensity, wherein the second intensity is different from the first intensity;

A twenty-fifth obtaining unit, the twenty-fifth obtaining unit is used to obtain binaural listening record information;

A twenty-sixth obtaining unit, the twenty-sixth obtaining unit is used to obtain the first simulation parameters according to the binaural listening record information;

A twenty-seventh obtaining unit, the twenty-seventh obtaining unit is used to obtain the binaural listening time difference according to the first simulation parameter;

The twenty-eighth obtaining unit is configured to sequentially send the test target sound and the test masking sound according to the binaural listening time difference to obtain binaural test results.

Furthermore, the system also includes:

A second input unit. The second input unit is used to input the word information and the preset sentence structure as input information into the sentence construction model. The sentence construction model is obtained through the convergence of multiple sets of training data. , the plurality of sets of training data all include the word information, the preset sentence structure and identification information identifying the short sentence information;

The twenty-ninth obtaining unit is used to obtain a second output result of the sentence construction model, where the second output result is a test short sentence information set.

Furthermore, the system also includes:

The thirtieth obtaining unit, the thirtieth obtaining unit is used to obtain test user information;

The thirty-first obtaining unit, the thirty-first obtaining unit is used to obtain the test user interference factors according to the test user information;

The thirty-second obtaining unit is configured to input the test user interference factors into the masking analysis model to obtain predicted speech understandable information;

A thirty-third obtaining unit, the thirty-third obtaining unit is configured to obtain the first loss data by performing loss analysis on the predicted speech intelligibility information and the speech intelligibility information;

The thirty-fourth obtaining unit is used to bring the first loss data into the masking analysis model for incremental learning, and obtain the incremental masking analysis model. The incremental masking analysis model is a new model obtained by incremental learning of the mask analysis model.

The various variations and specific examples of the de-information masking evaluation method based on the closed-concentrated Chinese short sentence test in the first embodiment of Figure 1 are also applicable to the de-information masking evaluation method based on the closed-concentrated Chinese short sentence test in this embodiment. Masking evaluation system, through the foregoing detailed description of a de-information masking evaluation method based on a closed-concentrated Chinese short sentence test, those skilled in the art can clearly understand a de-information based on a closed-concentrated Chinese short sentence test in this embodiment The implementation method of the masking evaluation system is not described in detail here for the sake of simplicity.

Example electronic device

The electronic device according to the embodiment of the present application will be described below with reference to FIG. 3 .

FIG. 3 illustrates a schematic structural diagram of an electronic device according to an embodiment of the present application.

Based on the inventive concept of a de-information masking evaluation method based on closed-concentration Chinese sentence testing in the foregoing embodiments, the present invention also provides a computer device with a computer program stored thereon. When the program is executed by a processor, the above mentioned The steps of any method of the information-de-masking evaluation method based on the closed-concentration Chinese short sentence test.

Among them, in Figure 3, the bus architecture (represented by bus 300), bus 300 may include any number of interconnected buses and bridges, bus 300 will include one or more processors represented by processor 302 and memory 304. The various circuits of memory are linked together. Bus 300 may also link together various other circuits such as peripherals, voltage regulators, power management circuits, etc., which are all well known in the art and therefore will not be described further herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303 . The receiver 301 and the transmitter 303 may be the same element, a transceiver, providing a unit for communicating with various other systems over a transmission medium.

The processor 302 is responsible for managing the bus 300 and general processing, while the memory 304 may be used to store data used by the processor 302 in performing operations.

One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

Embodiments of the present invention provide a method and system for de-information masking evaluation based on closed-concentration Chinese phrase testing. By obtaining test phrase information, a masking sound information type set is obtained. The masking sound information type set includes steady-state background noise. , dynamic noise, spectrogram noise, multi-person speaking noise; according to the set of masking sound information types, the first masking sound information is obtained; according to the first masking sound information and the test phrase information, the first test sound is obtained , wherein the first masking sound information in the first test tone is energy masking or information masking; based on the first test tone, a first test result is obtained; according to the first test result, the first test tone, obtain a first test adjustment tone; obtain a second test result based on the first test adjustment tone; obtain a second test adjustment tone based on the first test tone and the first test adjustment tone; based on the The second test adjusts the tone to obtain the third test result; the first test result, the second test result, and the third test result are screened according to the first set signal-to-noise ratio; and the first test result is obtained according to the screened test results. evaluation result. By classifying the masking sounds and continuously adjusting the noise dynamically, we can accurately evaluate the information masking effect of cochlear implants in children's multi-person competition context, and can separate energy masking from information masking to provide information masking. The evaluation field provides the technical effect of a more scientific and reliable auditory perception evaluation method, thus solving the existing technical problems that mainly focus on the evaluation of energy masking effects. There is a lack of strict control over the factors that affect the information masking effect, and there are technical problems that affect the evaluation results.

In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or can be integrated into another system, or some features can be ignored, or not implemented. On the other hand, the coupling or direct coupling or communication connection between each other shown or discussed may be through some interfaces, and the indirect coupling or communication connection of the devices or units may be in electrical, mechanical or other forms. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or software functional units. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code.

Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art will understand that various changes, modifications, and substitutions can be made to these embodiments without departing from the principles and spirit of the invention. and modifications, the scope of the invention is defined by the appended claims and their equivalents.

Claims (7)

1. A method of mask evaluation of information removal based on closed-set text phrase testing, wherein the method comprises:

obtaining test phrase information;

obtaining a masking sound information category set, wherein the masking sound information category set comprises steady background noise, dynamic noise, speech spectrum noise and multi-person speaking noise;

obtaining first masking sound information according to the masking sound information category set;

obtaining a first test sound according to the first masking sound information and the test phrase information, wherein the first masking sound information in the first test sound is energy masking or information masking;

obtaining a first test result based on the first test sound;

according to the first test result and the first test sound, a first test adjustment sound is obtained;

obtaining a second test result based on the first test adjustment sound;

Obtaining a second test adjustment sound according to the first test sound and the first test adjustment sound;

obtaining a third test result based on the second test adjustment sound;

screening the first test result, the second test result and the third test result according to a first set signal-to-noise ratio;

obtaining a first evaluation result according to the screened test result;

wherein, the obtaining test phrase information includes:

obtaining a preset test word library, wherein the preset test word library comprises preset type word information;

obtaining a preset sentence structure, wherein the preset sentence structure corresponds to a preset type of word information;

according to the preset type, extracting words from the preset test word library;

inputting the extracted word information and the preset sentence structure into a sentence construction model to obtain the test short sentence information;

the method further comprises the steps of:

obtaining the number requirement of target keywords;

screening from the preset test word library according to the target keyword quantity requirement to obtain a first target keyword;

obtaining an interference evaluation content set according to the masking sound information category set;

inputting the interference evaluation content set and the first target keyword into a masking analysis model, wherein the masking analysis model is obtained through training convergence by a plurality of groups of training data, and each group of training data in the plurality of groups of training data comprises interference evaluation content, target keywords and identification information for identifying speech intelligibility;

Obtaining a first output result of the masking analysis model, wherein the first output result comprises speech intelligibility information, and constructing a speech intelligibility ranking table based on all the obtained speech intelligibility information, wherein each piece of speech intelligibility information in the speech intelligibility ranking table is ranked from easy to difficult;

obtaining test playing information according to the speech intelligibility ordering table and the first target keywords;

according to the test playing information, speech test results are sequentially obtained;

and obtaining a second evaluation result according to all the speech test results.

2. The method of claim 1, wherein the method comprises:

obtaining test target sound according to the first target keyword;

obtaining a first incidence angle and a second incidence angle, wherein the first incidence angle is different from the second incidence angle;

obtaining a plurality of groups of test masking sounds according to the masking sound information class set, wherein the masking sounds of each group in the plurality of groups of test masking sounds are different in number;

transmitting the test target sound according to the first incidence angle, sequentially transmitting a plurality of groups of test masking sounds according to a second incidence angle, and sequentially obtaining a plurality of groups of test results;

and obtaining a third evaluation result according to the multiple groups of test results.

3. The method of claim 2, wherein the method comprises:

setting the test target sound according to a first intensity;

setting the test masking sound according to a second intensity, wherein the second intensity is different from the first intensity;

obtaining binaural listening recording information;

obtaining a first simulation parameter according to the binaural listening record information;

obtaining a binaural listening time difference according to the first simulation parameter;

and according to the binaural listening time difference, sequentially transmitting the test target sound and the test masking sound to obtain a binaural test result.

4. The method of claim 1, wherein said inputting the extracted word information, the preset sentence structure, and the input sentence construction model to obtain the test phrase information comprises:

inputting the word information and the preset sentence structure as input information into the sentence construction model, wherein the sentence construction model is obtained through training convergence of multiple sets of training data, and the multiple sets of training data comprise the word information, the preset sentence structure and identification information for identifying the short sentence information;

and obtaining a second output result of the sentence construction model, wherein the second output result is a test short sentence information set.

5. The method of claim 1, wherein the method further comprises:

obtaining test user information;

obtaining a test user interference factor according to the test user information;

inputting the tested user interference factors into the masking analysis model to obtain predicted speech intelligibility information;

obtaining first loss data by carrying out loss analysis on the predicted speech intelligibility information and the speech intelligibility information;

and carrying the first loss data into the masking analysis model to perform incremental learning to obtain an incremental masking analysis model, wherein the incremental masking analysis model is a new model obtained by performing incremental learning on the masking analysis model.

6. A closed-set text-phrase-test-based information-masking evaluation system, wherein the system is applied in the method of any one of claims 1-5, the system comprising:

the first obtaining unit is used for obtaining test phrase information;

a second obtaining unit, configured to obtain a masking sound information category set, where the masking sound information category set includes steady background noise, dynamic noise, speech spectrum noise, and multi-person speaking noise;

A third obtaining unit for obtaining first masking sound information according to the masking sound information category set;

a fourth obtaining unit, configured to obtain a first test sound according to the first masking sound information and the test phrase information, where the first masking sound information in the first test sound is energy masking or information masking;

a fifth obtaining unit configured to obtain a first test result based on the first test tone;

a sixth obtaining unit, configured to obtain a first test adjustment sound according to the first test result and the first test sound;

a seventh obtaining unit configured to obtain a second test result based on the first test adjustment sound;

an eighth obtaining unit, configured to obtain a second test adjustment sound according to the first test sound and the first test adjustment sound;

a ninth obtaining unit configured to obtain a third test result based on the second test adjustment sound;

the first screening unit is used for screening the first test result, the second test result and the third test result according to a first set signal-to-noise ratio;

A tenth obtaining unit, configured to obtain a first evaluation result according to the screened test result;

wherein the system further comprises:

an eleventh obtaining unit, configured to obtain a preset test word library, where the preset test word library includes preset type word information;

a twelfth obtaining unit for obtaining a preset sentence structure corresponding to a preset type of the word information;

the first execution unit is used for extracting words from the preset test word library according to the preset type;

a thirteenth obtaining unit for inputting the extracted word information and the preset sentence structure into a sentence construction model to obtain the test phrase information;

a fourteenth obtaining unit configured to obtain a target keyword number requirement;

a fifteenth obtaining unit, configured to screen from the preset test word library according to the number of target keywords, to obtain a first target keyword;

a sixteenth obtaining unit configured to obtain an interference evaluation content set from the masking sound information category set;

The first input unit is used for inputting the interference evaluation content set and the first target keyword into a masking analysis model, the masking analysis model is obtained through training convergence by a plurality of sets of training data, and each set of training data in the plurality of sets of training data comprises interference evaluation content, target keyword and identification information for identifying speech intelligibility;

a seventeenth obtaining unit, configured to obtain a first output result of the masking analysis model, where the first output result includes speech intelligibility information, and construct a speech intelligibility ranking table based on all obtained speech intelligibility information, where each speech intelligibility information in the speech intelligibility ranking table is ranked from easy to difficult;

an eighteenth obtaining unit, configured to obtain test playing information according to the speech intelligibility ordered list and the first target keyword;

a nineteenth obtaining unit, configured to obtain speech test results sequentially according to the test playing information;

and the twentieth obtaining unit is used for obtaining a second evaluation result according to all the speech test results.

7. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any of the preceding claims 1-5 when the computer program is executed by the processor.