Cognitive ability assessment device, mobile terminal, and speech acquisition device

Abstract

The embodiments of the present disclosure can provide a user's cognitive ability assessment results in a simple manner by analyzing the linguistic characteristics of the user's speech data, and through simple tests in daily life, can easily detect mild cognitive impairment or dementia in the user, allowing for early medical management and improving the quality of life of patients with degenerative brain diseases.

Description

[Technical Field]

[0001] An embodiment of the present disclosure comprises: Speech and EEG (Electro Encephalography) The present invention relates to a cognitive ability assessment device, a mobile terminal, and a speech acquisition device. [Background technology]

[0002] Advances in medical technology have increased the average lifespan of people. This increase in lifespan has led to a global trend toward an increasing proportion of elderly people in the overall population. Elderly populations are more susceptible to a variety of geriatric diseases, and healthcare for the elderly population has emerged as a national and societal issue.

[0003] In particular, with the continuing increase in the elderly population, Alzheimer's disease, which the elderly fear most, is becoming a serious problem. -( The number of elderly people suffering from degenerative brain diseases such as Alzheimer's disease and Parkinson's disease is constantly increasing. Once these degenerative brain diseases develop, there are no cures available, and they rapidly deteriorate quality of life, as well as incur high costs for management and maintenance. Therefore, medical management of the elderly population suffering from these degenerative brain diseases poses a significant personal and social burden. Summary of the Invention [Problem to be solved by the invention]

[0004] The embodiment of the present disclosure relates to speech in people's daily lives. or EEG The linguistic (or language) and acoustic (or speech) characteristics detected through this method can be used to identify cognitive decline, providing a method for easily identifying early-stage mild cognitive impairment (MCI) and early-stage dementia. The embodiments of the present disclosure can provide a device that allows elderly people to conveniently collect speech and brain waves in their daily lives. [Means for solving the problem]

[0005] An embodiment of the present disclosure can provide a cognitive ability assessment device that includes a task providing unit that provides a user with a plurality of tasks between a first test period and a second test period after the first test period, an utterance data collecting unit that acquires digital utterance data converted from test audio in which the volume of at least a portion of the basic audio is adjusted according to the user's response to at least one of the plurality of tasks, and a cognitive ability measuring unit that calculates an assessment result regarding the user's cognitive ability based on at least one linguistic characteristic of the utterance data.

[0006] An embodiment of the present disclosure includes a task provision module that provides a user with a plurality of tasks during a test period, and acquires speech data in a digital format converted from a test voice in which the volume of at least a portion of a base voice is adjusted according to the user's response to at least one of the plurality of tasks. EEG acquisition around the ear and A mobile terminal may be provided that includes a speech data collection module.

[0007] The mobile terminal may further include a cognitive ability measurement module that calculates an assessment result regarding the user's cognitive ability based on at least one linguistic characteristic of the speech data.

[0008] An embodiment of the present disclosure can provide a speech acquisition device that includes an output module that outputs audio corresponding to multiple tasks provided to a user, an input module that acquires basic audio corresponding to the user's response to at least one of the multiple tasks, and a processing module that analyzes the volume or frequency band of the basic audio, amplifies the volume of low-volume basic audio based on the analysis, adjusts so as not to amplify high-volume basic audio that is easy to recognize, generates test audio that can recognize even the user's low-volume voice, and converts the test audio into digital speech data.

[0009] The speech capture device may further include a communication module for transmitting the speech data to the outside. [Effects of the Invention]

[0010] According to an embodiment of the present disclosure, changes in cognitive ability can be easily confirmed from speech produced by a user in an everyday life environment, rather than in a medical facility, and evaluation results for diseases such as mild cognitive impairment or dementia can be provided through an absolute evaluation of the cognitive ability of an individual speaker based on the user's individual speech information, a comparative evaluation of an individual's cognitive ability by time period based on periodic evaluation information, or a relative comparison of cognitive ability based on an individual's age, educational background, etc. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a diagram illustrating an example of a configuration of a cognitive ability assessment system according to an embodiment of the present disclosure. [Figure 2] FIG. 1 illustrates an example of how a cognitive performance assessment system performs a cognitive performance assessment according to an embodiment of the present disclosure. [Figure 3] FIG. 1 illustrates an example of how a cognitive performance assessment system performs a cognitive performance assessment according to an embodiment of the present disclosure. [Figure 4] FIG. 1 illustrates an example of how a cognitive performance assessment system performs a cognitive performance assessment according to an embodiment of the present disclosure. [Figure 5] FIG. 1 illustrates an example of how a cognitive performance assessment system performs a cognitive performance assessment according to an embodiment of the present disclosure. [Figure 6] FIG. 10 is a diagram schematically illustrating another example of the configuration of a cognitive ability assessment system according to an embodiment of the present disclosure. [Figure 7] FIG. 7 is a diagram illustrating an example of the configuration of the utterance capture device shown in FIG. 6. [Figure 8] FIG. 7 is a diagram illustrating an example of the configuration of the utterance capture device shown in FIG. 6. [Figure 9] FIG. 10 is a diagram schematically illustrating yet another example of the configuration of a cognitive ability assessment system according to an embodiment of the present disclosure. [Figure 10] FIG. 1 illustrates an example of how a cognitive performance assessment system performs a cognitive performance assessment according to an embodiment of the present disclosure. [Figure 11] FIG. 1 illustrates an example of how a cognitive performance assessment system performs a cognitive performance assessment according to an embodiment of the present disclosure. [Figure 12] 1 is a flowchart illustrating an example of the steps of a cognitive ability assessment method according to an embodiment of the present disclosure. [Figure 13] 1 is a flowchart illustrating an example of the steps of a cognitive ability assessment method according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0012] Some embodiments of the present disclosure will be described in detail below with reference to exemplary drawings. When adding reference numerals to components in each drawing, the same components may be assigned the same numerals whenever possible, even if they appear in different drawings. In describing the present disclosure, if it is determined that a specific description of related publicly known configurations or functions may obscure the gist of the present disclosure, such detailed description will be omitted. When terms such as "include," "have," and "comprise" are used in this specification, other parts may be added unless "only" is used. When a component is expressed as a singular, it may also include a plural number unless otherwise explicitly stated.

[0013] Furthermore, when describing components of the present disclosure, terms such as first, second, A, B, (a), (b), etc. are used. These terms are used only to distinguish the components from other components, and do not limit the nature, order, sequence, number, etc. of the components.

[0014] When two or more components are described as being "coupled," "coupled," or "connected" in a description of the positional relationship of components, it should be understood that the two or more components may be directly "coupled," "coupled," or "connected," but that the two or more components may also be "coupled," "coupled," or "connected" to other components through further "intervening" connections. Here, the other components may be included in one or more of the two or more components that are "coupled," "coupled," or "connected" to each other.

[0015] In describing the temporal relationship between components, methods of operation, methods of production, etc., when a temporal or sequential relationship is described using, for example, "after," "following," "next," or "before," this may also include cases where the relationship is not consecutive, unless "immediately" or "directly" is used.

[0016] On the other hand, when a numerical value or its corresponding information (e.g., level, etc.) relating to a component is mentioned, the numerical value or its corresponding information may be interpreted as including a range of error that may occur due to various factors (e.g., process factors, internal or external impacts, noise, etc.) even if there is no other explicit statement.

[0017] Various embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings.

[0018] Recently, various methods for early detection of dementia have been investigated. According to the 4th Comprehensive Dementia Management Plan, dementia screening tests, diagnostic tests, and differential tests are conducted. Among these, dementia screening tests are conducted at Dementia Relief Centers and use testing tools such as MMSE-KC, MMSE-DS, and CIST. In the second stage of diagnostic testing, CERAD-K and SNSB are used. If abnormalities are found through these screening and diagnostic tests, detailed tests such as beta protein analysis using PET imaging analysis are conducted to accurately diagnose the disease. These tests can be conducted as detailed tests and diagnosed directly at a hospital.

[0019] The cognitive ability assessment method using the cognitive ability assessment system according to the embodiment of the present disclosure can be performed in a short time at welfare centers or residential areas frequently visited by elderly people, rather than at designated medical institutions. In addition, the proposed method can select high-risk groups for mild cognitive impairment. (Screening performed) By conducting detailed and diagnostic tests only on selected high-risk groups, unnecessary testing costs for those who are not at high risk can be reduced, and the government can provide national support only to high-risk groups, thereby reducing national costs. Furthermore, the proposed testing method can be easily and conveniently performed by elderly people anytime, anywhere, and can be repeated periodically (e.g., every six months) for continuous progress, allowing for continuous management and monitoring of high-risk groups with symptoms of early dementia over time, rather than being a one-time method at existing medical institutions.

[0020] As is known to date, once dementia progresses to moderate stage, it is difficult to treat or alleviate symptoms. Therefore, detecting early dementia and slowing the progression of dementia through drug therapy or hospital treatment is being promoted as a good solution, allowing patients to continue their daily activities.

[0021] An efficient method for detecting actual early-stage dementia symptoms is to identify high-risk groups for dementia through selective testing and periodically test and diagnose these high-risk groups, which is much more efficient than existing random (about 1 in 10 elderly people) or testing of all elderly people. In particular, it would be extremely useful if high-risk groups for dementia could be identified early through continuous selective testing, and the state could actively manage these high-risk elderly people and delay their progression to a stage of dementia that requires state management.

[0022] Recently, several methods have been proposed to identify high-risk individuals, such as monitoring the walking habits of elderly people in their daily lives and long-term monitoring of sleep patterns.

[0023] Another method is the active research into monitoring degenerative brain diseases such as dementia based on the utterances of the elderly. Data expressed in sound, such as conversations (acoustic data or speech data), contains various information about a person's health condition. For example, a method based on analyzing the characteristics of the sound (timbre, speed, volume, etc.) can be proposed. Alternatively, there is a method of converting the utterance information into the frequency domain and integrating artificial intelligence technology in the frequency domain to find acoustic features related to the disease. Here, utterances are learned through the five senses, such as listening to sounds, seeing questions, and acquiring various information that should be expressed in words. When you get it, it's expressed in audio. The response to this This could be the answer.

[0024] This method involves either listening to or showing prepared questions to induce speech, or presenting undetermined questions (such as questions about daily life or conversations) to the subject using audio or images, and then saving, recording, or filming the speech data. Tasks such as The analysis is performed in real time or after the fact through the voice recognition system. The analysis of voice characteristics may mainly involve analyzing timbre, speed, volume, etc. Such acoustic feature-based methods may have excellent screening capabilities for dementia. In addition, it may be necessary to extract common phonetic features taking into account differences in dialect usage and individual differences in vocal structure.

[0025] Another method using dialogue or speech data from the elderly is to analyze the level of language comprehension, vocabulary used, explanations of shapes, descriptions of daily life, etc. or dialogue Using data, we can analyze response times, interjections used, words used, and sentence structure. Linguistic features in answers to questions such as Linguistic features )of There is a way to analyze it.

[0026] In this method, the examiner uses pre-tested questions in a one-on-one face-to-face interview to obtain speech data for the elderly. (called a task)The test is played or shown to the elderly, and their responses to each question are recorded or saved in a specific storage area (e.g., the memory of a smart device). To analyze the results, the recorded audio is converted into text (called a transcription) and saved so that each response is matched to the question used. For each response, the transcription also displays the elderly's response time to the question, the number of words used, the accuracy of the sentence structure (e.g., whether the subject, object, and predicate are in the correct order), the appropriateness of the words used, the difficulty of the words used, and the interjections used. Until now, the audio recorded during the test has been manually transcribed, and the transcribed data is then organized according to analysis rules. In addition to the above, these rules may include the degree of repetition of similar words, the limit of the words used, and, if a word needed for an explanation cannot be remembered, the interjections used when the word cannot be remembered.

[0027] Furthermore, a process of synchronizing the recorded speech segments with the transcription text is required for analysis. This is called labeling. There are also tools available for labeling Hangul.

[0028] These methods take a long time, several hours or more, just to collect the speech data, and not only do they require the speaker and the client to conduct the test in person at a specific location such as a hospital, but they also require a lot of manpower and time to transcribe the collected voice data.

[0029] In recent years, research has been actively conducted into integrating artificial intelligence (AI) technology to recognize spoken speech and analyze its linguistic and acoustic structure. For example, a speech recognizer can be used to convert spoken speech data into text. AI speech recognizers can also be used for labeling and transcription.

[0030] By using such speech recognizers, various speech recognition devices and equipment have been developed, and services using them are being activated. Devices that acquire voice or spoken information can be smart devices capable of two-way information exchange, such as smartphones and tablets, conversational robots, and AI speakers. The basic configuration of these devices can be a microphone that listens to voice for conversation, an AI-based speech recognizer that recognizes the input voice, response technology based on the recognized voice, and finally a speaker that explains or listens to the question.

[0031] Current methods for collecting speech information using microphones built into smart devices such as portable tablets require the subject to look at or listen to the smart device. As another example, speech questions can be provided and speech information can be acquired through fixed devices with built-in AI speech recognition, such as AI speakers, social robots, and chatbots. Whether the device is portable or fixed, there is a certain distance between the speaker and the microphone used to collect speech data. This causes the microphone to receive not only the speaker's voice but also other voices and ambient noise. In this case, the speech recognizer of an AI speaker or social robot may have poor speech recognition performance in noisy environments, failing to correctly recognize the speaker's voice and convert it into text. Alternatively, accurate speech recognition or conversion of speech data into text may not be achieved. For example, speech spoken close to the built-in microphone of the device used to collect speech has a good recognition rate, but the speech recognition rate decreases with even a small distance. The reason for this is that the distance between the speaker and the microphone reduces the volume of the input voice, and other surrounding sounds (environmental noise, everyday noise, voices other than the speaker, etc.) are also included. In other words, the spectrum of the input voice contains noise components in addition to the speaker's original voice, which causes many errors in acoustic analysis. This is because the noise and the speaker's voice mix together and cannot be physically separated.

[0032] One way to solve this problem is to use multiple directional microphones in smart devices to collect only the speaker's voice, or to incorporate a noise reduction algorithm into smart devices to remove unwanted noise from the voice input through the microphone.

[0033] Furthermore, if the distance is a little farther, the subject may not be able to hear the questions or dialogue properly. In the absence of general assistive devices or in an environment with external noise, the subject may have problems hearing the questions correctly, which may result in the subject being unable to answer. However, the analysis shows that there is an error that can be determined to be a cognitive impairment.

[0034] Furthermore, in order for the subject (speaker) to hear the question and be guided to speak, the process of the subject seeing and hearing the content of the question may be very important. In order to enable the subject (e.g., elderly people) to hear the question clearly, the embodiment of the present disclosure may provide a function of amplifying the audio of the question to a level suitable for the user so that the question can be clearly heard not only by elderly people with hearing loss but also by elderly people with normal hearing.

[0035] While existing proposed methods can produce normal evaluation results in quiet spaces without external noise, there are various limitations to applying related technologies in spaces where elderly people are actually active. Another method is to use wireless earphones with microphone functionality. The explanations on the questionnaire can be clearly heard through the wireless earphones, and the audio received through the wireless earphones can be transmitted to a smart device (smartphone, tablet, computer, etc.). Furthermore, the volume of the questions can be kept constant regardless of the speaker's condition, improving the reliability of the test.

[0036] However, when this method is used in daily life, in addition to the speaker's speech data, the speaker's voice is transmitted together with external noise.

[0037] The embodiments of the present disclosure can provide a solution that can overcome the above-mentioned problems in practical application of screening testing technology for cognitive disorders in the elderly that applies linguistic / acoustic analysis technology based on actual speech data from the daily lives of the elderly.

[0038] Therefore, in order to commercialize a convenient and efficient screening test technology for cognitive impairment and early dementia using a language testing technology, an embodiment of the present disclosure may provide a method for easily and conveniently recording speech data while listening to and viewing test questions not only in non-medical public institutions such as senior welfare centers that are easily accessible to the elderly, but also in actual individual living spaces, etc. Furthermore, testing and evaluation may be possible even in everyday living spaces, even when there are noises or other disturbances.

[0039] Since most elderly people have hearing loss due to aging, methods to better understand spoken information should also be proposed. It is preferable to amplify the sound during the test depending on the degree of hearing loss. Examples of such methods include adjusting the volume when using earphones, and amplifying the sound appropriately when no earphones or other devices are available. Therefore, embodiments of the present disclosure may also include adjusting the volume of the sound to suit the elderly's ability to hear the questions.

[0040] Furthermore, a method for transmitting speech information to a cloud server as quickly as possible and a method for analyzing speech information are also required so that the results of speech data analysis can be provided to the user immediately after the test.

[0041] In particular, technology is needed to select optimal question sets for various tasks to make testing as efficient and fast as possible. Furthermore, technology may also be needed to shorten testing time and guide test subjects to stay focused on the test while it is being conducted.

[0042] The information gathering method in such an audio approach generally requires a method for the speaker to learn and provide questions for speaking, and a device to store the spoken information when the speaker speaks.

[0043] Generally, in language tests, questions are asked through a smart device, and if necessary, the test subject can view and respond to the questions on the screen of the smart device. These smart devices can be of several types. For example, they can be mobile tablets, smartphones, notebooks, mobile robots, etc., or fixed devices such as personal computers.

[0044] [Method for collecting speech data using smart devices and wearable devices, and analysis method using the cloud and AI]

[0045] An embodiment of the present disclosure proposes a method for conducting cognitive impairment or language tests via an app or the web using a wearable device and a smart device. The method also proposes a method for screening for mild cognitive impairment and early dementia by utilizing the cloud and AI technology for the tests. This method enables tests to be conducted not only at services, senior citizen centers, dementia relief centers, and senior welfare centers, but also in the everyday spaces where seniors actually live, by utilizing their own smart devices, allowing seniors to conveniently, easily, and periodically use the testing service.

[0046] In particular, to solve these problems, an embodiment of the present disclosure uses a wearable device that can wirelessly listen to questions posed via a web or app on a smart device, receive audio responses to the questions, and transmit the audio responses to the smart device without distortion. The audio responses transmitted from the wearable device to the smart device are stored on the smart device, and the stored audio is transmitted to the cloud.

[0047] As another example, while the question is being asked, the audio collected through the microphone of the wearable device is stored directly on the wearable device, and the audio data is delivered all at once.

[0048] Labeling can also be easily handled by listening to questions and saving answers via the wearable device's web or app. This is because questions and answers are saved consistently. A built-in gyro or acceleration sensor can also be used to record answers using acceleration information in the case of multiple-choice questions. The time it takes to answer a question is particularly important, and providing information on the end and maintenance of the answer allows for more accurate information about the answer being spoken.

[0049] The proposed method builds standardized big data on speech from Korean elderly people and uses it to learn the linguistic features of subjects with mild cognitive impairment and early dementia symptoms. The big data can also include long-term test data on the same individuals, allowing for the extraction of linguistic features over time. The classifier thus trained preferably indicates the degree of normality and abnormality in cognitive impairment. For example, it can indicate the degree of deviation from the average for normal elderly people. Furthermore, it can indicate the degree of deviation from the existing accumulated database with respect to the current state of mild cognitive impairment based on speech data extracted from patients with mild cognitive impairment. That is, the results of the language test are displayed by converting the evaluation for each individual question into a score, and the score for each question may vary depending on the type of task used to guide speech. For example, weights and importance scores for linguistic features such as the number of words used, frequency, sentence structure, and speech start time after question are prepared in advance and applied to the collected speech data. The question speech data is quantified as the sum of the weighted linguistic features.

[0050] In speech testing, a method for speaking about a given topic within a limited time can be selected. The given topic can be presented using audio, images, or text. A variety of topics can be presented, including family names, animal names, place names, and country names. Any topic word can be selected as long as the number of answers can be counted. Therefore, it is possible to count not only nouns but also predicates such as beautiful, pretty, and adorable, or verbs such as run, sprint, and escape. Therefore, any topic word that can be spoken or that can derive the number of related words is acceptable.

[0051] The time limit should be limited to approximately a few minutes for a general topic, but at least 30 seconds of speech must be possible, and it is recommended that speech be designed to be completed within three minutes per topic.

[0052] A simple method of evaluation involves assessing cognitive function by simply evaluating the number of words for each topic, the difficulty of the words, and the number of times the same word is repeated.

[0053] Furthermore, by evaluating two or more themes rather than one, the accuracy of cognitive function can be improved. In other words, by evaluating multiple themes, the results of the evaluation for one theme and the evaluation for the second or third theme can be presented separately, and the three results can be combined to perform the evaluation.

[0054] Tasks include comprehension tests, expression / speaking tests, and memory tests, and each task has a variety of questions. Quantification methods include quantifying the spoken data of the questions used in the task and calculating the average value. It is also possible to quantify the test results for each task independently.

[0055] Once each test result has been quantified as described above, the speech data can be translated into linguistic data collected from tailored language tests administered to normal subjects, MCI patients, dementia patients, etc., and transcribed to derive data indicating linguistic features (word frequency, speech time after question, words used, exclamations, sentence structure, etc.).

[0056] In particular, current language tests include a variety of tasks, each with several corresponding questions. For example, language tests include comprehension tasks, expression tasks, and memory tasks. Expression tasks also include a variety of questions, such as describing shapes and telling fairy tales. Memory tasks involve reading a list of words related to animals and fruits, and then reciting the words that are recalled. Questions in each task are of similar difficulty, and some are highly difficult. However, depending on the actual subject, it may not be necessary to ask similar questions for all tasks. To address this issue, it is desirable to select optimal questions for each task. Methods for selecting these optimal tasks and questions can also be derived based on big data from language tests. For example, the difficulty of tasks can be varied depending on the subject's condition. Alternatively, tasks with overlapping characteristics can be optimized and used. To date, there has been little academic research on question sets that take task-specific correlations into account. Similar to currently used methods, this optimal question set can be extracted through AI learning by extracting correlation characteristics between questions based on the subject's gender, age, educational background, etc., based on big data collected through almost all tasks and the questions for each task. For example, it is possible to select a question set for a task that matches the difficulty of describing a picture according to educational background.

[0057] Alternatively, there may be task-specific questions prepared in advance, and the selection of questions may be based on the information and answers from previous questions. The information and answers from previous questions are analyzed in real time. The analysis includes the speaker's understanding of the question, their ability to answer, and linguistic features. Therefore, the selection of the next question is based on basic information about the speaker, such as their gender, age, and educational background, as well as the analysis results from the previous questions. Artificial intelligence techniques can also be applied to create appropriate models for selecting these next questions.

[0058] In this way, the optimal task set for the task can be selected (the selection method utilizes an AI model that selects the optimal task set based on big data on existing subjects) and the test can be carried out. This shortens the test time and makes it easier for the elderly to take the test.

[0059] Therefore, in order to obtain speech information for each task, it is necessary to determine the optimal task list for each task of each question.

[0060] [Function definition of wearable devices]

[0061] As mentioned above, in the voice reception method using a microphone installed in a smart device, there is a certain distance between the smart device and the speaker, and in addition to the speaker's speech information, ambient noise and other people's voices are added together, which reduces the accuracy of the results of linguistic and acoustic analysis using the input speech information.

[0062] The present disclosure provides a method for accurately collecting speech data using a wearable device. The proposed wearable device is preferably wirelessly connected to a smart device. A separate app / web application may be used to connect the wearable device to the smart device.

[0063] Wearable speech collection devices for seniors can include noise reduction technology to filter out external noise. Noise reduction technology not only improves the speaker's voice recognition rate by directly removing various everyday noises and irregular sounds, but also removes unnecessary signals in the voice characteristic analysis method, allowing for clearer confirmation and analysis of the characteristics of the voice itself, making it possible to easily perform language testing even in noisy environments.

[0064] Voice signal processing, which is performed to improve the voice recognition rate of AI voice recognition equipment in wearable devices and to preserve as much speech information as possible from the speaker, can have two possible functions, for example.

[0065] First, there may be a function to remove noise depending on the environment. For example, noise removal may be necessary when there is noise other than the speaker's voice or other sounds coming from the surrounding environment. This can improve the speaker's voice recognition characteristics in the final evaluation stage.

[0066] Second, if the speaker's voice volume is low, it needs to be increased. Furthermore, if there is low-pitched speech in the speaker's speech, it needs to be further searched for in the final speech recognition process. When a person listens and transcribes, they can easily understand the speech if they listen carefully. However, in everyday life, the situation can be dramatically different when using a speaker's speech information for recognition by an AI speech recognition device. Specifically, any monologue or interjections included in the speech may be quiet and be filtered out by the AI ​​speech recognition device, potentially resulting in the loss of information needed for analysis. Therefore, in order to ensure that speech data displayed quietly in addition to the speech accurately produced by a wearable device is also obtained, it is necessary to amplify the speech to improve the speech recognition rate of the AI ​​speech recognition device. Furthermore, since noise is also amplified during amplification, it may be desirable to perform noise reduction and amplification simultaneously.

[0067] The noise reduction process is a method of converting a sound signal input through a microphone into a frequency domain and removing noise power in the frequency domain.

[0068] To remove noise, a speaker's voice is learned and stored in advance. At this time, the speaker's voice characteristics, such as the frequency characteristics and loudness information of the voice, can be extracted and utilized. A method of removing both voices containing information other than the acquired speaker's voice characteristics and everyday noise can also be used. To remove noise, speech data input in real time is converted into the frequency domain, and the previously stored speaker's voice information is utilized to remove everyday noise and voice information that does not correspond to the speaker's characteristics.

[0069] Another method is to use a classifier or detector that can classify noise and speech environments. In other words, if noise and speech can be efficiently separated, erroneous speech recognition will not occur, and noise power can be removed even in speech frequency analysis. Therefore, by improving the signal-to-noise ratio (SNR), speech can be heard more clearly and the content of noise can be heard in clearer speech. In addition, a method can be used in which an intelligent classifier is used through learning of traffic noise, everyday noise, etc. to remove data other than the speaker's voice from speech data input through a microphone.

[0070] In addition to this, various noise reduction techniques can be applied, in particular, by transforming the input signal into a frequency band, several methods can be applied to remove noise from the frequency band.

[0071] The wearable device can have a receiver (speaker) that can hear the spoken task questions from the smart device. The speaker is preferably inserted into the ear and can also be made as a bone conduction type. This allows the task questions to be heard more accurately by using only the smart device, which can be difficult to hear due to external noise. However, by using the proposed method, the questions can be heard more accurately, which is advantageous for testing.

[0072] The microphone of the wearable device is advantageously located in the vicinity of where the subject speaks. For example, it may be a neckband-type speech collector or an earphone-type. A lightweight wearable device is used, which, if possible, can collect speech information directly from the subject.

[0073] Wearable devices can use directional microphones, which can be used to more clearly receive only the sound coming from the speaker's mouth.

[0074] Also, When you send audio, the sound quality itself may be distorted, which may result in a lower voice recognition rate. Instead of transmitting audio, it transmits digitized data, which can be converted into audio from smart devices. 。

[0075] The wearable device can selectively amplify the sound received through the microphone, for example, by amplifying a specific frequency band, a high frequency band, or a low frequency band. .Ma The audio characteristics of audio received via a microphone may change slightly depending on the characteristics of the microphone or the analog-to-digital converter (ADC) that converts analog to digital signals. This digitized audio source can then be wirelessly transmitted to a smart device after small signals in certain frequency bands are removed according to the characteristics of the wireless communication and the signal-to-noise ratio. For example, if an elderly person with cognitive impairment cannot remember the answer to a question, they may mutter quietly to themselves while thinking about the answer. This information can be used as an important item in linguistic characteristic analysis, and it is necessary to ensure as much original speaker data as possible. Alternatively, if a certain expression is too quiet, the answer to the question may not be recognized, leading to incorrect evaluation results. In other cases, the elderly person's voice itself may be quiet, resulting in a low audio signal being input to the smart device.

[0076] Generally, when a signal input from a microphone is digitized, small sounds may be removed due to the characteristics of the device and the audio signal processing process. Although this does not pose a major problem for communication in everyday life, various information may be lost in terms of linguistic and acoustic characteristics. If small sounds are removed, they may not be included in the evaluation results even though they should be reflected in the evaluation score, which may result in an incorrect final evaluation result.

[0077] Therefore, this problem can be solved by using a method of amplifying quiet sounds and not amplifying loud sounds in the speech data received by a microphone. Also, by analyzing the frequency band of the input sound and selectively amplifying it according to frequency, the degree of speech recognition can be improved, expressions with speech characteristics can be prevented from being removed, and quiet sounds can be amplified to accurately evaluate the speech data.

[0078] The advantage of using a wearable device for amplification is that it allows the original speaker's voice signal to be transmitted and analyzed with the least distortion possible. In other words, this can be done on a server with very powerful computing power, but there is a problem in that much linguistic and phonetic information is lost during transmission to the server.

[0079] The wearable device can store answers to individual task questions directly on the wearable device, and can send multiple answers together to the smart device. The speech data received for each question can also be wirelessly transmitted to the smart device before the next question begins.

[0080] [App / Web function definition]

[0081] The speech collection app / website must include the following features: When performing a task, the voice instructions for the task must be transmitted to the wearable device. In addition, by monitoring the process performed by the elderly via the smart device, it must have a function to assist in case the test is incorrect due to an elderly person's mistake or error. In general, if an error occurs during the test via the app / website or if a retest is required, the tester should be able to rerun the test. Also, if the test is interrupted by pressing the wrong button during the test, the tester's assistance can be provided to resolve the related issue.

[0082] The spoken data for each question is synchronized and saved, and the data saved on the smart device is sent to the cloud according to a specified protocol.

[0083] The answers to the questions sent to the cloud for each task are passed through a model that has already been trained on the disease to determine whether the individual is at risk for mild cognitive impairment, dementia, or a steady state. These determinations can also be made as scores, showing the degree of risk compared to the normal standard. For example, if the results indicate a high risk, the hospital can provide detailed screening and diagnostic testing, which can lead to early treatment.

[0084] As such, cloud-based evaluation methods can derive highly reliable results by using a large amount of data, but they require the speech data to be sent directly to the cloud, which requires the cloud.Therefore, we propose a method that can provide services using only wearable devices, even in environments where the cloud is not available.

[0085] In this case, a wearable device is equipped with a speech database that classifies mild cognitive impairment (MCI) or cognitive impairment. The subject speaks while viewing and listening to the task questions via the smart device. The speech data is input via the wearable device's microphone, and noise is removed and the audio is amplified as needed. The wearable device is equipped with a pre-trained cognitive impairment classifier, which uses the speech data input to the microphone to quickly identify cognitive impairment. Taking into account the computing and memory capabilities of the wearable device, the classifier preferably uses smaller linguistic and volume features than cloud-based classifiers.

[0086] As a method for collecting speech data, each of the various tasks can be automatically saved to facilitate labeling.

[0087] For example, a method of skipping if there is no audio for a certain period of time or more can be used; a method of recording answers using physical sensor information in the case of multiple choice questions using a built-in gyro or acceleration sensor; a method of specific interaction with the device (for example, a method using a command, a method using questions such as "Next" or "Do you want to proceed?"), or a method of skipping using a physical function unique to the device (touch type, button type, etc.). In addition, a message may be output at regular intervals to alert the subject to the need to concentrate on the progress of the task. The content and format of the message may vary, and for example, the name of the subject may be called, or a comment such as "Please concentrate on the test" may be output in the form of a sound. Alternatively, a method of outputting light, an image, or the like in a form that is advantageous for focusing attention, such as flashing a light, may be used.

[0088] This proposed method not only makes it easier to label data when saving it, but also makes it easier to separate each task, subdivide the various questions included in each task, and select the optimal question set for each task, which significantly reduces the screening time.

[0089] Such a wearable device can be used only to store the speaker's speech and transmit it to a smart device. That is, the subject can hear and see the questions directly through the smart device. The wearable device collects, stores, and transmits the speech for each question to the smart device. The connection between the wearable device and the smart device is preferably wireless. Speech data collected through the microphone in the wearable device can be converted to digital data via an ADC, transmitted as digital data, or converted back to audio and transmitted as audio. During this process, noise and sounds other than the speaker are filtered out to facilitate speech recognition and analysis.

[0090] [About service methods]

[0091] The service method scenario is as follows:

[0092] The subject wears a wearable device and can view or listen to test questions via a smart device app or a PC via the web. The test proceeds while listening to and viewing the test questions administered via the smart device app through the wearable device. During the questions, the speaker's voice is collected via the wearable device's microphone and transmitted to the smart device as digital or audio data. The smart device then sends the speaker's responses to each question to the cloud, where an evaluation model for each answer is created and used to score each answer. Each answer is evaluated and analyzed using a separate speech recognizer and a normal / mild cognitive impairment classifier. The input speech data is converted into text via the speech recognizer. The converted text is then analyzed and evaluated according to certain rules. The classifier learns and stores linguistic and speech characteristics for normal subjects and mild cognitive impairment patients using answers to previously presented questions. Once the speaker has completed answering the entire test, the scores are totaled and the speaker is provided with an indication of how much their level of mild cognitive impairment, depression, etc. deviates from that of a normal person.

[0093] Another method utilizes a wearable device equipped with a speech recognizer and a classifier capable of diagnosing mild cognitive impairment (MCI) and depression. In other words, the subject wears the wearable device and can see, hear, and answer questions while looking at the smart device. Speech data is stored via the wearable device's microphone. The wearable device has a built-in speech recognizer and classifier. The speech recognizer converts the input speech data into text. The converted text is analyzed and evaluated according to certain rules. While it is preferable for the wearable device to know the speaker's question, it is also possible to do this without knowing the question. The wearable device has a built-in classifier that has learned the linguistic and speech characteristics of normal subjects and patients with mild cognitive impairment (MCI) based on previously presented questions. This method allows for instant confirmation of the subject's test results without using the cloud. Furthermore, there is no need to transmit the subject's voice data to the cloud.

[0094] The foregoing embodiments of the present disclosure will now be described, by way of example only, with reference to the accompanying drawings.

[0095] Fig. 1 is a diagram illustrating an example of the configuration of a cognitive ability assessment system according to an embodiment of the present disclosure. Fig. 2 to Fig. 5 are diagrams illustrating an example of a method by which the cognitive ability assessment system according to an embodiment of the present disclosure performs cognitive ability assessment.

[0096] Referring to FIG. 1, a cognitive ability assessment system according to an embodiment of the present disclosure may include, as an example, a speech acquisition device 200 and a cognitive ability assessment device 300.

[0097] The speech acquisition device 200 can acquire speech uttered by the user 100 (or a subject). The speech acquisition device 200 can output questions to guide the user 100 to speak.

[0098] The speech capture device 200 may be realized in a wearable form or may be implemented in the form of a movable speaker, but is not limited thereto.

[0099] The cognitive ability assessment device 300 communicates with the speech acquisition device 200 via a wired or wireless method, and can provide tasks to the user 100 via the speech acquisition device 200. In addition, the cognitive ability assessment device 300 can receive speech data based on the voice of the user 100 acquired by the speech acquisition device 200.

[0100] 1 exemplarily illustrates a configuration in which a cognitive ability assessment device 300 performs an assessment on a user 100 via an utterance acquisition device 200. In some cases, the cognitive ability assessment device 300 can directly provide a task to the user 100 and receive utterance data of the user 100.

[0101] The cognitive ability assessment device 300 may be realized in the form of, but is not limited to, a mobile terminal or a server. In some cases, some of the components of the cognitive ability assessment device 300 may be included in the terminal, and the remaining components may be included in the server.

[0102] The cognitive ability assessment device 300 may include, for example, a task providing unit 310 , a speech data collecting unit 320 , a cognitive ability measuring unit 330 , and a control unit 340 .

[0103] The task providing unit 310 can provide a task for evaluating the user's cognitive ability. The speech data collecting unit 320 can acquire speech data based on the user's voice corresponding to their response to the task. The cognitive ability measuring unit 330 can calculate an evaluation result of the user's cognitive ability based on the speech data. The control unit 340 can control the operation of each component included in the cognitive ability assessment device 300.

[0104] The task providing unit 310 can select tasks to be provided to the user 100 during a test period for evaluating the cognitive ability of the user 100, and provide the selected tasks to the user 100 via the speech acquisition device 200.

[0105] The task providing unit 310 can select tasks to provide to the user 100 in different ways based on personal information of the user 100, such as educational background, gender, and age.

[0106] The task providing unit 310 can provide, as a task, a question generated based on information recognized by the user 100. The information recognized by the user 100 may be information that the user 100 personally recognizes, or may be information that is commonly recognized by the general public.

[0107] The information personally recognized by the user 100 can be acquired via a device used by the user 100, such as a smart device. The information commonly recognized by ordinary people can be, for example, a widely known fairy tale. The commonly recognized information can be used to generate a task after first confirming whether the user 100 is aware of it.

[0108] The task provider 310 can generate a task that includes questions for assessing the cognitive ability of the user 100 .

[0109] The task providing unit 310 can provide basic questions used in cognitive ability assessment of the user 100. The task providing unit 310 can provide questions that require the user 100 to make multiple judgments simultaneously. As an example, the service providing unit 310 can provide a task that requires judgments on two or more elements simultaneously to solve the task through content that is expressed by pictures, colors, etc. in addition to the content of the text.

[0110] The task providing unit 310 may change some of the content from the information recognized by the user 100 and provide the information to the user 100, and then provide a question generated from the changed recognition information to the user 100. As an example, the task providing unit 310 may change some of the words from the information recognized by the user 100. The task providing unit 310 may change words corresponding to specific parts of speech (e.g., nouns, pronouns, particles, etc.). By providing the user 100 with a question based on information whose content has been slightly changed from the information recognized by the user 100, the task providing unit 310 can confirm whether the user 100 recognizes the changed content and provides an accurate answer.

[0111] The task providing unit 310 can provide voice questions to the user 100. In some cases, the task providing unit 310 may provide an image to be displayed on the screen or a task that requests an action from the user 100.

[0112] The task providing unit 310 can induce the user 100 to speak through the provided task.

[0113] The task providing unit 310 may provide a task that guides the user 100 to make a specific utterance. For example, the task providing unit 310 may provide a task that guides the user 100 to pronounce a voiced sound or a plosive sound. The task providing unit 310 may provide a task that guides the user 100 to repeat a specific utterance for a certain period of time. In this manner, a variety of tasks that can guide pronunciation that can be distinguished between a normal person and a person with mild cognitive impairment can be provided to the user 100.

[0114] Furthermore, the task providing unit 310 may adjust the task to be provided subsequently in accordance with the answer of the user 100. By adjusting the task in accordance with the answer of the user 100, it is possible to continuously guide the user 100 in speaking and acquire speech data for evaluating cognitive ability. Furthermore, in order to maintain the user 100's level of concentration on the task, the task providing unit 310 can output sounds, images, etc. that guide the user 100's concentration at regular intervals while the task is being provided. This makes it possible to maintain the user 100's level of concentration at or above a certain level while the task is being performed.

[0115] When an utterance is generated in accordance with the user's 100 response to the task, the utterance data collection unit 320 can acquire the voice generated by the utterance.

[0116] The utterance data collection unit 320 can receive utterance data acquired from a user via the utterance acquisition device 200.

[0117] When the speech acquisition device 200 acquires the speech of the user 100, it can convert the speech into digital speech data. The speech acquisition device 200 can convert the acquired speech directly into digital speech data and transmit it to the speech data collection unit 320.

[0118] Alternatively, the speech capture device 200 may convert the captured speech into digital speech data after adjusting the captured speech in a manner that can improve the performance of cognitive ability assessment. As an example, the speech capture device 200 may remove noise from the captured speech before converting it into speech data. As another example, the speech capture device 200 may amplify at least a portion of the captured speech before converting it into speech data.

[0119] The noise removal or the sound amplification may be performed in sections during the period when the task is provided by the task providing unit 310 and the user 100 is answering the task.

[0120] For example, referring to FIG. 2, the task providing unit 310 can provide tasks for a test period divided into a first test period T1 and a second test period T2.

[0121] The first test period T1 and the second test period T2 may be periods that are separated in time, and the first test period T1 may be a period before the second test period T2.

[0122] The first test period T1 may be, for example, a period during which a preparatory task for the main test is performed before the main test for cognitive ability assessment is performed, and the second test period T2 may be, for example, a period during which the main test for cognitive ability assessment is performed.

[0123] For example, the first testing period T1 may be a period during which guidance to the main test is provided before the main test is administered in the second testing period T2. Alternatively, the first testing period T1 may be a period during which a pre-test is administered before the second testing period T2. Alternatively, the first testing period T1 may be a period during which a test or task for a purpose different from the test for cognitive ability assessment is administered simultaneously.

[0124] As an example, a guide for the test may be provided during the first test period T1. When the guide is provided during the first test period T1, there is a possibility that the user 100 will not produce any speech.

[0125] The speech acquisition device 200 can acquire noise occurring around the user 100 during the first test period T1. The speech acquisition device 200 can analyze the frequency band of the noise acquired during the first test period T1. The speech acquisition device 200 can remove noise occurring during the second test period T2 in which the main test is performed, based on the frequency band of the noise acquired during the first test period T1.

[0126] For example, a main test for cognitive ability assessment can be executed during the second test period T2. The user 100 can provide answers to tasks provided according to the main test during the second test period T2. The speech capture device 200 can capture speech in response to the answers of the user 100 during the second test period T2. A sound in which the voice of the user 100 and ambient noise are mixed can be captured during the second test period T2.

[0127] Since the speech capture device 200 has confirmed the frequency band of noise around the user 100 during the first test period T1, it can remove noise from the sound obtained during the second test period T2 through analysis of the frequency band of the sound acquired during the second test period T2. The speech capture device 200 can convert the noise-removed voice into digital speech data and transmit it to the speech data collection unit 320. The cognitive ability assessment device 300 can acquire speech data based on the voice from which noise has been removed by an external device.

[0128] Furthermore, the utterance capture device 200 can adjust the volume of at least a part of the voice of the user 100 captured during the second test period T2, and then convert it into utterance data.

[0129] 3, noise around the user 100 can be acquired during a first test period T1. The speech acquisition device 200 can confirm the frequency band of the noise based on the noise acquired during the first test period T1.

[0130] During a second test period T2, the voice of the user 100 can be acquired. Based on the frequency band of the noise identified during the first test period T1, noise can be removed from the sound obtained during the second test period T2.

[0131] The speech capture device 200 can adjust the volume of at least a part of the speech before converting the speech of the user 100 captured in the second test period T2 into speech data in digital format.

[0132] As an example, if the voice of the user 100 acquired during the second test period T2 is referred to as the base voice, the speech acquisition device 200 can generate test voice in which the volume of at least a part of the base voice is adjusted. As an example, the speech acquisition device 200 can amplify the volume of the first part P1 of the base voice. The speech acquisition device 200 does not need to amplify the volume of the second part P2 or the third part P3, which are parts of the base voice other than the first part P1.

[0133] A cognitive ability assessment system according to an embodiment of the present disclosure can assess the cognitive ability of a user 100 through speech analysis of the user 100. Low-volume speech that can be removed as noise may also be necessary for utterance analysis. Since noise is removed from the sound obtained in the second test period T2 based on the frequency band of noise analyzed during the first test period T1, the remaining sound can be seen as the user's 100 speech. By amplifying the volume of a portion of the user's 100 speech, analysis can be performed using speech data based on the user's 100 overall speech acquired during the second test period T2.

[0134] The speech acquisition device 200 may amplify a portion of the user's voice acquired during the second test period T2, where the volume of the voice is equal to or less than the preset volume. Alternatively, the speech acquisition device 200 may amplify a portion of the voice where the length of the portion where the volume of the voice is equal to or less than the preset volume is equal to or greater than the preset length. Examples in which the speech acquisition device 200 amplifies a portion of the user's 100 voice are not limited to these. Any method of amplifying the volume of at least a portion of the user's 100 voice acquired during the second test period T2 to enable speech analysis by the cognitive ability assessment device 300 may be included in embodiments of the present disclosure.

[0135] As an example, referring to Figure 4, a compression amplification technique (WDRC) is illustrated. Figure 4 shows an example in which an original sound is partially amplified and provided. In Figure 4, the horizontal axis represents time, and the vertical axis represents the volume of the sound.

[0136] As shown in the example in Figure 4, the original sound may contain both very quiet parts and parts that are loud and clean enough to recognize the speech. Sometimes the first part may be very quiet and the last part may be loud. In everyday life, quiet sounds like the first part may not be recognized by an actual speech recognizer and may not be usable as information.

[0137] To compensate for this, for example, different gains can be applied depending on the part of the audio, amplifying parts of the audio. As shown in the example in Figure 4, a gain of about 25 dB can be applied in quiet sections, and almost no amplification can be applied in loud sections.

[0138] The loud sounds are recognized as they are, and the quiet sounds are amplified to an appropriate volume, resulting in good voice recognition. The amplification method can simply amplify the entire sound, but it can also select and amplify a specific frequency band.

[0139] In this way, amplification makes small sounds audible. In particular, when noise is amplified and noise reduction is performed, sounds that were not actually heard become audible, allowing you to fully obtain the necessary information you want to obtain through audio.

[0140] In this way, the speech capture device 200 can provide the user 100 with time to prepare for the test through the first test period T1, and at the same time perform a process that can remove noise from the user's 100 voice.

[0141] In some cases, speech data corresponding to the speech of the user 100 may also be acquired during the first test period T1. The speech data corresponding to the speech of the user 100 acquired during the first test period T1 may or may not be used for evaluating the cognitive ability of the user 100.

[0142] 5, during the first test period T1, a guide for the actual test or a preliminary test may be provided to the user 100. In some cases, depending on the answers of the user 100 obtained during the first test period T1, it may be possible to determine whether or not to administer the actual test during the second test period T2, the difficulty level of the actual test, etc.

[0143] Speech by the user 100 in response to the task provided during the first test period T1 can be acquired.

[0144] The speech acquisition device 200 can generate first speech data based on speech acquired in a first test period T1. The speech acquisition device 200 can generate second speech data based on speech acquired in a second test period T2.

[0145] The utterance capturing device 200 can transmit the first utterance data and the second utterance data to the utterance data collecting unit 320.

[0146] The cognitive ability assessment device 300 can calculate an assessment result regarding the cognitive ability of the user 100 based on the second speech data based on the speech acquired in the second test period T2. In some cases, the cognitive ability assessment device 300 may use the first speech data based on the speech acquired in the first test period T1 to calculate an assessment result regarding the cognitive ability of the user 100.

[0147] As an example, when the cognitive ability assessment device 300 calculates an evaluation result regarding the cognitive ability of the user 100 based on the second speech data, if the data necessary to calculate the evaluation result is insufficient, the cognitive ability assessment device 300 can calculate the evaluation result regarding the cognitive ability of the user 100, including the first speech data.

[0148] Alternatively, the cognitive ability assessment device 300 may calculate an assessment result regarding the cognitive ability of the user 100 by including both the first utterance data and the second utterance data.

[0149] When calculating the assessment result regarding the cognitive ability of the user 100, the cognitive ability assessment device 300 can reflect the first utterance data and the second utterance data at an equal level.

[0150] Alternatively, in some cases, the first test period T1 can be considered a period during which the reliability of the test results for the user 100 is lower than that of the second test period T2, and therefore different weights may be applied to the first utterance data and the second utterance data to calculate an evaluation result regarding cognitive ability. As an example, the evaluation result regarding the cognitive ability of the user 100 can be calculated by applying a first weight to the first utterance data and a second weight to the second utterance data. The first weight may be smaller than the second weight.

[0151] In this way, speech data based on the speech of the user 100 acquired during the first test period T1 can also be used to calculate the evaluation result regarding cognitive ability.

[0152] Furthermore, even if the speech of user 100 is acquired in both the first test period T1 and the second test period T2, the speech acquisition device 200 can remove noise from the speech of user 100 acquired in each test period or adjust the volume of the speech.

[0153] As an example, the speech capture device 200 can capture speech from the user 100 during a first test period T1. The speech capture device 200 can identify the frequency band of the user 100's speech by analyzing the frequency band of the user 100's speech captured during the first test period T1.

[0154] The speech capture device 200 can remove noise corresponding to a frequency band other than the frequency band of the voice of the user 100 from the sounds captured in the first test period T1 and the second test period T2. The speech capture device 200 can amplify the volume of a part of the voice of the user 100 from which the noise has been removed.

[0155] Therefore, even if the speech of user 100 is acquired during the first test period T1 and the second test period T2, speech data based on the speech with noise removed and some parts amplified can be provided.

[0156] The speech data transmitted by the speech capture device 200 may be provided to the cognitive ability measurement unit 330 .

[0157] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 by analyzing the speech data.

[0158] The cognitive ability measurement section 330 can calculate an evaluation result regarding the cognitive ability of the user 100 by using at least one of various elements included in the speech data of the user 100.

[0159] As an example, the cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on the number of words used by the user 100 during a previously set period included in the utterance data. The cognitive ability measurement unit 330 may also calculate an evaluation result regarding the cognitive ability of the user 100 based on the number of words used in each of a plurality of sections included in the previously set period.

[0160] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on the types of words used by the user 100 during a predetermined period included in the speech data. The types of words may refer to the parts of speech of the words. The evaluation result regarding the cognitive ability of the user 100 can be calculated based on the frequency with which words of a particular part of speech are used.

[0161] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on the number of overlapping words used by the user 100 during a pre-set period included in the speech data. The number of overlapping words may be counted by dividing it into multiple sections included in the pre-set period. The evaluation result regarding the cognitive ability of the user 100 can be calculated based on the number of overlapping words used in each section.

[0162] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on the number of words used by the user 100 for each section of a pre-set period included in the speech data.

[0163] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on the speech periods and pause periods included in the pre-set period of the speech data. The pause period may mean a period during which no speech occurs. Alternatively, the pause period may mean a period of a certain length of time or more between speech periods.

[0164] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on a rate of change in the length of at least one of an utterance period or a pause period included in a period that has already been set in the utterance data. As an example, the cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on a rate of change in the length of a pause period. Alternatively, the cognitive ability measurement unit 330 may calculate an evaluation result regarding the cognitive ability of the user 100 based on a rate of change in the length of a pause period in each of a plurality of sections included in a period that has already been set.

[0165] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on the amount of change in the ratio between speech periods and pause periods included in a previously set period of the speech data. The ratio between speech periods and pause periods can mean the ratio between speech periods and pause periods in one section in which the speech periods and pause periods are repeated. Alternatively, it can mean the ratio between speech periods and pause periods in each of multiple sections included in the previously set period. The evaluation result regarding the cognitive ability of the user 100 can be calculated based on the uniformity of the ratio of pause periods.

[0166] The cognitive ability measurement section 330 can calculate an evaluation result regarding the cognitive ability of the user 100 based on one or a combination of two or more of the above examples.

[0167] The cognitive ability measurement unit 330 can provide the user 100 with an evaluation result regarding the cognitive ability of the user.

[0168] The cognitive ability measurement unit 330 may present the evaluation result regarding cognitive ability as a score, or may present it in two or more grades, or may provide it with or without a warning regarding mild cognitive impairment.

[0169] The cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 by comparing the analysis result of the speech data of the user 100 with past data of the user 100. Alternatively, the cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 by comparing the analysis result of the speech data of the user 100 with average data of other users other than the user 100. Alternatively, the cognitive ability measurement unit 330 can calculate an evaluation result regarding the cognitive ability of the user 100 by comparing the analysis result of the speech data of the user 100 with data derived from patients with mild cognitive impairment or dementia. In some cases, the cognitive ability measurement unit 330 may calculate an evaluation result regarding the cognitive ability of the user 100 by combining two or more of the above-mentioned comparison methods.

[0170] Thus, according to an embodiment of the present disclosure, the speech acquisition device 200 and the cognitive ability assessment device 300 can easily assess the cognitive ability of the user 100, and if there is a risk of mild cognitive impairment, provide the user 100 with relevant information, thereby enabling early detection of mild cognitive impairment or dementia.

[0171] As mentioned above, the cognitive ability assessment device 300 can be realized in various forms such as a terminal or a server, and in some cases, the configuration of the cognitive ability assessment device 300 may be distributed between a terminal and a server to form a cognitive ability assessment system.

[0172] Fig. 6 is a diagram schematically illustrating another example of the configuration of a cognitive ability assessment system according to an embodiment of the present disclosure. Fig. 7 and Fig. 8 are diagrams exemplarily illustrating the configuration of the utterance acquisition device 200 shown in Fig. 6.

[0173] Referring to FIG. 6, a cognitive ability assessment system according to an embodiment of the present disclosure can be configured from an utterance acquisition device 200, a mobile terminal 400, and a cognitive ability assessment management server 500.

[0174] The speech capture device 200 may include, for example, an output module 210 , an input module 220 , a processing module 230 , and a communication module 240 .

[0175] The output module 210 can output audio corresponding to a plurality of tasks provided to the user 100. . Out The force module 210 may provide task-related audio that is output by adjusting the volume of the audio, as in the example above, taking into account the hearing ability of the user 100 .

[0176] The input module 220 can acquire speech in response to the user's 100 responses to the tasks provided by the output module 210 .

[0177] The processing module 230 can convert the audio captured by the input module 220 into speech data in a digital format. The processing module 230 can adjust the volume of at least a portion of the captured audio before converting it into speech data.

[0178] Through frequency band analysis, the processing module 230 can remove noise from the audio acquired by the input module 220. Through frequency band analysis, the processing module 230 can select a portion to amplify from the audio acquired by the input module 220.

[0179] The communication module 240 can transmit the speech data generated by the processing module 230 to the mobile terminal 400 .

[0180] The speech acquisition device 200 can be implemented in various forms such as a speaker, etc. In some cases, the speech acquisition device 200 may be implemented in the form of a wearable device.

[0181] In one example, the speech capture device 200 can be implemented in the form of a wearable device including an earphone 201, a microphone 202, a main body 203, and a connection unit 204. The microphone 202 can be attached to the main body 203. The earphone 201 can be coupled to the main body 203 via the connection unit 204.

[0182] The main body 203 may be configured to be wrapped around the neck of the user 100. When the main body 203 of the speech capture device 200 is wrapped around the neck of the user 100 and worn by the user 100, it is possible to provide a task to the user 100 and capture the speech of the user 100.

[0183] In some cases, the microphone 202 may be mounted in close contact with the neck of the user 100 and may be implemented to capture sound corresponding to the speech of the user 100 by detecting vibrations caused by the speech of the user 100. Even if the speech ability of the user 100 has declined, it is possible to evaluate the cognitive ability of the user 100 by analyzing the speech data of the user 100.

[0184] The earphones 201 may be provided in a state connected to the main body 203 via a connection unit 204. This can provide convenience for the user 100. In some cases, the earphones 201 may be implemented in a form that adds a function of measuring the brain waves of the user 100.

[0185] Furthermore, the speech capture device 200 can provide a hearing aid function that compensates for the hearing loss so that the user 100 can hear sounds more easily.

[0186] 7, the configuration of the speech capture device 200 may be exemplified by the input module 220, which may include multiple microphones. The input module 220 may use a directional microphone or apply an algorithm to more efficiently capture audio output from an external device such as a smart device. While an omnidirectional microphone may be used in some cases, using a directional microphone may improve noise reduction and better capture the speaker's voice.

[0187] The processing module 230 may convert and process the input audio into the frequency domain to provide amplified and noise-reduced audio. For example, the input audio may include environmental noise depending on the current location of the user 100. Such environmental noise may include characteristics of various outdoor environments, such as an office environment, a subway, or a factory environment. The characteristics of the environmental noise may be classified using information obtained by fast Fourier transforming the input signal, as shown in the example of FIG. 7. Furthermore, the fast Fourier transformed signal may be amplified by varying the amplification gain depending on the frequency domain. This amplification may increase the volume of soft sounds, thereby enabling users 100 with poor hearing (e.g., elderly people) to hear clearer sounds. The amplification method may be the same as the example described with reference to FIG. 4. The processing module 230 may also perform a function of removing environmental noise. This may amplify the audio frequency band while removing noise from the received audio. If necessary, the frequency-converted signal may be subjected to spectral enhancement.

[0188] The noise reduction and audio amplification functions of the processing module 230 may be controlled by a remote control device (e.g., a smartphone, a PC, etc.) via the communication module 240. The compression amplification size for each frequency band, the audio-to-noise ratio for noise reduction, and the inversion suppression parameters can be remotely controlled.

[0189] Furthermore, the speech capture device 200 can provide a speech processing function for improving the speech recognition rate of the user 100.

[0190] Referring to FIG. 8 , the configuration of the speech capture device 200 can be exemplified. The input module 220 can include at least one microphone. To efficiently capture the speech of the user 100, the input module 220 can include a directional microphone or can use multiple microphones and apply a directional algorithm. In some cases, an omnidirectional microphone can be included, but a directional microphone can improve noise reduction effects and better receive the speaker's voice. The input module 220 can transmit the sound captured via the microphone to the processing module 230 after initial processing. For example, the input module 220 can remove noise from the sound captured via the microphone or amplify at least a portion of the captured sound to transmit clearer sound to the processing module 230.

[0191] The processing module 230 may remove noise from a noisy voice input from a microphone, or may analyze environmental noise to extract information based on the classification of the external environment in which the speaker is located. The processing module 230 may perform a function of amplifying a soft voice of the speaker. The processing module 230 may also pass the voice signal without performing additional signal processing.

[0192] The processing module 230 may perform a fast Fourier transform on the signal input via the microphone. The frequency characteristics of the converted signal may be analyzed to provide information about the environment in which the speaker is located. For example, the environment in which the speaker is located may be analyzed as a quiet office environment, a subway environment, an external environment, etc. This information may be used to remove environmental noise other than the speaker's voice information input via the microphone.

[0193] In addition, the processing module 230 amplifies quiet speech, such as interjections or murmurs, input from the fast Fourier transformed signal via the microphone, and speech that is expressed softly depending on the speaker's characteristics, thereby improving the recognition rate of the speech recognizer. Since amplifying quiet speech also amplifies noise contained in the speech, reducing the speech recognition rate, it may be desirable to perform noise removal.

[0194] The processing module 230 may process and digitize analog sound received from the input module 220. For example, the processing module 230 may classify the acquired sound and either pass the entire sound or delay it. The processing module 230 may convert the acquired sound into a frequency spectrum and perform a process of amplifying a portion of the converted sound (dynamic range compression (WDRC)) and a process of removing noise (noise suppression). The processing module 230 may further include a selective sound amplification component to further amplify specific sounds. After completing the sound processing, the processing module 230 may convert the sound back into the time domain and output it. The output sound may be provided as feedback and reflected in the sound output by the processing module 230. The processing module 230 may provide the digitized audio data to the communication module 240.

[0195] When the communication module 240 receives the digitized voice data, i.e., speech data, it can transmit the speech data to a mobile terminal 400 such as a smart device.

[0196] The mobile terminal 400 may include a task provision module 410 , a speech data collection module 420 , and a cognitive performance measurement module 430 .

[0197] The task providing module 410 can select a task to be provided to the user 100 via the speech acquisition device 200, and provide the selected task according to the test period to the user 100 via the speech acquisition device 200. The task providing module 410 may be arranged in the cognitive ability assessment management server 500 in some cases.

[0198] The speech data collection module 420 can acquire the speech data transmitted by the speech acquisition device 200 .

[0199] The speech data collection module 420 can receive speech data based on a test voice in which noise has been removed from a base voice corresponding to the answer of the user 100 and the volume of at least a part of the base voice has been adjusted.

[0200] The speech data collection module 420 may receive speech data that has been digitally processed by the speech capture device 200. In some cases, the speech data collection module 420 may receive analog audio from the speech capture device 200 and process the received audio as digital audio data.

[0201] The cognitive ability measurement module 430 can calculate an assessment result regarding the cognitive ability of the user 100 based on the speech data.

[0202] The cognitive ability measurement module 430 can analyze the speech data acquired by the speech data collection module 420 and calculate an evaluation result regarding the cognitive ability of the user 100. The method by which the cognitive ability measurement module 430 calculates the evaluation result regarding the cognitive ability of the user 100 can be similar to the method performed by the cognitive ability measurement unit 330 included in the cognitive ability assessment device 300 described above.

[0203] The cognitive ability measurement module 430 may be located in the cognitive ability assessment management server 500 in some cases. That is, the cognitive ability evaluation may be performed by the cognitive ability evaluation management server 500. In this case, the evaluation result may be transmitted from the cognitive ability evaluation management server 500 to the mobile terminal 400.

[0204] 6 , when the task providing module 410, the speech data collecting module 420, and the cognitive ability measuring module 430 are included in the mobile terminal 400, the cognitive ability assessment managing server 500 can provide a function of updating tasks that can be provided to the user 100, or a function of storing and managing assessment results related to the cognitive ability of the user 100. In some cases, some of the functions executed by the mobile terminal 400 may be executed by the cognitive ability assessment managing server 500.

[0205] In this way, it is possible to provide a system in which the mobile terminal 400 and the cognitive ability evaluation management server 500 communicate via the network 600, and the cognitive ability test of the user 100 can be easily carried out.

[0206] In addition, in some cases, an embodiment of the present disclosure can obtain the electroencephalogram (EEG) of user 100 appearing during a test period, a speech period, or a period other than a speech period, along with the speech data of user 100, and calculate an evaluation result regarding the cognitive ability of user 100.

[0207] Fig. 9 is a diagram schematically illustrating yet another example of the configuration of a cognitive ability assessment system according to an embodiment of the present disclosure. Fig. 10 and Fig. 11 are diagrams illustrating an example of a method for performing cognitive ability assessment by a cognitive ability assessment system according to an embodiment of the present disclosure.

[0208] Referring to FIG. 9, a cognitive ability assessment system according to an embodiment of the present disclosure can be configured from an utterance acquisition device 200, a mobile terminal 400, and a cognitive ability assessment management server 500.

[0209] The speech acquisition device 200 may include an output module 210, a processing module 230, and a communication module 240, similar to the example described with reference to Fig. 6. The speech acquisition device 200 may include an utterance input module 221 and an electroencephalogram (EEG) input module 222. The speech acquisition device 200 can acquire audio uttered by the user 100 and acquire the electroencephalogram of the user 100. The speech acquisition device 200 can acquire the electroencephalogram of the user 100 via a part connected to the head or ear of the user 100.

[0210] The utterance capturing device 200 can capture the brain waves of the user 100 during at least a part of the period during which the voice uttered by the user 100 is captured. Alternatively, the utterance capturing device 200 may capture the brain waves of the user 100 during a period distinct from the period during which the voice uttered by the user 100 is captured.

[0211] The speech input module 221 and the brain wave input module 222 can acquire the voice and brain waves of the user 100. The communication module 240 can transmit, to the mobile terminal 400, speech data corresponding to the voice of the user 100 and brain wave data corresponding to the brain waves.

[0212] The mobile terminal 400 may include a task providing module 410, a speech data collecting module 420, and a cognitive ability measuring module 430, and may further include an electroencephalogram data collecting module 440.

[0213] The mobile terminal 400 can use the acquired electroencephalogram data in various ways in the process of calculating an evaluation result regarding the cognitive ability of the user 100 based on the speech data.

[0214] As an example, referring to FIG. 10, tasks may be provided during a test period including a first test period T1 and a second test period T2, and a test may be performed to assess the cognitive ability of the user 100.

[0215] In the second test period T2, a main test can be administered to assess the cognitive abilities of the user 100. During the first test period T1, instructions for the main test can be provided.

[0216] During a first test period T1, first electroencephalogram data can be acquired. During a second test period T2, second electroencephalogram data can be acquired.

[0217] The first electroencephalogram data may be data obtained from electroencephalograms related to the comprehension level of the user 100. The task providing module 410 or the cognitive ability measuring module 430 can confirm the comprehension level of the user 100 regarding the guidance to the main test based on the first electroencephalogram data. Based on the first electroencephalogram data, it can determine whether or not to administer the main test in the second test period T2, or the difficulty level, type, etc. of the main test.

[0218] The second electroencephalogram data may be data obtained from electroencephalograms related to the speech ability of the user 100. The cognitive ability measurement module 430 can calculate an evaluation result regarding the cognitive ability of the user 100 using the analysis result of the second electroencephalogram data together with the analysis result of the speech data.

[0219] In this way, when the speech acquisition device 200 is realized in a form that can simultaneously measure the brain waves of the user 100, the brain wave data collected during the first test period T1 and the second test period T2 can be used to determine the test method or calculate the evaluation results regarding cognitive ability.

[0220] Furthermore, if a test for cognitive ability assessment is conducted during the first test period T1, the electroencephalogram data acquired during the first test period T1 can also be used to calculate an evaluation result regarding the cognitive ability of the user 100. In this case, similar to the speech data, by applying different weights to the electroencephalogram data acquired during the first test period T1 and the second test period T2, the data can be used to calculate an evaluation result regarding the cognitive ability.

[0221] Alternatively, different abilities of the user 100 can be measured via electroencephalogram data and used to calculate an assessment result regarding the cognitive ability of the user 100. As described above, the cognitive ability assessment may be performed on the mobile terminal 400, or the cognitive ability assessment may be performed on the cognitive ability assessment management server 500. When the cognitive ability assessment is performed on the cognitive ability assessment management server 500, the assessment results may be transmitted from the cognitive ability evaluation management server 500 to the mobile terminal 400.

[0222] As an example, referring to FIG. 11, the speech capture device 200 can capture the speech of the user 100 and simultaneously capture the brain waves of the user 100.

[0223] The speech capture device 200 can simultaneously capture multiple types of electroencephalograms.

[0224] As an example, the speech capture device 200 can capture electroencephalograms related to the hearing ability of the user 100 during at least a part of the first test period T1 and the second test period T2.

[0225] Furthermore, the speech acquisition device 200 can acquire electroencephalograms related to the speech ability of the user 100 during at least a part of the first test period T1 and the second test period T2.

[0226] The mobile terminal 400 can calculate an evaluation result regarding the cognitive ability of the user 100 using the electroencephalogram data regarding the hearing ability acquired during the first test period T1 and the second test period T2. The mobile terminal 400 can calculate an evaluation result regarding the cognitive ability of the user 100 using the electroencephalogram data regarding the speaking ability acquired during the first test period T1 and the second test period T2.

[0227] If user 100's auditory activity is mainly performed during the first test period T1 and user 100's speech activity is mainly performed during the second test period T2, the evaluation results for user 100's cognitive ability can be calculated using EEG data regarding user 100's auditory ability acquired during the first test period T1 and EEG data regarding user 100's speech ability acquired during the second test period T2.

[0228] Alternatively, both the electroencephalogram data relating to hearing ability and the electroencephalogram data relating to speaking ability obtained during the first test period T1 and the second test period T2 can be used to calculate an evaluation result relating to the cognitive ability of the user 100. In this case, depending on the case, different weights can be applied to the electroencephalogram data for each test period to be used in calculating the evaluation result.

[0229] The mobile terminal 400 can calculate an evaluation result regarding the cognitive ability of the user 100 by using both the speech data and the electroencephalogram data acquired from the user 100.

[0230] A decline in cognitive ability may result in a decline in speech ability, a decline in hearing ability, or a simultaneous decline in speech and hearing ability. Therefore, a decline in the cognitive ability of user 100 can be measured by testing speech and hearing ability.

[0231] In some cases, the degree of decline in speech ability may differ from the degree of decline in hearing ability. For example, in a growing person, the rate of development of hearing ability may be faster than the rate of development of speech ability. On the other hand, in an elderly person, the rate of decline in speech ability may be faster than the rate of decline in hearing ability. Alternatively, unlike the developmental process, the rate of decline in both abilities may be similar. Alternatively, during decline, the rate of decline in hearing ability may be faster.

[0232] In this way, the state of cognitive ability can be determined based on the presence or absence of a decline in hearing ability. In some cases, the state of cognitive ability can be divided into various levels based on the difference between hearing ability and speaking ability, and an evaluation result regarding the cognitive ability of user 100 can be calculated. Furthermore, when considering whether or not hearing ability declines when determining the state of cognitive ability, data according to the level of hearing aid function provided by speech acquisition device 200 can also be used. The evaluation result regarding cognitive ability can also be calculated by reflecting data regarding the degree of decline in hearing ability according to the level of hearing aid function.

[0233] Since electroencephalogram data is used together with speech data, the accuracy of cognitive ability assessment by the cognitive ability assessment system according to the embodiment of the present disclosure can be improved.

[0234] 12 and 13 are flowcharts illustrating exemplary steps of a cognitive ability assessment method according to an embodiment of the present disclosure.

[0235] 12, the cognitive ability assessment system allows the user 100 to select a task to be provided to the user 100 (S1200). The user 100 can select the difficulty level, type, and content of questions included in the task to be provided.

[0236] A task can be provided by a cognitive ability assessment system (S1210).

[0237] An answer from the user 100 to the task is generated, and utterance data from the user 100 can be obtained (S1220).

[0238] The cognitive ability assessment system may analyze linguistic characteristics of the speech data (S1230). The cognitive ability assessment system may provide an assessment result of the cognitive ability of the user 100 based on the analysis of the linguistic characteristics of the speech data (S1240).

[0239] Alternatively, in some cases, the cognitive ability assessment system may provide an assessment of the cognitive ability of the user 100 based on linguistic and acoustic characteristics of the speech data.

[0240] The linguistic characteristics of the speech data may refer to the characteristics of vocabulary such as sentences and words spoken by the user 100. The acoustic characteristics of the speech data may refer to the characteristics related to the intensity, frequency, timbre, etc. of the voice spoken by the user 100.

[0241] Using the linguistic and acoustic characteristics of speech produced by normal individuals, mild cognitive impairment patients, dementia patients, etc., it is possible to calculate an evaluation result regarding the cognitive ability of user 100. In addition, based on the acoustic characteristics of the speech data, it is also possible to distinguish between diseases other than mild cognitive impairment, such as depression, and calculate an evaluation result regarding the cognitive ability of user 100.

[0242] The linguistic characteristics extracted through the speech data of the user 100 can be compared with various previously stored reference data to calculate an evaluation result regarding the cognitive ability of the user 100.

[0243] 13, the cognitive ability assessment system can extract linguistic features of the speech data (S1300). In some cases, acoustic features of the speech data can also be used, as described above.

[0244] The cognitive ability assessment system can check whether past data exists for the user 100 under test (S1310).

[0245] If past data of user 100 exists, the cognitive ability assessment system can calculate a first evaluation result through a comparison between the speech data of user 100 (S1320). By analyzing the speech data, changes in the speech ability of user 100 can be confirmed, and the first evaluation result can be calculated based on the changes. The cognitive ability assessment system can apply a first weight to the first evaluation result.

[0246] The cognitive ability assessment system can calculate a second evaluation result after comparing the speech data of user 100 with the speech data of other users (S1330). The speech data of other users may be speech data of other users who have taken a test, or may be speech data collected from a user with a disease such as mild cognitive impairment. The cognitive ability assessment system can apply a second weight to the second evaluation result.

[0247] If no past data of the user 100 exists, the cognitive ability assessment system can calculate an assessment result after comparing the speech data of the user 100 with that of other users (S1340).

[0248] The cognitive ability assessment system can provide an assessment result regarding cognitive ability based on the comparison result between the speech data of the person and the comparison result between the speech data of the person and others (S1350).

[0249] The first weight applied to the first evaluation result obtained by comparing the speech data of the person himself / herself and the second weight applied to the second evaluation result obtained by comparing the speech data of the person himself / herself and another person can be set in a variety of ways.

[0250] As an example, based on the age of user 100, the degree of change in cognitive ability, the level of risk, etc., it is possible to give priority to the rate of change in user 100's own cognitive ability, or to give priority to the difference between user 100 and others, thereby making a more detailed judgment as to whether user 100 is at risk of mild cognitive impairment.

[0251] According to the embodiment of the present disclosure, when a wearable device having noise removal and selective amplification functions is used, a language test can be conducted in daily life or in a public place to easily measure the cognitive ability of the user 100. In other words, according to the embodiment of the present disclosure, the user 100 has excellent accessibility and can conveniently conduct the test anytime and anywhere.

[0252] According to the embodiments of the present disclosure, the effectiveness of speech recognition can be improved by applying noise reduction techniques.

[0253] According to an embodiment of the present disclosure, detailed speaker information can be collected through the sound amplification function, thereby enhancing analytical capabilities.

[0254] According to an embodiment of the present disclosure, by selecting multiple-choice questions for each task, it is possible to select an optimal set of test questions taking into consideration the intellectual ability, language cognitive ability, etc. of the individual subject.

[0255] According to embodiments of the present disclosure, physical sensors, time-based question start and end information can be provided to enable question set selection.

[0256] According to the embodiment of the present disclosure, the selective questioning test method according to the individual's condition can shorten the test time and can significantly improve the fatigue level of the elderly and perform the test.

[0257] According to an embodiment of the present disclosure, speech information for each question can be quantified. Multiple questions can be asked for each task, and speech for each question can be quantified and presented as an average. All task-specific information can be combined.

[0258] The above description merely exemplifies the technical concept of the present disclosure, and various modifications and variations may be made by a person skilled in the art without departing from the essential characteristics of the present disclosure. Furthermore, the embodiments described in the present disclosure are for illustrative purposes only and do not limit the technical concept of the present disclosure. The scope of protection of the present disclosure should be interpreted by the scope of the following claims, and all technical concepts within the scope equivalent thereto should be interpreted as being within the scope of the present disclosure.

[0259] CROSS-REFERENCE TO RELATED APPLICATION

[0260] This patent application claims priority under 35 U.S.C. § 119(a) of Patent Application No. 10-2022-0105435, filed in Korea on August 23, 2022, the entire contents of which are incorporated herein by reference. If this patent application also claims priority from a country other than the United States for the same reason as above, the entire contents of that country will be incorporated herein by reference.

Claims

1. a task providing unit that provides a plurality of tasks to a user during a first test period and a second test period after the first test period; a speech data collection unit that acquires digital speech data converted from a test voice in which the volume of at least a portion of a base voice has been adjusted in accordance with the user's response to at least one of the plurality of tasks; and A cognitive ability assessment device comprising a cognitive ability measurement unit that calculates an assessment result regarding the cognitive ability of the user based on at least one linguistic characteristic of the speech data.

2. The cognitive ability assessment device according to claim 1 , wherein during the first test period, noise around the user is collected, and a noise frequency band is extracted by analyzing the frequency band of the ambient noise.

3. 3. The cognitive ability assessment device of claim 2, wherein during the second test period, the baseline sound is collected, noise is removed from the baseline sound based on the noise frequency band, and then the volume of at least a portion of the baseline sound is adjusted.

4. The cognitive ability assessment device of claim 1 , wherein a first portion of the base sound is amplified and a second portion of the base sound is not amplified.

5. The cognitive ability assessment device according to claim 1 , wherein the volume of at least a part of the basic sound is adjusted by an external device.

6. During the first test period, first speech data based on the baseline speech is acquired, and during the second test period, second speech data based on the baseline speech is acquired; The cognitive ability assessment device of claim 1, wherein a first weight is applied to the first speech data and a second weight is applied to the second speech data, and the cognitive ability assessment device is used to calculate the assessment result regarding the user's cognitive ability.

7. The cognitive ability assessment device according to claim 1 , wherein the plurality of tasks are provided to the user in the form of audio, and the volume of the audio relating to the plurality of tasks is adjusted according to the hearing ability of the user.

8. The cognitive ability measurement unit The cognitive ability assessment device according to claim 1 , wherein the assessment result regarding the cognitive ability of the user is calculated based on the at least one linguistic characteristic of the speech data and the hearing ability of the user.

9. The cognitive ability measurement unit 2. The cognitive ability assessment device of claim 1, wherein the assessment result regarding the user's cognitive ability is calculated based on a first evaluation result calculated by comparing the speech data with the user's past speech data, and a second evaluation result calculated by comparing the speech data with average speech data of other users other than the user.

10. The cognitive ability measurement unit 2. The cognitive ability assessment device according to claim 1, wherein the assessment result regarding the cognitive ability of the user is calculated based on at least one of the number of words used by the user during a previously set period included in the speech data, the types of words used by the user during the previously set period, the number of overlapping words used by the user during the previously set period, the number of words used by the user for each section of the previously set period, speech periods and pause periods included in the previously set period, a rate of change in the length of at least one of the speech periods or the pause periods, or an amount of change in the ratio of the speech periods to the pause periods.

11. The task providing unit The cognitive ability assessment device according to claim 1 , wherein a task is provided to induce an answer in which the content of a part of the information perceived by the user is changed.

12. The task providing unit The cognitive ability assessment device according to claim 1 , wherein the cognitive ability assessment device provides a task generated based on at least one of individual information and common information recognized by the user.

13. The cognitive ability assessment device according to claim 1 , wherein the speech data based on the user's answer to at least one task among the plurality of tasks is not used to calculate the assessment result regarding the user's cognitive ability.

14. The task providing unit The cognitive ability assessment device according to claim 1 , further comprising: adjusting a task provided after the user's answer based on the user's answer.

15. a task provision module that provides a plurality of tasks to the user during the test period; and A mobile terminal including a speech data collection module that acquires digital speech data converted from a test voice in which the volume of at least a portion of a base voice has been adjusted in response to the user's response to at least one of the plurality of tasks.

16. The mobile terminal of claim 15 , further comprising a cognitive ability measurement module that calculates an assessment result regarding the user's cognitive ability based on at least one linguistic characteristic of the speech data.

17. an output module for outputting sounds corresponding to a plurality of tasks provided to the user, the sound volume being adjusted according to the hearing ability of the user; an input module that acquires a basic speech according to the user's response to at least one of the plurality of tasks; and a processing module for providing speech data in digital form based on said base voice;

18. The processing module includes: The speech acquisition device of claim 17, further comprising: analyzing a frequency band of the base speech; adjusting the volume of at least a portion of the base speech through the analysis of the frequency band to generate a test speech; and converting the test speech into the digital speech data.

19. The processing module includes: The speech capture device of claim 17 , further comprising: collecting ambient noise of the user during a first period; and analyzing frequency bands of the ambient noise to extract noise frequency bands.

20. The processing module includes: The speech acquisition device of claim 19, wherein the basic speech is collected during a second period after the first period, noise is removed from the basic speech based on the noise frequency band, and then a volume of at least a portion of the basic speech is adjusted.

Images