Method and apparatus for predicting Alzheimer's disease based on voice characteristics

Abstract

A method and device for predicting Alzheimer's disease based on voice characteristics is provided. The Alzheimer's disease prediction device according to an embodiment includes a voice input unit configured to record a subject's voice to generate a voice sample, a data input unit configured to receive demographic information of the subject, a voice feature extraction unit configured to extract voice features from the generated voice sample, and a prediction model pre-trained to predict whether the subject has Alzheimer's disease based on the voice features and the demographic information.

Description

[Technical field]

[0001] The present invention relates to a method for diagnosing Alzheimer's disease (AD), which is the most common cause of dementia, and a method that helps medical professionals determine whether or not a patient has Alzheimer's disease through a non-invasive method using the patient's voice. It also provides a method and device for predicting the state of dementia caused by Alzheimer's disease through a non-invasive method for people at high risk of dementia. [Background technology]

[0002] The use of AI speakers is gradually increasing along with the fourth industry, but they are not yet widespread among the masses. There are various reasons for the delay in popularizing AI speakers, but the majority of consumers believe that there is a lack of content that utilizes AI speakers. In other words, there is a demand for the development of diverse content that uses AI speakers.

[0003] AI speakers and the healthcare industry are seen as a "win-win" combination that can achieve both profitability and public benefit, and many researchers both at home and abroad are using AI speakers to develop a variety of healthcare-related content.

[0004] The elderly population aged 65 and over in Korea accounted for 7.1% of the total population in 2000, and is expected to exceed 20% by 2026. In the case of dementia, a representative geriatric disease, it was estimated that there were 280,000 cases, or 8.3% of the elderly population aged 65 and over, in 2000, and this number is on the rise, reaching an estimated 430,000 cases, or 8.6% of the total elderly population, in 2010. The prevalence of dementia is showing a continuous upward trend due to the increase in the elderly population and the increase in average life expectancy.

[0005] Alzheimer's disease (AD) is the most common causative disease, accounting for approximately 70% of all dementia cases, and it is estimated that approximately 600,000 people will develop Alzheimer's disease in 2020. Alzheimer's disease is a representative degenerative dementia disease that causes cognitive function and daily life decline due to degeneration of cerebral nerve cells, and is a destructive disease that can lead to death, causing pain not only to patients but also to their families who care for them. As Alzheimer's disease is an irreversible disease and no completely curative treatment has been developed, the current best method is to diagnose Alzheimer's disease early and intervene promptly. As the prevalence of Alzheimer's disease increases, national health and medical expenditures are also being enormously affected.

[0006] Speech and language disorders are one of the symptoms associated with dementia due to cognitive function and neurodegeneration, and are a commonly used criterion for distinguishing between normal aging and dementia-causing diseases, and often appear as markers of early symptoms. Therefore, in automatic speech recognition research, various studies are being conducted to discover speech and language markers for dementia diagnosis. A representative example is research on predicting dementia using data from DementiaBank's Pitt corpus (e.g., Cookie Theft diagram).

[0007] Considering that cerebrospinal fluid tests and positron emission tomography (PET), which are used to accurately diagnose Alzheimer's disease, a major cause of dementia, are restricted by experts in specific spaces and times, there is a demand for methods and devices that can predict whether a person has Alzheimer's disease using AI speakers and smart devices. Summary of the Invention [Problem to be solved by the invention]

[0008] The present invention relates to a method for diagnosing Alzheimer's disease using the acoustic characteristics of voice, and to a method and device that can collect the responses of a conversation partner using an AI speaker or smart device through a voice recognition microphone and predict whether or not the conversation partner has Alzheimer's disease and the possibility of it. [Means for solving the problem]

[0009] An Alzheimer's disease prediction device based on voice characteristics according to an embodiment of the present invention includes a voice input unit configured to record a subject's voice to generate a voice sample, a data input unit configured to receive demographic information of the subject, a voice characteristic extraction unit that extracts voice characteristics from the generated voice sample, and a prediction model pre-trained to predict whether or not the subject has Alzheimer's disease based on the voice characteristics and the demographic information.

[0010] A method for predicting Alzheimer's disease based on voice characteristics according to an embodiment of the present invention includes the steps of recording a subject's voice to generate a voice sample, receiving demographic information of the subject, extracting voice characteristics from the generated voice sample, and inputting the voice characteristics and the demographic information into a pre-trained prediction model to predict whether the subject has Alzheimer's disease.

[0011] A recording medium according to an embodiment of the present invention is a computer-readable recording medium storing computer-readable instructions which, when executed by at least one processor, cause the at least one processor to perform steps including recording a subject's voice to generate a voice sample, receiving demographic information of the subject, extracting voice characteristics from the generated voice sample, and inputting the voice characteristics and the demographic information into a pre-trained prediction model to predict whether the subject has Alzheimer's disease. Effect of the Invention

[0012] The method and device for predicting Alzheimer's disease based on voice characteristics according to an embodiment of the present invention is a method and device for predicting whether a conversation subject has Alzheimer's disease using an AI speaker and a smart device, and is expected to help overcome the time, space, and resource limitations associated with conventional testing, and to assist medical professionals in diagnosing Alzheimer's disease.

[0013] The present invention provides a method for predicting the risk of Alzheimer's disease by utilizing the acoustic characteristics of voice, which can assist in the development of appropriate intervention strategies for the continuously increasing dementia population. [Brief description of the drawings]

[0014] [Figure 1] 1 is a block diagram showing the configuration of an Alzheimer's disease prediction device based on voice characteristics according to an embodiment of the present invention. FIG. [Diagram 2] 1 illustrates an exemplary operation process of an Alzheimer's disease prediction device based on voice characteristics according to an embodiment of the present invention. [Diagram 3] 1 is a block diagram showing the sequence of a method for predicting Alzheimer's disease based on voice characteristics according to an embodiment of the present invention. FIG. [Figure 4] 1 is a graph showing Alzheimer's disease prediction results taking into account demographics and voice characteristics. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, a preferred embodiment of the present invention will be described with reference to the accompanying drawings. The present invention has been described with reference to the embodiment shown in the drawings, but this is only one embodiment, and the technical idea and the core configuration and operation of the present invention are not limited thereto.

[0016] Fig. 1 is a block diagram showing the configuration of an Alzheimer's disease prediction device according to an embodiment of the present invention, Fig. 2 is an exemplary diagram showing the operation process of the Alzheimer's disease prediction device according to an embodiment of the present invention.

[0017] 1 and 2, an Alzheimer's disease prediction device 10 according to an embodiment of the present invention includes a voice input unit 100, a data input unit 110, a voice feature extraction unit 120, a prediction model 130 and a data storage unit 140.

[0018] The Alzheimer's disease prediction device 10 may be entirely hardware, or may have an aspect of being partially hardware and partially software. For example, the Alzheimer's disease prediction device 10 and each unit included therein in this specification may generally refer to a device for exchanging data of a specific format and content in an electronic communication manner and associated software. In this specification, terms such as "unit," "module," "server," "device," "apparatus," or "terminal" are intended to refer to a combination of hardware and software driven by the hardware. For example, the hardware here may be a data processing device including a CPU or other processor. In addition, the software driven by the hardware may refer to a running process, an object, an executable, a thread of execution, a program, etc.

[0019] In addition, each module constituting the Alzheimer's disease prediction device 10 is not necessarily intended to refer to a separate component that is physically separated. In FIG. 1, the voice input unit 100, the data input unit 110, the voice feature extraction unit 120, the prediction model 130, and the data storage unit 140 are shown as separate blocks separated from each other, but this is merely a functional division of the devices constituting the Alzheimer's disease prediction device 10 according to the operations performed by the corresponding devices. Therefore, in some embodiments, the voice input unit 100, the data input unit 110, the voice feature extraction unit 120, the prediction model 130, and the data storage unit 140 may be integrated in part or in whole within a single device. For example, the Alzheimer's disease prediction device 10 may be embodied through a device having voice recognition and data processing capabilities, such as an AI speaker. However, the present invention is not limited to this, and one or more of the above configurations may be embodied as separate devices that are physically separated from other units, and may be components connected to each other so as to be able to communicate with each other in a distributed computing environment.

[0020] The voice input unit 100 generates a voice sample of the subject. The voice input unit 100 may be configured to record the subject's voice at a certain frequency and generate a voice sample. The voice input unit 100 may include a condenser microphone and a device for controlling the same, and can record the subject's voice at a frequency of 16 Hz or more and generate a voice sample. The subject performs at least one of a picture drawing task, a standard sentence speech task, and a story recall task in a quiet room where recording is possible, and speaks naturally or recites aloud, and the voice input unit 100 records the subject's speech and generates a voice sample.

[0021] The data input unit 110 receives demographic information of the subject. The demographic information includes at least the subject's age, sex, and years of education (degree of education). The material input through the data input unit 110 means data from which the subject's age, sex, and years of education can be confirmed or from which such information can be extracted. Exemplarily, the demographic information can be acquired through the subject's interview material. The interview material can include information on the patient's age, sex, and years of education, and corresponds to material generated using the medical judgment of a trained medical specialist, and may be material collected under the supervision of other medical personnel and the patient's guardians. The data input unit 110 can receive demographic information through the interview material described above, but is not limited thereto. In some embodiments, the subject's demographic information can be information directly input by the subject.

[0022] The data inputted through the voice input unit 100 and the data input unit 110 may be stored in the data storage unit 140. The data storage unit 140 may be configured to store the inputted data or to provide a temporary or temporary storage required for data processing of the predictive model described below.

[0023] The voice feature extraction unit 120 can extract the voice features of the patient from the input voice sample of the subject. The voice feature extraction unit 120 can extract voice features related to the phonological features, root features, and spectral features of the voice sample. Specifically, the voice characteristic extraction unit 120 extracts the fundamental frequency of the target voice, information related to speech (speech rate, speech time, and speech length), information related to pauses (pause level, pause count, and pause length), shimmer, jitter, formant, harmonic-to-noise ratio, loudness, spectral centroid, Mel Frequency Cepstral Coefficients (MFCC), identity vector (i-vector), articulatory rate, zero-crossing rate (zcr), voicing probability (vp), line spectral pairs (LSP), period perturbation, and amplitude perturbation quality (APQ). At least one of the following voice characteristics can be extracted: quotient, stiffness, energy, intensity, and entropy.

[0024] Here, the voice characteristic extraction unit 120 may first perform a pre-processing process for quantifying a voice sample. Through the pre-processing process, the time, frequency, etc. of the voice sample may be adjusted to be constant. Furthermore, through the pre-processing process, human voices and non-human voices may be distinguished. The voice characteristic extraction unit 120 may include an artificial neural network model (e.g., a convolutional neural network) trained to select only human voices from input voice samples, and may perform the above-mentioned pre-processing to select only human voices through the trained artificial neural network model. The voice feature extraction unit 120 extracts from the pre-processed voice samples the following characteristics: fundamental frequency of the voice, speech rate, speech duration, speech length, pause level, pause count, pause length, shimmer, jitter, formant, harmonic-to-noise ratio, loudness, spectral centroid, Mel Frequency Cepstral Coefficients (MFCC), identity vector (i-vector), articulatory rate, zero-crossing rate (zcr), voicing probability (vp), line spectral pairs (LSP), period perturbation, and amplitude perturbation quality (APQ). The voice feature extraction unit 120 can extract voice features including at least one of quotient, stiffness, energy, intensity, and entropy. The voice feature extraction unit 120 can also extract voice features corresponding to phonemes, roots, and spectrum regions using a publicly available voice feature extraction program (e.g., Praat).

[0025] The prediction model 130 may be pre-trained to predict the presence or absence of Alzheimer's disease of a subject based on the voice characteristics extracted from the voice characteristic extraction unit 120 and the demographic information provided through the data input unit 110. The prediction model 130 may include at least one analytical model selected from the group consisting of a linear regression model, a logistic regression model, a machine learning model, and a neural network model.

[0026] The prediction model 130 includes a multivariate logistic regression model, which can be configured as shown in Equation 1 below.

[0027]

number

[0028] JPEG0007625299000002.jpg24155

[0029] The prediction model 130 configured based on the above-mentioned Equation 1 can output a dementia risk probability value and output status information that evaluates the patient's status based on the dementia risk probability value. The dementia risk probability value indicates the probability of being diagnosed with dementia, and the status information can determine the patient's status as Alzheimer's disease or normal cognitive function based on the diagnosis of the patient based on a specialist's diagnostic criteria. For example, the prediction model 130 can output the calculated dementia risk probability value If JPEG0007625299000003.jpg718 is greater than or equal to 0.5, then the patient's status can be determined as Alzheimer's disease, and if the calculated dementia risk probability value is less than 0.5, then the patient's status can be determined as normal cognitive function.

[0030] However, the prediction model 130 of the present invention is not limited thereto. The prediction model 130 may be configured as a multinomial multivariate logistic regression model, and the prediction model 130 may determine the state of the patient as Alzheimer's disease, mild cognitive impairment, or normal cognitive function based on the calculated dementia risk probability value.

[0031] Hereinafter, a method for predicting Alzheimer's disease based on voice characteristics according to another embodiment of the present invention will be described.

[0032] 3 is a flowchart of an Alzheimer's disease prediction method based on voice characteristics according to an embodiment. The method can be performed by the prediction device according to Fig. 1 and Fig. 2, and Fig. 1, Fig. 2 and the related description can be referred to for the description of the embodiment.

[0033] Referring to FIG. 3, the method for predicting Alzheimer's disease based on voice characteristics according to an embodiment includes step S100 of recording a subject's voice to generate a voice sample, step S110 of receiving demographic information of the subject, step S120 of extracting voice characteristics from the generated voice sample, and step S130 of inputting the voice characteristics and the demographic information into a pre-trained prediction model to predict whether the subject has Alzheimer's disease.

[0034] In the steps of the Alzheimer's disease prediction method based on voice characteristics according to the embodiment, steps S100 and S110 are described sequentially for convenience of explanation, and are not limited to being performed sequentially according to the order of description. In some embodiments, step S110 may be performed prior to step S100. Also, a step of training a prediction model may be performed prior to the execution of the prediction method.

[0035] S100 records the subject's voice to generate a voice sample.

[0036] The subject performs at least one of picture drawing, standard sentence utterance, and story recollection tasks in a quiet room where recording is possible, and speaks spontaneously or recites aloud, and the subject's speech is recorded to generate a voice sample. The voice input unit 100 may include a condenser microphone and a device for controlling the same, and can record the subject's voice at a frequency of 16 Hz or more to generate a voice sample.

[0037] Demographic information of the subject is received S110.

[0038] The demographic information includes the subject's age, sex, and years of education. Exemplarily, the demographic information may be acquired through a questionnaire of the subject. The questionnaire may include information on the patient's age, sex, and years of education, and may correspond to materials generated using the medical judgment of a trained medical specialist, or may be materials collected under the supervision of other medical personnel and the patient's guardians. The data input unit 110 may receive the demographic information through the above-mentioned questionnaire materials, but is not limited thereto. In some embodiments, the subject's demographic information may be information directly input by the subject.

[0039] Next, voice features are extracted S120 from the generated voice samples.

[0040] The voice characteristics of the patient can be extracted from the input voice sample of the subject. An artificial neural network model can be used in a pre-processing process for quantifying the patient's voice sample. Through the pre-processing process, the time, frequency, etc. of the voice sample can be adjusted to be constant. In addition, if there are a large number of input voice samples, a pre-processing process can be performed to select the voice samples. The voice feature extraction unit 120 extracts from the pre-processed voice samples the following characteristics: fundamental frequency of the voice, speech rate, speech duration, speech length, pause level, pause count, pause length, shimmer, jitter, formant, harmonic-to-noise ratio, loudness, spectral centroid, Mel Frequency Cepstral Coefficients (MFCC), identity vector (i-vector), articulatory rate, zero-crossing rate (zcr), voicing probability (vp), line spectral pairs (LSP), period perturbation, and amplitude perturbation quality (APQ). The speech characteristics can be extracted including at least one of the following: quotient, stiffness, energy, intensity, and entropy.

[0041] Next, the voice characteristics and the demographic information are input into a pre-trained prediction model to predict whether the subject has Alzheimer's disease (S130).

[0042] The prediction model 130 may be pre-trained to predict the presence or absence of Alzheimer's disease of a subject based on the voice characteristics extracted from the voice characteristic extraction unit 120 and the demographic information provided through the data input unit 110. The prediction model 130 may include at least one analytical model selected from the group consisting of a linear regression model, a logistic regression model, a machine learning model, and a neural network model.

[0043] The prediction model 130 includes a multivariate logistic regression model, which may be configured as shown in Equation 1 below.

[0044]

number

[0045] JPEG0007625299000005.jpg24155

[0046] The prediction model 130 configured based on the above-mentioned Equation 1 can output a dementia risk probability value and output status information that evaluates the status of a patient based on the dementia risk probability value. The dementia risk probability value means the probability of being diagnosed with dementia, and the status information can determine the status of the patient as Alzheimer's disease or normal cognitive function based on the diagnosis of the patient based on the diagnostic criteria of a specialist. For example, the prediction model 130 can determine the patient's status as Alzheimer's disease if the calculated dementia risk probability value is 0.5 or more, and can determine the patient's status as normal cognitive function if the calculated dementia risk probability value is less than 0.5.

[0047] The method for predicting Alzheimer's disease based on voice characteristics according to the present embodiment may be implemented as an application or may be implemented in the form of program instructions that can be executed by various computer components and may be recorded in a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, and the like, either alone or in combination.

[0048] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like.

[0049] Examples of program instructions include not only machine code, such as produced by a compiler, but also high level language code that can be executed by a computer using an interpreter, etc. The hardware devices can be configured to operate as one or more software modules to perform the processes of the present invention, and vice versa.

[0050] The present invention has been described above with a focus on its preferred embodiments. Those skilled in the art will understand that the present invention can be embodied in modified forms without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered in an illustrative rather than restrictive sense. The scope of the present invention is set forth in the claims, not the above description, and all differences within the scope of the equivalents thereto should be construed as being included in the present invention. EXPERIMENTAL EXAMPLES

[0051] A prediction model was constructed for the method and apparatus for predicting Alzheimer's disease based on voice characteristics according to the above-described embodiment, and an experiment was conducted to verify the constructed prediction model.

[0052] A total of 210 subjects were included in the subject data for constructing the prediction model, and voice and diagnosis information was obtained from patients visiting Boramae Hospital in Seoul, Korea, and people registered at the Dongjak-gu Dementia Safety Center. Of the total subjects, 106 were in the Alzheimer's dementia group and 104 in the normal group. To collect the subjects' voices, voices were collected while the examiner and the subjects interacted with each other.

[0053] To quantify the subject's voice, we first preprocessed the voice samples using an artificial neural network model (convolutional neural network). In other words, if there were a large number of input voice samples, a preprocessing process was performed first to select human voice samples, preventing noise data from being input into the learning process. We extracted the voice characteristics of the voice using an automated voice characteristic extraction method for the preprocessed voice. The following voice characteristics were extracted: fundamental frequency (f0) mean, f0 std), information related to speech (speech rate, speech time, speech length), information related to pauses (pause rate, pause count, pause duration mean, pause duration standard deviation (std)), shimmer, jitter, formant, harmonic-to-noise ratio, loudness, and spectral values ​​(spectral centroid mean, spectral centroid std).

[0054] A prediction model was constructed that uses demographic information (age, sex, years of education) and each of the extracted voice characteristics as input values, outputs a probability value of dementia risk for the subject using an Alzheimer's dementia prediction model, and outputs condition information evaluating the patient's condition based on the dementia risk probability value. The prediction model was embodied as a multivariate logistic regression model and was constructed to output a dementia risk probability value. The prediction model can determine the patient's condition as Alzheimer's disease if the calculated dementia risk probability value is 0.5 or more, and can determine the patient's condition as normal cognitive function if the calculated dementia risk probability value is less than 0.5.

[0055] The prediction performance was tested through the constructed prediction model. Specifically, the area under the curve (AUC) of the receiver operating characteristic (ROC) curve, an index of prediction performance, was calculated for the prediction model. AUC is the value calculated by calculating the area under the ROC curve and is a representative index that indicates the overall performance of a prediction model, with the closer to 1 the better the performance. Figure 4 is a graph showing the results of the constructed prediction model, and the performance is shown in Table 1 below according to the optimal cutoff score by the Youden Index.

[0056] [Table 1]

[0057] The predictive power was AUC=0.816, the sensitivity was 0.802, and the specificity was 0.699. The positive predictive value (PPV) was 0.733, and the negative predictive value (NPV) was 0.774.

[0058] Previously, tests for diagnosing Alzheimer's disease used cerebral amyloid deposition PET and cerebrospinal fluid testing. Amyloid PET is expensive and not easily available at hospitals that operate specialized medical centers such as tertiary care facilities, and it also has risk factors such as exposure to radiation. Cerebrospinal fluid analysis has limitations such as the use of invasive lumbar puncture therapy, the need for labor to carry out the procedure and analysis, and the difference in reliability depending on the institution.

[0059] In contrast, the Alzheimer's disease prediction method and device based on voice characteristics of the present invention diagnoses Alzheimer's disease through a test using a non-invasive method, and is a psychiatric screening test that transcends time, space, and experts and can be performed easily outside of the home or hospital, via AI speaker / smartphone / tablet / PC, etc. In addition, it can determine the risk of Alzheimer's disease at first and second level medical facilities, minimizing false positives, which has the effect of saving costs and can be expanded to treatment programs in the future.

[0060] In other words, considering that cerebrospinal fluid tests and positron emission tomography (PET), which are used to accurately diagnose Alzheimer's disease, a representative dementia-causing disease, are restricted by experts in specific spaces and times, the method of testing whether a conversation subject has Alzheimer's disease using an AI speaker and smart device such as the present invention is expected to help overcome the time, space, and resource limitations associated with conventional testing.

[0061] As a result, it is expected that diagnosing the risk of Alzheimer's disease using the acoustic characteristics of voice will enable appropriate intervention measures to be taken for the continuously increasing dementia population.

Claims

1. a voice input unit configured to record a voice of the subject to generate a voice sample; a data input configured to receive demographic information of the subject; a voice feature extraction unit for extracting voice features from the generated voice samples; a predictive model pre-trained to predict whether the subject has Alzheimer's disease based on the voice characteristics and the demographic information; The prediction model includes a multivariate logistic regression model including a plurality of terms determined by a combination of a plurality of independent variables and a plurality of regression coefficients, and a term including an initial constant value; the plurality of independent variables includes a plurality of first independent variables representing voice characteristics and a plurality of second independent variables representing demographic information; the plurality of regression coefficients include constant values ​​that are regression coefficients for each of the first independent variable and the second independent variable; the plurality of regression coefficients are trained to determine a decision boundary for the multivariate logistic regression model; The Alzheimer's disease prediction device is configured such that the multivariate logistic regression model uses the determined decision boundary to output a dementia risk probability value according to the correlation between the voice characteristics and the demographic information, and outputs status information determining whether the patient has Alzheimer's disease or normal cognitive function based on a comparison of the dementia risk probability value with a reference value.

2. The Alzheimer's disease prediction device of claim 1 , wherein the demographic information includes the subject's age, sex, and years of education.

3. The voice characteristic extraction unit extracts voice fundamental frequency, speech rate, speech duration, speech length, pause level, pause count, pause length, shimmer, jitter, formant, harmonic-to-noise ratio, loudness, spectral centroid, Mel Frequency Cepstral Coefficients (MFCC), identity vector (i-vector), articulation rate, zero-crossing rate (zcr), voicing probability (vp), and 2. The Alzheimer's disease prediction device according to claim 1, wherein at least one of the following voice characteristics is extracted as the voice characteristics: frequency probability, line spectral pairs (LSP), period perturbation, amplitude perturbation quotient (APQ), stiffness, energy, intensity, and entropy.

4. The voice feature extraction unit includes an artificial neural network model that performs preprocessing to select human voices from the voice samples; The Alzheimer's disease prediction device according to claim 3 , wherein the voice feature extraction unit extracts the voice features from preprocessed voice samples.

5. The Alzheimer's disease prediction device according to claim 1 , wherein the prediction model includes at least one analytical model selected from the group consisting of a linear regression model, a logistic regression model, a machine learning model, and a neural network model.

6. The Alzheimer's disease prediction device of claim 5 , wherein the logistic regression model is a multivariate logistic regression model, and the multivariate logistic regression model is configured as shown in Equation 1 below. [0010] 【number】

7. A method for predicting Alzheimer's disease based on voice characteristics, which is performed by an Alzheimer's disease prediction device, comprising: a voice input unit recording a voice of a subject to generate a voice sample; a data input unit receiving demographic information of the subject; a speech feature extractor extracting speech features from the generated speech samples; The prediction model includes inputting the voice characteristics and the demographic information into a pre-trained prediction model to predict the presence or absence of Alzheimer's disease of the subject; The prediction model includes a multivariate logistic regression model including a plurality of terms determined by a combination of a plurality of independent variables and a plurality of regression coefficients, and a term including an initial constant value; the plurality of independent variables includes a plurality of first independent variables representing voice characteristics and a plurality of second independent variables representing demographic information; the plurality of regression coefficients include constant values ​​that are regression coefficients for each of the first independent variable and the second independent variable; the plurality of regression coefficients are trained to determine a decision boundary for the multivariate logistic regression model; The multivariate logistic regression model is configured to use the determined decision boundary to output a dementia risk probability value according to the correlation between the voice characteristics and the demographic information, and to output status information determining whether the patient has Alzheimer's disease or normal cognitive function based on a comparison of the dementia risk probability value with a reference value.

8. A computer readable medium having stored thereon computer readable instructions that, when executed by at least one processor, cause the at least one processor to perform steps including: recording a voice of a subject to generate a voice sample; receiving demographic information of the subject; extracting speech features from the generated speech samples; inputting the voice characteristics and the demographic information into a pre-trained prediction model to predict whether the subject has Alzheimer's disease; The prediction model includes a multivariate logistic regression model including a plurality of terms determined by a combination of a plurality of independent variables and a plurality of regression coefficients, and a term including an initial constant value; the plurality of independent variables includes a plurality of first independent variables representing voice characteristics and a plurality of second independent variables representing demographic information; the plurality of regression coefficients include constant values ​​that are regression coefficients for each of the first independent variable and the second independent variable; the plurality of regression coefficients are trained to determine a decision boundary for the multivariate logistic regression model; The multivariate logistic regression model is configured to use the determined decision boundary to output a dementia risk probability value according to the correlation between the voice characteristics and the demographic information, and to output status information determining as Alzheimer's disease or normal cognitive function based on a comparison of the dementia risk probability value with a reference value.

Images