Ai-based alzheimer's disease detection method using disfluency feature in speech

Abstract

The present invention enables the detection of Alzheimer's disease (AD) by utilizing a disfluency feature extracted from spontaneous speech of a participant. To this end, the present invention enables more accurate detection of Alzheimer's disease (AD) by utilizing a disfluency feature including a silent feature, a participant feature, an interviewer feature, and a metadata feature through voice segmentation and speaker segmentation using a voice activity detection (VAD) system. In particular, the present invention relates to an AI-based Alzheimer's disease detection method using a disfluency feature in speech, wherein from among various silent features, the present invention considers the following silent features with respect to the participant's speech, such as: the number of silent segments (Num_PS) between consecutive participant segments; the average of the lengths (AVG_PS) of the silent segments between consecutive participant segments; and the variance of the lengths (VAR_PS) of the silent segments between consecutive participant segments, and also considers the following features with respect to the interviewer feature associated with the level of interviewer intervention, such as: the number of interviewer segments (NUM_I) in a voice recording; and the number of speaker changes (Num_chn) between a participant segment and an interviewer segment in the voice recording, thereby enabling more accurate detection of Alzheimer's disease (AD).

Description

AI-based Alzheimer's disease detection method using disfluency features in speech

[0001] The present invention relates to a method for diagnosing Alzheimer's disease using only acoustic features, and more specifically, to detect Alzheimer's disease (AD) by utilizing disfluency features extracted from a spontaneous speech of a participant, and for this purpose, by utilizing disfluency features including static features, participant features, interviewer features, and metadata features through speech segmentation and speaker segmentation using a voice activity detection (VAD) system, Alzheimer's disease (AD) can be detected more accurately, and in particular, among various static features, in relation to the static features in the participant's speech, the number of static segments between consecutive participant segments (Num_PS), the average length of static segments between consecutive participant segments (Avg_PS), and the variance of the length of static segments between consecutive participant segments (Var_PS) are considered, and in relation to the interviewer features, in relation to the degree of interviewer intervention, the number of interviewer segments in a voice recording (Num_I), and the number of speaker changes between a participant segment and an interviewer segment in a voice recording (Num_chn) are considered, thereby enabling more accurate detection of Alzheimer's disease (AD) in speech. This is about an AI-based Alzheimer's disease detection method using non-fluent features.

[0002] Alzheimer's Disease (AD) is a neurodegenerative disease characterized by a long-term, progressive decline in cognitive function. It is a common form of dementia that initially progresses to mild cognitive impairment (MCI) and then to AD. Early detection of Alzheimer's disease is essential for timely treatment, which can slow cognitive decline and reduce the risk of developing dementia.

[0003] Typical symptoms of Alzheimer's disease include memory and cognitive decline, but language impairments are also common. Apraxia of speech, logopanic progressive aphasia (LPA), and anomia commonly occur. In more severe cases, patients may stutter, repeat thoughts and questions, and have difficulty forming simple sentences. In this regard, speech can be used to characterize the loss of verbal fluency associated with Alzheimer's disease, such as pauses and silences. Therefore, speech can serve as an important biomarker for rapid diagnosis of Alzheimer's disease and for monitoring cognitive status.

[0004] In order to accurately detect Alzheimer's disease (AD) from speech, it is important to automatically extract AD-related features from the input signal. Previous studies have explored various types of features for AD detection in voice recordings, and these features can be broadly classified into two main groups: acoustic features and text-based features.

[0005] Acoustic features are extracted directly from the raw audio signal using methods such as eGeMAPS (extended Geneva Minimalistic Acoustic Parameter Set), MFCC (Mel-Frequency Cepstral Coefficients), and x-vector, which capture various aspects of the speech signal such as pitch, intensity, and spectral content from audio signals of individuals with Alzheimer's disease (AD), and use them to identify the disease.

[0006] Meanwhile, text-based features convert speech recordings into text using an automatic speech recognition (ASR) system, and then extract metrics such as word proportions and BERT features. This approach assumes that individuals with AD use a different language than the general population, i.e., they have a lower vocabulary and less accurate word choice, and thus these text-based features are being used to identify the disease.

[0007] However, these methods can only achieve similar performance using linguistic approaches such as manual transcription or automatic speech recognition (ASR), and text-based methods often suffer from low accuracy and time-consuming tasks in noisy conditions, and have limited generalizability to multilingual AD detection.

[0008] (Patent Document) Domestic Patent No. 10-2274072 (registered on July 1, 2021) "Method and Device for Determining the Degree of Dementia in a User"

[0009] The prior patent disclosed in the above <patent document> also only discloses the concept of visualizing the characteristics of the voice input by the user and using this to determine the degree of dementia.

[0010] Therefore, there is a growing need for a method that can quickly and accurately diagnose Alzheimer's disease (AD) by easily extracting features from participants' voice recordings (voice data).

[0011] The present invention has been devised to solve the above problems.

[0012] The purpose of the present invention is to provide an AI-based Alzheimer's disease detection method using disfluency features in speech that can be applied across various languages, by detecting Alzheimer's disease (AD) by utilizing disfluency features extracted from spontaneous speech of a participant.

[0013] Another object of the present invention is to provide an AI-based Alzheimer's disease detection method using disfluency features in speech, which can more accurately detect Alzheimer's disease (AD) by utilizing disfluency features including static features, participant features, and interviewer features through speech segmentation and speaker segmentation using a voice activity detection (VAD) system.

[0014] Another object of the present invention is to provide an AI-based Alzheimer's disease detection method using disfluency features in speech, which can more accurately detect Alzheimer's disease (AD), by considering, among various static features, in particular, the number of static segments between consecutive participant segments (Num_PS), the average length of static segments between consecutive participant segments (Avg_PS), and the variance of the length of static segments between consecutive participant segments (Var_PS) related to static features in the participant's speech.

[0015] Another object of the present invention is to provide an AI-based Alzheimer's disease detection method using disfluency features in speech, which can more accurately detect Alzheimer's disease (AD), by considering the number of interviewer segments (Num_I) in a voice recording and the number of speaker changes between participant segments and interviewer segments (Num_chn) in a voice recording, in particular, in relation to interviewer characteristics, and in relation to the degree of interviewer involvement.

[0016] Another object of the present invention is to provide an AI-based Alzheimer's disease detection method using disfluency features in speech, which can more accurately detect Alzheimer's disease (AD) by utilizing disfluency features including metadata features such as the age and education level of the participant.

[0017] Another object of the present invention is to provide an AI-based Alzheimer's disease detection method using disfluency features in speech, which can detect Alzheimer's disease (AD) simply, quickly, and accurately by inputting extracted disfluency features and using a random forest (RF) model.

[0018] In order to achieve the above-mentioned purpose, the present invention is implemented by an embodiment having the following configuration.

[0019] According to one embodiment of the present invention, an AI-based Alzheimer's disease detection method using disfluency features in speech according to the present invention is characterized by including the steps of performing voice segmentation in speech data (S1); extracting disfluency features based on the segmented speech patterns (S3); and receiving the extracted disfluency features as input and detecting Alzheimer's disease (AD) using a neural network (S5).

[0020] According to another embodiment of the present invention, the step S1 in the present invention is characterized in that it targets voice data from a speaker's spontaneous speech, and includes a step (S11) of segmenting voices from the voice data and a step (S12) of segmenting speakers from the voice data.

[0021] According to another embodiment of the present invention, step S11 of the present invention is characterized in that it divides the input voice data into voice segments and static segments using a neural network, and step S12 divides the divided voice segments into participant segments and interviewer segments using a neural network.

[0022] According to another embodiment of the present invention, the disfluency features in the S3 step of the present invention are characterized by including static features, participant features, and interviewer features.

[0023] According to another embodiment of the present invention, the static feature in the present invention is characterized by including a sum of the lengths of static segments in a voice recording (Len_S), a variance of the lengths of static segments in a voice recording (Var_S), an average of the lengths of static segments in a voice recording (Avg_S), a maximum length of the lengths of static segments in a voice recording (Max_S), a ratio of static segments to the total voice length (Rat_S), and a number of static segments in a voice recording (Num_S).

[0024] According to another embodiment of the present invention, the static feature in the present invention is characterized in that it further includes the number of static segments between consecutive participant segments (Num_PS), the average of the lengths of static segments between consecutive participant segments (Avg_PS), and the variance of the lengths of static segments between consecutive participant segments (Var_PS).

[0025] According to another embodiment of the present invention, the participant characteristics include a ratio of participant segments to the total voice length (Rat_P) and a number of participant segments in a voice recording (Num_P), and the interviewer characteristics include a number of interviewer segments in a voice recording (Num_I) and a number of speaker changes between participant segments and interviewer segments in a voice recording (Num_chn).

[0026] According to another embodiment of the present invention, the disfluency feature in the S3 step of the present invention additionally includes a metadata feature, and the metadata feature is characterized in that it includes the total recording length of the voice recording (Len), the age of the participant (Age), and the annual education level of the participant (Edu).

[0027] According to another embodiment of the present invention, the step S5 in the present invention is characterized in that it receives the extracted non-fluent features as input and detects Alzheimer's Disease (AD) using a random forest (RF) model.

[0028] The present invention can obtain the following effects through the combination and use of the configuration described below with the previously described embodiment.

[0029] The present invention has the effect of being applicable across various languages, as it can detect Alzheimer's disease (AD) by utilizing disfluency features extracted from spontaneous speech of a participant.

[0030] The present invention has the effect of detecting Alzheimer's disease (AD) more accurately by utilizing non-fluent features including static features, participant features, and interviewer features through voice segmentation and speaker segmentation using a voice activity detection (VAD) system.

[0031] The present invention has the effect of detecting Alzheimer's disease (AD) more accurately by considering, among various static features, the number of static segments between consecutive participant segments (Num_PS), the average length of static segments between consecutive participant segments (Avg_PS), and the variance of the length of static segments between consecutive participant segments (Var_PS) in relation to static features in the participant's speech.

[0032] The present invention has the effect of detecting Alzheimer's disease (AD) more accurately by considering the number of interviewer segments (Num_I) in a voice recording and the number of speaker changes between participant segments and interviewer segments (Num_chn) in a voice recording, particularly in relation to interviewer characteristics, and in relation to the degree of interviewer intervention.

[0033] The present invention has the effect of detecting Alzheimer's disease (AD) more accurately by utilizing non-fluent features including metadata features such as the participant's age and education level.

[0034] The present invention has the effect of detecting Alzheimer's disease (AD) simply, quickly, and accurately by inputting extracted non-fluent features and using a random forest (RF) model.

[0035] Figure 1 is an image visualizing the disfluency characteristics in speech with and without AD.

[0036] Figure 2 is a reference diagram of the non-fluent characteristics presented in the present invention.

[0037] Figure 3 is a graph showing the AD identification accuracy in various AD classification models.

[0038] Figure 4 is a graph showing the contribution of disfluency features.

[0039] Figure 5 is a comparison chart of AD identification accuracy in the ADReSSo test set.

[0040] Figure 6 is a comparison chart of AD identification accuracy in the MADReSS test set.

[0041] Hereinafter, preferred embodiments of an AI-based Alzheimer's disease detection method utilizing disfluency characteristics in speech according to the present invention will be described in detail with reference to the attached drawings. It should be noted that identical components are represented by the same reference numerals throughout the drawings, as much as possible. Unless otherwise defined, all terms in this specification have the same general meaning as those terms understood by a person skilled in the art to which the present invention pertains, and if there is a conflict with the meaning of a term used in this specification, the definitions used in this specification shall prevail. Throughout the specification, when a part is said to "include" a certain component, this does not mean that other components are excluded, but rather that other components may be included, unless specifically stated otherwise. In addition, terms such as "... part" and "... module" described in the specification mean a unit that processes at least one function or operation, and this may be implemented by hardware, software, or a combination of hardware and software.

[0042]

[0043] Referring to FIGS. 1 to 4, an AI-based Alzheimer's disease detection method using disfluency features in speech according to one embodiment of the present invention may include a step (S1) of performing voice segmentation in speech data; a step (S3) of extracting disfluency features based on the segmented speech patterns; and a step (S5) of receiving the extracted disfluency features and detecting Alzheimer's disease (AD) using a neural network.

[0044] The above S1 step is a process of performing voice segmentation in voice data, and as previously explained, the present invention focuses on identifying Alzheimer's (AD) related features in speech, and in particular, focuses on disfluency in a speaker's spontaneous speech. Therefore, the S1 step also targets voice data in a speaker's spontaneous speech, and performs voice segmentation using a Voice Activity Detection (VAD) system (algorithm) to extract disfluency features from a voice recording without relying on transcription.

[0045] More specifically, the above step S1 may include a step of segmenting voices from voice data (S11) and a step of segmenting speakers from voice data (S12).

[0046] The above step S11 is a process of dividing the input voice data into voice segments and static segments using a neural network (e.g., the VAD algorithm or a pre-trained audio neural network (PANN)). The goal of the S11 step is to divide the voice data into voice parts (voice segments) and pause parts (static segments). To this end, pre-trained audio neural networks (PANNs: large-scale pretrained audio neural networks for audio pattern recognition) can be used to capture voice patterns. The PANN receives raw recorded voice data as input and outputs audio tags. Audio segments tagged as 'speaking' are regarded as voice parts (voice segments), and other parts are regarded as pause parts (static segments).

[0047] The above step S12 is a process of segmenting the segmented voice segments into participant segments and interviewer segments using a neural network. For this speaker segmentation, the audio segments (voice segments) previously tagged as 'speaking' in the above step S11 can be automatically separated into different speakers in the audio segments using a neural network (e.g., Pyannote (Pyannote.Audio: neural building blocks for speaker diarization), etc.). For example, if the minimum number of speakers is set to 1 and the maximum number of speakers is set to 2, the speaker with the longer speech time in the audio segment can be regarded as the participant, and the other speakers can be automatically separated as the interviewer.

[0048] As a result, through the above steps S11 and S12, the voice data can be expressed as a sequence of P representing the voice segment of the participant, I representing the voice segment of the interviewer (or investigator), and S representing the silence segment (see Fig. 1).

[0049] For reference, Figure 1 is an image visualizing the disfluency features in speech with and without AD. Referring to Figure 1, it shows that participants with AD have longer response times, more interviewer intervention, and longer and more pauses between utterances regardless of language. These parts are utilized in designing the disfluency features described later.

[0050] The above S3 step is a process of extracting disfluency features based on the segmented speech pattern, and the disfluency features in the S3 step may include static (S) features, participant (P) features, interviewer (I) features, and metadata (M) features (see Fig. 2).

[0051] First, the above static features can be seen as the most important features for AD detection, because abnormal speech interruptions, longer durations of silence, and greater number of pauses in speech are symptoms that appear in Alzheimer's disease (AD) patients. The above static features basically include statistical features. That is, they can include the sum of the lengths of static segments in a voice recording (Len_S), the variance of the lengths of static segments in a voice recording (Var_S), the average of the lengths of static segments in a voice recording (Avg_S), the maximum length of static segments in a voice recording (Max_S), the ratio of static segments to the total voice length (Rat_S), and the number of static segments in a voice recording (Num_S). Among the above static features, the sum of the lengths of static segments in a voice recording (Len_S), the average of the lengths of static segments in a voice recording (Avg_S), the ratio of static segments to the total voice length (Rat_S), and the number of static segments in a voice recording (Num_S) are static features that are considered to increase the accuracy of AD diagnosis through statistical features related to static segments in the input voice data. In particular, among the above static features, the maximum length of static segments in a voice recording (Max_S) and the variance of the lengths of static segments in a voice recording (Var_S) are features that are considered to enable more accurate AD detection using static features while minimizing the influence of abnormal data that may be included in the voice data.

[0052] Meanwhile, in the above static features, only features related to static features in the participant voice data, i.e., features induced by the participant (S) among speakers, can be additionally included. Specifically, the number of static segments between consecutive participant segments (Num_PS), the average length of static segments between consecutive participant segments (Avg_PS), and the variance of the length of static segments between consecutive participant segments (Var_PS) can be additionally included. The number of static segments between consecutive participant segments (Num_PS) can be used to identify the characteristics of the number of static segments in the participant's continuous speech process, that is, how often pauses (silence pauses) appear in the speech process. The average length of the static segments between consecutive participant segments (Avg_PS) can be used to identify the characteristics of how long the static segments, i.e., pauses (silence pauses, voiceless pauses are easier to extract from speech signals (speech data) than filled pauses such as 'uh', 'um', etc., and can be used for analysis regardless of language) continue on average in the participant's continuous speech process. The variance of the length of the static segments between consecutive participant segments (Var_PS) can be used to identify the characteristics of how evenly the length of the static segments, i.e., pauses (silence pauses), appear in the participant's continuous speech process. Therefore, by utilizing these characteristics, it is possible to identify the characteristics of how slowly the original speech is made (even in the absence of AD disease). This can improve the accuracy of AD diagnosis by filtering out participants with a speech tendency to have regular pauses between pronunciations.

[0053] The above participant features include the proportion of participant segments in the total speech length (Rat_P) and the number of participant segments in the speech recording (Num_P), and the above interviewer features include the number of interviewer segments in the speech recording (Num_I) and the number of speaker changes between participant segments and interviewer segments in the speech recording (Num_chn). The interviewer features do not include the proportion of interviewer segments (Rat_I) to avoid unnecessary duplication (i.e., Rat_S+Rat_P+Rat_I is always 1).

[0054] In particular, the above interviewer characteristics serve as an important factor in increasing the accuracy of AD diagnosis, because the interviewer's (investigator's) intervention occurs more frequently during the interview (investigation) process for AD patients, and therefore, not only the number of interviewer segments in the voice recording (Num_I), but especially the number of speaker changes between the participant segment and the interviewer segment in the voice recording (Num_chn) can be viewed as disfluency characteristics that contribute to increasing the accuracy of AD diagnosis.

[0055] The above metadata features are features that allow for the consideration of the relativity of voice data and disfluency features extracted therefrom in AD diagnosis using acquired voice data, and may include the total recording length (Len) of the voice recording, the participant's age (Age), and the participant's educational level (Edu) in years. In other words, it allows for the consideration that the older the participant, the longer it takes to speak overall and the higher the risk of AD, and that the participant's educational level affects spontaneous speech, thereby allowing for the consideration of the relativity of disfluency features for each participant, thereby enabling a more accurate AD diagnosis.

[0056] The above step S5 is a process of detecting Alzheimer's Disease (AD) using a neural network by inputting the extracted disfluent features. In general, the dataset for diagnosing cognitive impairment (including AD) is small compared to other fields, so overfitting to the training data may be a problem. To prevent this overfitting, a simple machine learning classifier can be used for AD classification / diagnosis. Examples include a support vector machine (SVM) using a linear kernel and a regularization parameter 1, a multilayer perceptron (MLP) with a single hidden layer of 20 hidden units, a K-nearest neighbor (KNN) classifier with K=5, and a random forest (RF) with 20 trees and a maximum depth of 5. Among these, as will be described later, the model that can most accurately detect Alzheimer's Disease (AD) by inputting the extracted non-fluent features and using a random forest (RF) model is the model.

[0057]

[0058] Below, we verify the accuracy of the AD detection model using a simple classifier by utilizing the disfluency feature in speech proposed in the present invention through experiments.

[0059] - Data set used

[0060] To evaluate the performance of our model, we utilized two publicly available datasets: ADReSSo and MADReSS. These datasets were used in the AD Recognition Challenge held in 2021 and 2023, respectively. These datasets contain audio recordings of photo captions generated by individuals with AD and cognitively normal individuals. The ADReSSo dataset consists of 156 English training data and 71 English test data, while the MADReSS dataset consists of 237 English training data and 46 Greek test data. Both datasets provide only audio recordings, without any transcription. For the AD classification / diagnosis task, the audio recordings were labeled as "normal" and "suspected AD," and gender and age were balanced to minimize potential bias.

[0061] - Performance comparison of machine learning classifiers

[0062] The accuracy of AD identification was verified through four machine learning models, including SVM, MLP, KNN, and RF, utilizing the disfluency features in the utterances proposed in the present invention (for this, the 5-fold cross-validation accuracy was used on the ADReSSo training set). As shown in Fig. 3, it can be confirmed that the proposed disfluency features showed reasonable performance in all of the various models, and in particular, the Random Forest (RF) model showed the best performance (accuracy of 78.3%, low variance). Since the RF model is an ensemble learning method that builds various decision trees during training, the following experiments were conducted using the RF model.

[0063] - Analysis of disfluency features

[0064] In order to examine the characteristics (contribution) of the disfluency features proposed in the present invention, we used the results of 5-fold cross-validation on the ADReSSo training set while sequentially removing some features from the results utilizing all disfluency features. As shown in Fig. 4, the results show that removing the number of interviewer segments (Num_I) from the voice recording has a greater impact on the performance than simply removing the features of the proportion of participant segments to the entire speech length (Rat_P) and the number of participant segments in the voice recording (Num_P). This confirms that the interviewer feature serves as a more important feature because participants with AD are more influenced and intervened by the interviewer during their speech. In addition, when the static features such as the ratio of static segments to the total speech length (Rat_S), the sum of the lengths of static segments in the voice recording (Len_S), and the maximum length of static segments in the voice recording (Max_S) were removed, the results showed that they had a great impact on the AD detection performance. This confirms that static features during speech, namely silence pauses, are important features in detecting cognitive impairments such as AD. Finally, when the metadata feature such as the participant's year of education (Edu) was removed, there was a slight decrease in performance, confirming that these metadata features provide an auxiliary function in AD diagnosis. In conclusion, among the disfluency features, static features are the most important factor in AD detection, and interviewer features, especially features related to the number of interventions, are also very important features.

[0065] - Comparison with other AD identification / diagnosis methods (models)

[0066] To verify the superiority of the model proposed in the present invention, we compared it with other models that only use audio features, such as wav2vec, MFCC eGeMAPS, or ComParE. We trained the model on the ADReSSo training set and compared the test results on the test set. As shown in Figure 5, the baseline method presented the ADReSSo dataset, built a simple model using eGeMAPS and an SVM classifier, and achieved an accuracy of 64.8%, which is similar to the results of Pappagari et al. ["Automatic detection and assessment of alzheimer's disease using speech and language technologies in low-resource scenarios," in Proc. Interspeech, 2021] and Pan et al. ["Using the outputs of different automatic speech recognition paradigms for acoustic- and BERT-based alzheimer's dementia detection through spontaneous speech," in Proc. Interspeech, 2021] achieved 74.7% accuracy on the ADReSSo test set using ASR embeddings and wav2vec 2.0. However, ASR embeddings and wav2vec features are extracted from a model trained solely on English data, making them difficult to transfer to other languages. In contrast, the disfluency feature of our proposed model (Ours) outperforms other features (76.6% accuracy), demonstrating its potential for use in other languages.

[0067] Additionally, to evaluate the performance of the model proposed in the present invention in different language scenarios, the model was trained with MADReSS English training data and the results were compared on Greek test data. As shown in Fig. 6, the performance of the model of the present invention was compared with the MADReSS baseline ["Multilingual alzheimer's dementia recognition through spontaneous speech: a signal processing grand challenge," in Proc. IEEE Int. Conf. Acoust. Speech Signal Process., 2023]. The results showed that the model of the present invention (Ours) using the disfluency feature achieved 80.4% accuracy on the Greek data, which was 6.5% higher than the baseline. Therefore, this result indicates that the disfluency feature is a speech feature that can be generalized across different languages ​​in the context of AD detection.

[0068]

[0069] In the above, the applicant has described various embodiments of the present invention, but such embodiments are only examples of implementing the technical idea of ​​the present invention, and any change or modification that implements the technical idea of ​​the present invention should be interpreted as falling within the scope of the present invention.

Claims

1. Step of performing voice segmentation from voice data (S1); Step (S3) of extracting disfluency features based on segmented speech patterns; An AI-based Alzheimer's disease detection method using disfluency features in speech, characterized by including a step (S5) of detecting Alzheimer's disease (AD) using a neural network by inputting extracted disfluency features.

2. In paragraph 1, The above S1 step targets voice data from the speaker's spontaneous speech, An AI-based Alzheimer's detection method using disfluency features in speech, characterized in that it includes a step of segmenting speech from speech data (S11) and a step of segmenting speakers from speech data (S12).

3. In paragraph 2, The above step S11 uses a neural network to divide the input voice data into voice segments and static segments, An AI-based Alzheimer's detection method using disfluency features in speech, characterized in that the above step S12 divides the segmented speech segments into participant segments and interviewer segments using a neural network.

4. In paragraph 3, An AI-based Alzheimer's detection method using disfluency features in speech, characterized in that the disfluency features in the above S3 step include static features, participant features, and interviewer features.

5. In paragraph 4, An AI-based Alzheimer's detection method using disfluency features in speech, characterized in that the static features include a sum of the lengths of static segments in a voice recording (Len_S), a variance of the lengths of static segments in a voice recording (Var_S), an average of the lengths of static segments in a voice recording (Avg_S), a maximum length of static segments in a voice recording (Max_S), a ratio of static segments to the total voice length (Rat_S), and a number of static segments in a voice recording (Num_S).

6. In paragraph 5, An AI-based Alzheimer's detection method using disfluency features in speech, characterized in that the static features further include the number of static segments between consecutive participant segments (Num_PS), the average of the lengths of static segments between consecutive participant segments (Avg_PS), and the variance of the lengths of static segments between consecutive participant segments (Var_PS).

7. In paragraph 6, The above participant characteristics include the ratio of participant segments to the total speech length (Rat_P), the number of participant segments in the speech recording (Num_P), The above interviewer features are the number of interviewer segments in the voice recording (Num_I), and the number of speaker changes between participant segments and interviewer segments in the voice recording (Num_chn). An AI-based Alzheimer's disease detection method utilizing disfluency features in speech characterized by including:

8. In paragraph 4, The non-fluency feature in the above S3 step additionally includes metadata features, An AI-based Alzheimer's detection method using disfluency features in speech, characterized in that the above metadata features include the total recording length (Len) of the voice recording, the age of the participant (Age), and the annual education level (Edu) of the participant.

9. In any one of paragraphs 1 to 8, The above S5 step is an AI-based Alzheimer's disease detection method using disfluency features in speech, characterized in that it receives extracted disfluency features as input and detects Alzheimer's disease (AD) using a random forest (RF) model.

Images