CN114299996A - AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease - Google Patents
AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease Download PDFInfo
- Publication number
- CN114299996A CN114299996A CN202111677396.XA CN202111677396A CN114299996A CN 114299996 A CN114299996 A CN 114299996A CN 202111677396 A CN202111677396 A CN 202111677396A CN 114299996 A CN114299996 A CN 114299996A
- Authority
- CN
- China
- Prior art keywords
- voice
- parkinson
- frozen gait
- training
- symptoms
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a speech analysis method of key characteristic parameters of frozen gait symptoms of Parkinson's disease based on an AdaBoost algorithm. Collecting continuous and stable vowels of a Parkinson disease patient, and recording whether the Parkinson disease patient has frozen gait symptoms; performing denoising pretreatment on the voice signals and removing silent segments; step three, extracting a plurality of voice features; performing feature selection on the original features by using a CART algorithm, and screening out key features capable of effectively representing frozen gait symptom information; step five, training an AdaBoost model; and step six, inputting the feature vector of the voice to be detected into the model to obtain key feature parameters of the Parkinson's disease frozen gait symptoms. The method uses the AdaBoost algorithm to analyze the symptoms of the frozen gait of the Parkinson disease, improves the model precision by utilizing the ensemble learning, and reduces the cost of early analysis of the symptoms of the frozen gait of the Parkinson disease.
Description
Technical Field
The invention belongs to the field of machine learning and medical application, and relates to a speech analysis method and system for key characteristic parameters of frozen gait symptoms of Parkinson's disease based on AdaBoost algorithm.
Background
Parkinson's disease is the second major neurodegenerative disease next to alzheimer's disease, and its main symptoms include motor symptoms including muscle stiffness, tremor and some other movement disorders, and non-motor symptoms including hyposmia, constipation, abnormal sleep behavior, depression, and the like. These symptoms are caused by a decrease in the number of dopaminergic neurons.
Frozen gait is one of the most serious motor impairment symptoms of Parkinson's disease, and refers to that the patient has a short, sudden stop or obvious reduction of step when attempting to walk or during the advancing process, and the most common symptoms are that the patient is hesitant to start, has a blocked step and is suddenly difficult to walk. Frozen gait is disabling, and most patients need to move with the help of a wheelchair after symptoms appear for five years on average, which greatly affects the quality of life of the patients.
Existing studies have shown that there is a pathological link between parkinson's disease and dysphonia. Speech impairment may be one of the earliest signs of Parkinson's disease, with symptoms manifested as slow speech, hoarseness, low volume, and vocalization tremor. These speech impairments are due to loss of control of the laryngeal, vocal and respiratory muscles of parkinson's disease patients.
The voice is mainly generated by the cooperation of internal sounding organs, and the cooperation among human organs is uniformly coordinated and controlled by neurons. Parkinson patients cannot stably control vocal organs due to related neuron loss, so that the patients have different degrees of vocal disorder. Patients with parkinson's disease often do not produce smooth and accurate sounds compared to healthy people. Therefore, the early symptom analysis of the Parkinson disease can be carried out by utilizing the voice signals. Compared with the traditional Parkinson disease analysis method, the method for analyzing the Parkinson disease by using the voice signal is economical and efficient, and the voice monitoring is non-contact, simple and convenient. For pronunciation, the sustained vowels/a/,/i/,/o/, may be used. The sustained vowel/a/is the easiest to emit and experience has shown that it is the most conveying clinically useful information. Physiologically, vowels/a/involve a combination of various muscles in the vocal cords and vocal tract, thus it increases the probability that neurological problems can be identified. When the speech signal is used for analyzing the Parkinson's disease, the speech signal to be detected needs to be analyzed through a speech signal processing algorithm so as to extract speech characteristic information capable of representing pathological characteristics of the Parkinson's disease.
Based on the above situation, after the speech features are extracted by the speech signal processing algorithm, the number of the extracted speech features is often large, irrelevant features may exist, and relevance may exist between the features, which easily causes dimension disaster. The feature selection can eliminate irrelevant or redundant features and select key features useful for the current classification task, so that the aims of improving the accuracy of the model, reducing the time of model prediction and providing key features with certain interpretability are fulfilled. And then, the frozen gait symptom of the Parkinson disease can be analyzed by using a technical model in the field of machine learning.
Disclosure of Invention
The invention provides a voice analysis method and system of key characteristic parameters of Parkinson's disease frozen gait symptoms based on AdaBoost, aiming at the problem that the traditional analysis of the Parkinson's disease frozen gait symptoms is difficult. The collected voice signals are subjected to feature extraction and interpretable feature selection, and then the analysis of the frozen gait of the Parkinson patient is carried out by combining an integrated learning algorithm in machine learning, so that the interpretable key feature analysis of the frozen gait symptom of the Parkinson patient is realized, and a subsequent specific treatment strategy is adopted in time.
In order to achieve the above object, the method of the present invention comprises the steps of:
step one, collecting continuous and stable vowels of a Parkinson disease patient, and recording whether the Parkinson disease patient has a frozen gait symptom.
And step two, denoising the voice signals and removing the mute segments.
And step three, extracting various voice characteristics by using a voice signal processing algorithm.
Extracting voice features by using a voice signal processing algorithm, wherein the extracted features comprise: the method comprises the following steps of obtaining a fundamental frequency profile F0_ contourr, an average fundamental frequency F0_ ave, a minimum fundamental frequency F0_ min, a maximum fundamental frequency F0_ max, four characteristics Jitter, RAP, PPQ and DDP of fundamental frequency change, five characteristics Shimmer, APQ3, APQ5, APQ11 and DDA of amplitude change, a noise-harmonic ratio NHR, a harmonic-noise ratio HNR, a cycle period density entropy RPDE, a trend fluctuation analysis DFA, a gene period entropy PPE and a Mel frequency cepstrum coefficient MFCC obtained by converting characteristics in a Mel cepstrum domain.
And step four, carrying out feature selection by using a CART algorithm, and screening out key features with interpretability, which can effectively represent the frozen gait symptom information.
The method comprises the following steps of (1) utilizing a CART algorithm to select features, wherein the specific process comprises the following steps:
the specific formula of the Gini index (D) of the data set D is:
wherein p iskRepresenting the probability that the sample point belongs to the kth class, wherein K represents K classification problems;
gini index of characteristic aindex(D, a) is defined as:
wherein V represents the feature aThere are V possible values; therefore, the feature value which minimizes the divided kini index is selected as the optimal division feature a*Namely:
a*＝argmaxa∈AGiniindex(D,a)(3)
and step five, training an AdaBoost model.
Step six, voice analysis: and inputting the feature vector of the voice to be detected into the model to obtain the key feature parameters of the frozen gait symptom of the person to be detected.
The invention also discloses a speech analysis system of key characteristic parameters of the Parkinson's disease frozen gait symptoms based on the AdaBoost algorithm, which comprises the following steps:
the voice signal acquisition module is used for executing the first step and the acquisition of the voice signals: collecting continuous and stable vowels of a Parkinson disease patient, and recording whether the Parkinson disease patient has a frozen gait symptom;
the voice signal processing module is used for executing the second step and the voice signal preprocessing: denoising the voice signal and removing a mute segment;
and the voice feature extraction module is used for executing the third step and the voice feature extraction: extracting various voice features by using a voice signal processing algorithm; (ii) a
The voice feature selection module is used for executing the fourth step and the feature selection: carrying out feature selection by using a CART algorithm, and screening out key features capable of representing frozen gait symptoms;
and the AdaBoost classification model training module is used for executing the fifth step and training the model: training an AdaBoost classification model by using a decision tree as a base classifier;
and the voice analysis module is used for executing the sixth step and the voice analysis: and inputting the feature vector of the voice to be detected into the model to obtain the key feature parameters of the frozen gait symptom of the person to be detected.
The Parkinson's disease frozen gait voice analysis method based on the AdaBoost algorithm has the beneficial effects that:
1. the cost of early analysis of the frozen gait of the Parkinson disease patient is reduced. Because the invention performs the characteristic extraction on the voice signal and performs the frozen gait symptom analysis by using the machine learning method, the higher cost of diagnosis in a hospital is avoided, and the early frozen gait symptom analysis can be performed only by acquiring the voice signal and the gait characteristic information of the Parkinson disease patient to establish the model, thereby saving the diagnosis cost.
2. The efficiency of analyzing the symptoms of the frozen gait of the Parkinson disease is improved. The traditional Parkinson disease analysis method is to comprehensively analyze whether a patient has frozen gait symptoms of the Parkinson disease, the degree of the disease and the possible disease development trend of the patient according to the test of a series of movements, tremors and the like of the patient by a doctor, and the disease analysis means is complex and time-consuming. The invention is obtained by a machine learning method, and the dimensionality of the input speech feature vector can be reduced by effectively selecting the speech features, thereby improving the program analysis efficiency of a machine learning algorithm.
3. Provides a reliable Parkinson's disease frozen gait symptom key characteristic parameter analysis means. The traditional frozen gait symptom analysis tests a patient in multiple movements, tremors and other aspects through doctors, and the analysis accuracy is greatly influenced because the movement symptoms are not obvious in the early onset of the Parkinson's disease. The invention utilizes the physiological relation between the voice and the Parkinson disease, analyzes the early frozen gait symptom according to the voice of a patient, has small subjective influence, and screens out key voice characteristics with interpretability through CART characteristic selection to be used as a reference value for analyzing and evaluating the state of illness by a doctor.
Drawings
FIG. 1 is a flow chart of classifier training of the present invention;
FIG. 2 is a block diagram of a system module of the present invention.
Detailed Description
The beneficial effects of the invention are verified by adopting the following experiments:
the experiment selects the Parkinson disease patient data set collected by the invention as a research object. A total of 212 speech samples were collected from 53 Parkinson's disease patients. And extracting various voice characteristics after denoising pretreatment. Each sample contains 59 speech features, wherein each speech feature contains 7 statistical values; the sample contains a label identifying gait feature information, 1 indicates that there is a frozen gait symptom, and 0 indicates that there is no frozen gait. As shown in fig. 1, the speech-based parkinson's disease frozen gait analysis method proceeds according to the following steps, according to the attributes in the data set:
step one, collecting continuous and stable vowels of a Parkinson disease patient, and recording whether the Parkinson disease patient has a frozen gait symptom.
And secondly, preprocessing the voice signals, including denoising and removing the mute segments.
And step three, extracting a plurality of voice features by using a voice signal processing algorithm, wherein each sample comprises 59 voice features.
Step four, feature selection: and (4) carrying out feature selection by using a CART algorithm, and screening out key features capable of representing frozen gait symptoms.
The depth of a decision tree of the CART algorithm is set to be 5, the impurity calculation method is a Gini coefficient, and the minimum sample number required by the branch child node is set to be 3. The CART algorithm is run for 10 times to obtain key features selected under the average condition, wherein the key features comprise: the 9 th and 11 th frequency ranges MFCC _ mean _9 and MFCC _ mean _11 of the mel-frequency cepstral coefficients MFCC, the fundamental frequency profile f0_ contour _ mean, the fundamental frequency perturbation signature DDP, the trend fluctuation analysis DFA, Shimmer, and the like.
And dividing the data set subjected to feature selection into a training set and a testing set. And a five-fold cross verification method is adopted to randomly divide the data set into five mutually exclusive data subsets with similar sizes. And selecting the union of four subsets as a training set for training the decision tree model, and using the rest subset as a test set for testing the performance of the model. Finally, the mean value of the test results of the five groups of training/test sets is obtained and used as the performance measurement index of the model.
Step five, training the model: and training an AdaBoost classification model by using a decision tree as a base classifier.
Before model training, the training set and test set data are normalized, and all data are mapped into a numerical range of [0,1 ]. The role of the normalization algorithm is: 1. each attribute in the data set has a physical background and therefore their units and range are different. The normalization can eliminate the influence of unit or order of magnitude, and maps all data into a preset range, thereby providing convenience for the subsequent data processing; 2. the normalization can improve the running speed of the program and accelerate convergence; 3. singular sample data (sample vectors that are particularly large or small relative to other input samples) may increase training time and even cause the algorithm to fail to converge. Normalization is performed before training, so that the influence of singular sample data on the training process can be eliminated.
The specific process for constructing the AdaBoost model comprises the following steps:
1) initializing a weight distribution of training data, each training sample being initially given the same weight: 1/N;
wherein w1iRepresents the weight at the beginning of the ith training sample, and N represents the total number of samples;
2) performing M iterations, each iteration performing the following steps:
a. using a weight distribution DnThe training data set is learned to obtain a base classifier Gm(x)：
Gm(x):χ→{-1，+1}
b. Calculation of Gm(x) Classification error rate e on training data setm：
Wherein G ism(xi) Presentation basis classifier Gm(x) In training data xiClassification result of (1), yiRepresenting training data xiTrue classification of, wmiRepresenting sample x at the m-th iterationiWeight of (P)() Representing the probability of an event;<() Indicating that the result is 1 when the event in the brackets is true, and otherwise the result is 0;
c. calculation of Gm(x) To obtain the weight alpha of the basic classifier in the final classifierm：
Wherein emRepresents Gm(x) A classification error rate on the training data set;
d. updating the weight distribution of the training data set:
Dm+1＝(wm+1,1,wm+1,2,…,wm+1,i,…,wm+1,N)
wherein wm+1,iRepresenting the updated weight of the + th training sample after iteration for m times; zmRepresenting a normalization factor, exp () representing an exponential function with a natural constant e as the base;
3) and combining all the base classifiers to obtain a final classifier, wherein the final classification result is obtained by all the base classifiers through weighted voting:
where f (x) represents a weighted combination of the individual base classifiers, G (x) represents the final classifier, sign () represents a sign function.
Step six, voice analysis: and inputting the feature vector of the voice to be detected into the model to obtain the key feature parameters of the frozen gait symptom of the person to be detected. Here, 100 voice samples to be tested are selected for voice analysis, and partial key characteristic parameter values of the frozen gait are obtained as shown in the following table.
After the key speech features are selected, early analysis of the symptoms of the frozen gait is performed by using the features, and the test accuracy is 87.6%.
Fig. 2 shows a block diagram of the system module of the present invention, and the system includes:
a voice signal acquisition module; the method is used for executing the step one and acquiring the voice signals: collecting continuous and stable vowels of a Parkinson disease patient, and recording whether the Parkinson disease patient has a frozen gait symptom;
a voice signal processing module; the preprocessing of the voice signal is carried out in the second step: denoising the voice signal and removing a mute segment;
a voice feature extraction module; the method is used for executing the third step and the voice feature extraction: extracting various voice features by using a voice signal processing algorithm;
a voice feature selection module; and step four, feature selection: carrying out feature selection by using a CART algorithm, and screening out key features capable of representing frozen gait symptoms;
an AdaBoost classification model training module; and (5) executing the step five, training the model: training an AdaBoost classification model by using a decision tree as a base classifier;
a voice analysis module; and the speech analysis step six is executed: and inputting the feature vector of the voice to be detected into the model to obtain the key feature parameters of the frozen gait symptom of the person to be detected.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. A speech analysis method of key characteristic parameters of frozen gait symptoms of Parkinson's disease based on AdaBoost algorithm is characterized by comprising the following steps:
step one, acquiring a voice signal: collecting continuous and stable vowels of a Parkinson disease patient, and recording whether the Parkinson disease patient has a frozen gait symptom;
step two, preprocessing the voice signal: denoising the voice signal and removing a mute segment;
step three, voice feature extraction: extracting various voice features by using a voice signal processing algorithm;
step four, feature selection: carrying out feature selection by using a CART algorithm, and screening out key features capable of representing frozen gait symptoms;
step five, training the model: training an AdaBoost classification model by using a decision tree as a base classifier;
step six, voice analysis: and inputting the feature vector of the voice to be detected into the model to obtain the key feature parameters of the frozen gait symptom of the person to be detected.
2. The method according to claim 1, wherein the step three is specifically:
extracting a plurality of voice features by using a voice signal processing algorithm, wherein the extracted features comprise: fundamental frequency profile F0_ constant, average fundamental frequency F0_ ave, minimum fundamental frequency F0_ min, maximum fundamental frequency F0_ max, four characteristics Jitter, RAP, PPQ, DDP for measuring fundamental frequency changes, five characteristics shim, shim: APQ3, shim: APQ5, shim: APQ11, shim: DDA, noise harmonic ratio NHR, harmonic noise ratio HNR, cycle period density entropy RPDE, trend fluctuation analysis DFA and gene period entropy PPE; and Mel cepstral coefficients MFCC obtained by converting the speech in the Mel cepstral domain.
3. The method according to claim 1, wherein the step four comprises the following specific processes:
the specific formula of the Gini index (D) of the data set D is:
wherein p iskRepresenting the probability that the sample point belongs to the kth class, wherein K represents K classification problems;
gini index of characteristic aindex(D, a) is defined as:
wherein V represents that V possible values exist in the characteristic a; therefore, the feature value which minimizes the divided kini index is selected as the optimal division feature a*Namely:
a*＝argmaxa∈AGiniindex(D，a)。
4. the method according to claim 1, wherein the step five comprises the following specific processes:
1) initializing a weight distribution of training data, each training sample being initially given the same weight: 1/N;
wherein w1iRepresents the weight at the beginning of the jth training sample, and N represents the total number of samples;
2) performing M iterations, each iteration performing the following steps:
a. make itBy having a weight distribution DmThe training data set is learned to obtain a base classifier Gm(x)：
Gm(x)：χ→{-1，+1}
b. Calculation of Gm(x) Classification error rate e on training data setm：
Wherein G ism(xi) Presentation basis classifier Gm(x) In training data xiClassification result of (1), yiRepresenting training data xiTrue classification of, wmiRepresenting sample x at the m-th iterationiP () represents the probability of a certain event; i () represents that the result is 1 when the event in brackets is true, otherwise the result is 0;
c. calculation of Gm(x) To obtain the weight alpha of the basic classifier in the final classifierm：
Wherein emRepresents Gm(x) A classification error rate on the training data set;
d. updating the weight distribution of the training data set:
Dm+1＝(wm+1，1，wm+1，2，...，wm+1，i，...，wm+1，N)
wherein wm+1，iRepresents the ith training sampleUpdating weight after iteration for m times; zmRepresenting a normalization factor, exp () representing an exponential function with a natural constant e as the base;
3) and combining all the base classifiers to obtain a final classifier, wherein the final classification result is obtained by all the base classifiers through weighted voting:
where f (x) represents a weighted combination of the individual base classifiers, G (x) represents the final classifier, sign () represents a sign function.
5. Speech analysis system for key characteristic parameters of frozen gait symptoms of Parkinson's disease based on the AdaBoost algorithm, characterized in that the system performs the method according to any of claims 1 to 4.
6. The system of claim 5, wherein the system comprises:
the voice signal acquisition module is used for executing the first step and the acquisition of the voice signals: collecting continuous and stable vowels of a Parkinson disease patient, and recording whether the Parkinson disease patient has a frozen gait symptom;
the voice signal processing module is used for executing the second step and the voice signal preprocessing: denoising the voice signal and removing a mute segment;
and the voice feature extraction module is used for executing the third step and the voice feature extraction: extracting various voice features by using a voice signal processing algorithm;
the voice feature selection module is used for executing the fourth step and the feature selection: carrying out feature selection by using a CART algorithm, and screening out key features capable of representing frozen gait symptoms;
and the AdaBoost classification model training module is used for executing the fifth step and training the model: training an AdaBoost classification model by using a decision tree as a base classifier;
and the voice analysis module is used for executing the sixth step and the voice analysis: and inputting the feature vector of the voice to be detected into the model to obtain the key feature parameters of the frozen gait symptom of the person to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111677396.XA CN114299996A (en) | 2021-12-31 | 2021-12-31 | AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111677396.XA CN114299996A (en) | 2021-12-31 | 2021-12-31 | AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114299996A true CN114299996A (en) | 2022-04-08 |
Family
ID=80976045
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111677396.XA Pending CN114299996A (en) | 2021-12-31 | 2021-12-31 | AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114299996A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115954099A (en) * | 2022-11-18 | 2023-04-11 | 临沂中科睿鹤智慧科技有限公司 | Brain stroke correlation quantitative evaluation method based on multi-mode gait parameters |
| CN116312484A (en) * | 2023-05-18 | 2023-06-23 | 南京邮电大学 | Cross-language domain invariant acoustic feature extraction method and system |
-
2021
- 2021-12-31 CN CN202111677396.XA patent/CN114299996A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115954099A (en) * | 2022-11-18 | 2023-04-11 | 临沂中科睿鹤智慧科技有限公司 | Brain stroke correlation quantitative evaluation method based on multi-mode gait parameters |
| CN115954099B (en) * | 2022-11-18 | 2023-09-15 | 临沂中科睿鹤智慧科技有限公司 | Cerebral apoplexy associated quantitative evaluation method based on multi-modal gait parameters |
| CN116312484A (en) * | 2023-05-18 | 2023-06-23 | 南京邮电大学 | Cross-language domain invariant acoustic feature extraction method and system |
| CN116312484B (en) * | 2023-05-18 | 2023-09-08 | 南京邮电大学 | Cross-language domain invariant acoustic feature extraction method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107657964B (en) | Depression auxiliary detection method and classifier based on acoustic features and sparse mathematics | |
| CN111461176B (en) | Multi-mode fusion method, device, medium and equipment based on normalized mutual information | |
| Huang et al. | Exploiting vocal tract coordination using dilated cnns for depression detection in naturalistic environments | |
| JP2021506486A (en) | A method for analyzing cough sounds using disease signatures to diagnose respiratory disease | |
| Travieso et al. | Detection of different voice diseases based on the nonlinear characterization of speech signals | |
| CN114299996A (en) | AdaBoost algorithm-based speech analysis method and system for key characteristic parameters of symptoms of frozen gait of Parkinson's disease | |
| Cheng et al. | Automated sleep apnea detection in snoring signal using long short-term memory neural networks | |
| CN109285551A (en) | Disturbances in patients with Parkinson disease method for recognizing sound-groove based on WMFCC and DNN | |
| Wang et al. | Automatic assessment of pathological voice quality using multidimensional acoustic analysis based on the GRBAS scale | |
| Reddy et al. | The automatic detection of heart failure using speech signals | |
| Nouhaila et al. | An intelligent approach based on the combination of the discrete wavelet transform, delta delta MFCC for Parkinson's disease diagnosis | |
| Chen et al. | Cough detection using selected informative features from audio signals | |
| Benayad et al. | Features selection by genetic algorithm optimization with k-nearest neighbour and learning ensemble to predict Parkinson disease | |
| CN108766462A (en) | A kind of phonic signal character learning method based on Meier frequency spectrum first derivative | |
| Shabber et al. | A review and classification of amyotrophic lateral sclerosis with speech as a biomarker | |
| Ding et al. | Automatic recognition of student emotions based on deep neural network and its application in depression detection | |
| CN114299925A (en) | Method and system for obtaining importance measurement index of dysphagia symptom of Parkinson disease patient based on voice | |
| Cebola et al. | Speech-Based Supervised Learning Towards the Diagnosis of Amyotrophic Lateral Sclerosis. | |
| Melek | Responding to challenge call of machine learning model development in diagnosing respiratory disease sounds | |
| Gidaye et al. | Unified wavelet-based framework for evaluation of voice impairment | |
| Kumar et al. | Parkinson’s Speech Detection Using YAMNet | |
| Képešiová et al. | Comparative Analysis of Advanced Machine Learning Algorithms for Early Detection of Parkinson Disease | |
| Chen et al. | Acoustic characterization and machine prediction of perceived masculinity and femininity in adults | |
| Miliaresi et al. | A deep multimodal voice pathology classifier with electroglottographic signal processing capabilities | |
| Kho et al. | Malaysian cough sound analysis and COVID-19 classification with deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |