TWI622980B - Disease detecting and classifying system of voice - Google Patents

Abstract

The voice disease detection and classification system of the invention has an application program, a voice database, a medical history database and a servo terminal, and the application can transmit a measured medical history data and a measured voice data to a servo terminal, and the voice database and medical history The database can provide a plurality of comparison voice data and a plurality of comparison medical history data to the servo terminal, so that the servo terminal compares the medical history data and the plurality of comparison voice data according to the measured medical history data and the measured voice data respectively. Sorting and screening to form a diagnostic data, so that anyone can understand the voice status through the diagnostic data, and then can know whether the vocal cord is healthy, and can further determine whether it should be treated.

Description

Voice disease detection classification system

The invention relates to a voice disease detection and classification system for detecting a voice state, in particular to a voice disease detection and classification system capable of detecting a voice state by anyone.

With the evolution of the times, medical technology is also gradually improving. Further studies have pointed out that the health of the patient's throat does affect the patient's voice. Among them, the pathology that currently affects the patient's voice includes: structural lesions (eg, vocal cord nodules, Polyps and cysts), tumors and neurogenic diseases (eg vocal cord paralysis and spastic dysphonia).

When a patient influences the sound due to pathological phenomena in the vocal cords, the patient usually goes to the hospital to seek a doctor for treatment. At present, it is necessary to use the laryngoscope to detect the throat condition of the patient through the otolaryngology specialist, and then to determine which patient the patient receives. A condition, however, when using an endoscope to detect, the tube of the laryngoscope needs to be inserted into the patient's throat, and the patient may feel uncomfortable because of a foreign body in the throat, and the patient passes through the laryngoscope. Testing often also costs a medical bill.

In addition, when the patient's voice changes due to illness, not all patients will go to the hospital immediately for medical treatment, and some patients will wait for the disease to heal itself, so that the patient can not only detect whether the disease has healed or even delay the medical treatment. The condition is serious and it takes more time and money to treat.

In this way, today's patients can not only detect the health of the throat to achieve the preventive effect, even when the health of the throat is problematic, it must take time and money to the hospital for testing, and must also endure the uncomfortable detection, and nowadays There are still many inconveniences in detecting the health of the throat.

The main purpose of the present invention is that anyone can detect their own voice condition at any time, and thus can immediately know the health status of the vocal cords. If the health condition of the vocal cords is known to be problematic, the medical treatment can be immediately sought to improve the chance of cure.

To achieve the above objective, the voice disease detection and classification system of the present invention mainly comprises a medical history database, a voice database, an application program and a servo terminal.

The medical history database is configured to provide a plurality of comparison medical history data, and the sound data database is configured to store a plurality of original voice data, and a first extraction module is provided, wherein the first extraction module is configured to The original voice data is extracted to form a plurality of comparative voice data corresponding to the comparison medical history data, and the voice database further has a label module, and the label module can mark at least one of each of the compared voice data Pathology, and the above label module can be labeled with nodules, polyps, cysts, tumors, cancer, atrophy, tremor, paralysis and vocal cords.

The application can be installed on the electronic device and electrically connected to the servo terminal, and the application is used to form a measured medical history data transmitted to the servo terminal, and can be transmitted to the servo through the user's voice. The measured voice data of the terminal, wherein the application has a second extraction module, and the second extraction module can extract the input voice data input by the user to form the measured voice data.

The servo terminal has a medical history questionnaire corresponding to the measured medical history data, a comparison classification module, and a processing module electrically connected to the comparison classification module, wherein the comparison classification module is used for simultaneously connecting to the processing module. The medical history database and the voice database, wherein the processing module is electrically connected to the application program, wherein the servo terminal receives the plurality of comparison medical history data, a plurality of comparative voice data, the measured medical history data, and the measured The voice data can be combined and screened according to the above-mentioned medical history data and the measured voice data to form a diagnostic data, respectively, according to the plurality of comparative medical history data and the plurality of comparative voice data.

In a preferred embodiment, the processing module compares the measured medical history data with the measured voice data, so that the comparison classification module separately transmits the plurality of comparative medical history data, the plurality of comparative voice data, and the foregoing The medical history data and the above-mentioned measured voice data are used to form the above diagnostic data.

In another preferred embodiment, the comparison classification module forms a plurality of classification data according to different pathological classifications of the plurality of comparison medical history data and the plurality of comparison voice data, and the comparison classification module and One of the processing modules forms the diagnostic data according to the plurality of classified data according to the measured medical history data and the measured voice data, and, when the diagnostic data is formed, the comparison classification module or the processing module The group will stop screening for the remaining classification data.

However, in the foregoing two embodiments, the first extraction module extracts the original voice data by using a Mel frequency cepstrum coefficient to form the comparison voice data, and the second extraction module transmits the Mel frequency cepstrum coefficient. The input voice data is extracted to form the measured voice data, and the comparison classification module is selected from one of the system modules in the supervised machine learning, such as One or more system modules of the decision tree, adaptive enhanced classification, nearest neighbor method, support vector machine, Gaussian mixture model or deep neural network are classified.

It can be seen from the foregoing description that the application program can transmit the measured voice data and the measured medical history data to the servo terminal, so that the servo terminal respectively according to the received voice data and the measured medical history data to the voice database and The data in the medical history database is screened and judged whether there is a disease, so that as long as the user installs the application on the electronic device, any person can input his or her own measured voice data and the measured medical history data through the electronic device at any time to detect Your own voice status, you can immediately know if the vocal cords are healthy. If you know that there is a problem with the health of the vocal cords, you can seek medical treatment immediately to improve the chance of cure.

1‧‧‧Voice Disease Detection and Classification System

10‧‧‧Servo terminal

11‧‧‧Comparative classification module

12‧‧‧Processing module

20‧‧‧ medical history database

30‧‧‧Audio database

31‧‧‧First extraction module

32‧‧‧ label module

40‧‧‧Application

41‧‧‧Second extraction module

50‧‧‧Electronic devices

51‧‧‧Input module

52‧‧‧ Display screen

1 is a block diagram of a voice disease detection and classification system according to the present invention; FIG. 2 is a schematic diagram of the use of the voice disease detection and classification system of the present invention in a first preferred embodiment; FIG. 3 is a schematic diagram of using an input module; 4 is a schematic diagram of the use of the voice sickness detection and classification system of the present invention in the second preferred embodiment.

In order to facilitate further understanding and understanding of the structure, use and features of the present invention, a preferred embodiment will be described in detail with reference to the drawings as follows: In the preferred embodiment, the voice disease detection and classification system 1 of the present invention is mainly composed of a server terminal 10, a medical history database 20, a voice database 30, and an application 40.

The servo terminal 10 has a comparison classification module 11, a processing module 12, and a medical history questionnaire. The comparison classification module 11 is configured as a plurality of machine learning modules, such as a decision tree and a random forest. Adaptive enhanced classification (Adaboost), nearest neighbor method (K Nearest-neighbor), Support Vector Machine (SVM), Gaussian Mixture Model (GMM) or Deep Neural Networks (Deep Neural Networks) One or more of the DNNs are used in combination, and are electrically connected to the processing module 12, and the medical history database 20 stores a plurality of comparison medical history data, and is electrically connected to the comparison classification module 11 of the servo terminal 10. The voice database 30 is electrically connected to the comparison classification module 11 and has a first extraction module 31 and a label module 32, and the voice database 30 has a plurality of original voice data therein. In the embodiment, each of the medical history data records that the patient has a change in the voice due to the disease, and each of the voice data is composed of sound waves, and each of the voice data The wave is not the same as the sound wave of another voice data, in which part of the voice data records the voice as normal, while the other part records the voice because of the disease (eg, nodules, polyps, cysts, tumors, cancer, atrophy, tremors) The numbness or the vocal folds are not normal, and the application 40 is connected to the processing module 12 of the servo terminal 10 and has a second extraction module 41.

Referring to FIG. 2 to FIG. 3 , an application device 40 is installed on an electronic device 50 in a specific application, so that the electronic device 50 can provide data to the servo terminal 10 or receive data transmitted by the server 10 . The electronic device 50 has an input module 51 and a display screen 52. The input module 51 and the display screen 52 together form a touch screen, and the electronic device 50 can be set as a mobile phone, a computer or a tablet.

Next, the server 10 transmits the medical history questionnaire to the application 40, so that the display screen 52 of the electronic device 50 displays the content of the medical history questionnaire, so that the user can follow the medical history. The questionnaire is used to generate a measured medical history data through the input module 51, and the user inputs his own voice through the input module 51 to form an input voice data. In this embodiment, the user first records the voice formation. a recording file, and inputting the recording file to the input module 51 to form the input voice data. However, the user can directly input the voice into the input module 51 through the microphone built in the electronic device 50 to form the input voice. Information, wherein the medical history data includes user gender, age, personal medical history, voice related symptoms, lifestyle habits (such as smoking, drinking), severity self-assessment, occupational category and voice status, and duration of symptoms Or a variety of related content.

Next, the input module 51 of the electronic device 50 transmits the input voice data and the measured medical history data to the application 40, and the application 40 transmits the measured medical history data to the processing module 12 of the servo terminal 10. The second extraction module 41 of the application 40 extracts the input voice data through a Mel-scale Frequency Cepstral Coefficients (MFCC) to form a measured voice transmitted to the processing module 12. The data is processed by the processing module 12 to the comparison classification module 11 that transmits the measured medical history data and the measured arpeggio data to the servo terminal 10.

In addition, the medical history database 20 transmits the plurality of comparison medical history data to the comparison classification module 11 of the servo terminal 10, and the first extraction module 31 of the voice data library 30 transmits the above by the Mel frequency cepstrum coefficient. An original voice data is extracted to form a plurality of comparison voice data transmitted to the comparison classification module 11, wherein when the comparison voice data is formed, the label module 32 can mark each of the comparison voice data Normal or abnormal voice, and at least one or more pathologies such as nodules, polyps, cysts, tumors, cancer, atrophy, tremors, paralysis, or vocal folds can be classified for abnormal voices.

When the comparison classification module 11 of the servo terminal 10 receives the above-mentioned medical history data, the measured arpeggio data, the plurality of medical history data, and the plurality of comparison voice data, the comparison classification module 11 is separately determined according to the above-mentioned test. The medical history data and the input voice data respectively classify and filter the plurality of comparative medical history data and the plurality of comparative voice data to form a diagnostic data transmitted to the processing module 12, and the processing module 12 detects the diagnostic data. The content is formed to display a display material transmitted to the display screen 52, so that the display screen 52 displays whether the user's voice state is healthy through the display data.

In this embodiment, when the display screen 52 of the electronic device 50 displays that the throat state of the user is healthy through the display data, the user can know that the voice state of the user is clear by the display screen 52, and vice versa. When the processing module 12 detects the diagnostic data and knows that the user's voice state is unhealthy due to a tumor or other pathology, the display screen 52 displays the user's voice state as unhealthy and possible diseases through the display data. In turn, the user can seek medical treatment as soon as possible.

Referring to FIG. 4, in the second preferred embodiment, the difference from the first preferred embodiment lies in the manner in which the diagnostic data is formed. As for the servo terminal 10, the medical history database 20, the voice database 30, and the application program. The connection relationship between 40 is the same as that of the first preferred embodiment, and will not be described herein.

When the processing module 12 of the servo terminal 10 receives the measured medical history data and the measured voice data, the processing module 12 of this embodiment does not receive the above-mentioned receiving module 12 as in the first preferred embodiment. The medical history data and the measured voice data are transmitted to the comparison sorting module 11 of the servo terminal 10.

When the comparison classification module 11 of the servo terminal 10 receives a plurality of comparison medical history data and a plurality of comparison voice data, the comparison classification module 11 is based on nodules, polyps, cysts, tumors, cancer, atrophy, tremors, paralysis and paralysis. Eight pathological records of vocal folds are classified into eight categories of data by comparing multiple medical history data with a plurality of comparative voice data. Subsequently, the comparison classification module 11 sequentially transmits eight classified data to The processing module 12 of the servo terminal 10 causes the processing module 12 to classify and filter the classified data according to the measured medical history data and the measured voice data to jointly form the diagnostic data, wherein the comparison classification module 11 may The single alignment history data and the single comparison voice data are simultaneously classified in different classification data. In the preferred embodiment, when the diagnosis data is formed, the processing module 12 stops performing screening comparison for the remaining classification data. .

In this embodiment, the processing module 12 classifies and screens the classified data according to the measured medical history data and the measured voice data to form the diagnostic data for convenience of description, and may also be used by the comparison classification module 11 The classified data is classified and screened according to the above-mentioned measured medical history data and the measured voice data to jointly form the above-mentioned diagnostic data.

The above-mentioned embodiments are not intended to limit the scope of the present invention, and various simple modifications and modifications made by those skilled in the art in accordance with the scope of the invention and the description of the invention should still be made without departing from the spirit of the invention. It is included in the scope of the following patent application.

Claims (11)

A voice disease detection and classification system comprises: a medical history database, which can provide a plurality of comparison medical history data; a voice database can provide a plurality of comparative voice data corresponding to the comparison medical history data; a servo terminal, Electrically connected to the medical history database and the voice database, and having a medical history questionnaire; and an application that can be installed on the electronic device, electrically connected to the server terminal, and the application program transmits the above-mentioned medical history questionnaire to the above The measured medical history data of the servo terminal, and then the user's voice is used to form a measured arpeggio data transmitted to the servo terminal; wherein the servo terminal receives the plurality of comparative medical history data and the plurality of comparative audio data And the above-mentioned medical history data and the above-mentioned tested voice data, and can filter the above-mentioned plurality of medical history data and a plurality of comparative voice data according to the measured medical history data and the measured voice data respectively to form a transmission to the above The diagnostic data of the application. The voice detection and classification system of claim 1, wherein the voice database is provided with a first extraction module, and the first extraction module is configured to store a plurality of original voices stored in the voice database. The data extraction forms the above-mentioned comparative voice data. The arpeggio disease detection and classification system according to claim 2, wherein the first extraction module extracts the original voice data by using a Mel frequency cepstrum coefficient to form the comparison voice data. The arpeggio disease detection and classification system according to claim 1, wherein the voice database further has a tag module, and the tag module can mark at least one pathology for each of the compared voice data, and the foregoing The label module can be labeled with nodules, polyps, cysts, tumors, cancer, atrophy, tremors, paralysis, and vocal folds. The sounding disease detection and classification system of claim 1, wherein the application has a second extraction module, and the second extraction module is configured to extract input audio data input by a user to form the above Tested voice data. The arpeggio disease detection and classification system according to claim 5, wherein the second extraction module extracts the input voice data by using a Mel frequency cepstrum coefficient to form the measured voice data. The vocal cord disease detection and classification system according to claim 1, wherein the servo terminal has a comparison classification module and a processing module electrically connected to the comparison classification module, and the comparison classification The module is electrically connected to the medical history database and the voice database, and the processing module is electrically connected to the application. The arpeggio disease detection and classification system according to claim 7, wherein the upper processing module compares the measured medical history data with the measured voice data, so that the comparison classification module separately passes the plurality of comparison medical records. The data, the plurality of comparative voice data, the above-mentioned measured medical history data, and the above-mentioned measured voice data are used to form the above-mentioned diagnostic data. The arpeggio disease detection and classification system according to claim 7, wherein the comparison classification module forms a plurality of classifications according to different pathological classifications of the plurality of comparison medical history data and the plurality of comparative voice data. And the one of the comparison classification module and the processing module forms the diagnostic data according to the measured medical history data and the measured voice data. The vocal cord disease detection and classification system according to claim 9, wherein when the diagnostic data is formed, the comparison classification module or the processing module stops screening and screening the remaining classification data. The arpeggio disease detection and classification system according to claim 7, wherein the comparison classification module is selected from the group consisting of a decision tree, a random forest, an adaptive enhanced classification, a nearest neighbor method, a support vector machine, and a Gaussian mixture model. Or classify one or more system modules of a deep neural network.