Apparatus and method for classificating senile diseases using neural network

Abstract

An apparatus and method for classifying geriatric diseases using a neural network are disclosed. In an apparatus for classifying geriatric diseases using a neural network, an apparatus for classifying geriatric diseases according to one embodiment comprises: a receiver for receiving an utterance generated from a classification target; and a processor for generating a preprocessed utterance by performing preprocessing including filtering on the utterance, extracting an acoustic feature of the classification target by inputting the preprocessed utterance into a first neural network, and generating a classification result for a geriatric disease of the classification target by inputting the acoustic feature into a second neural network.

Description

Technology Field

[0001] The embodiments relate to an apparatus and method for classifying age-related diseases using a neural network. Background Technology

[0003] As population aging accelerates globally, the importance of diagnosis for the prevention of age-related diseases is increasing. The diagnosis of diseases is one of the key tasks of pathology.

[0004] With the advancement of artificial intelligence technology, AI-based disease diagnostic technologies are being developed. Due to the development of machine learning, active attempts are also being made to automate tasks such as image recognition and classification using computer systems.

[0005] In particular, attempts are being made to automate diagnoses performed by skilled medical professionals using deep learning of neural networks, a type of machine learning.

[0006] Diagnosis through deep learning does not merely automate the experience and knowledge of conventionally skilled medical professionals; rather, in that it derives desired answers by identifying characteristic elements through self-learning, it can even detect characteristics of disease factors in images that were previously unknown to skilled medical professionals. (Prior Art Patent Document 1) KR 10-2018-0002234 A

[0007] delete means of solving the problem

[0008] In a geriatric disease classification device using a neural network, the geriatric disease classification device according to one embodiment includes a receiver that receives an utterance generated from a classification target, and a processor that generates a preprocessed utterance by performing preprocessing including filtering on the utterance, extracts an acoustic feature of the classification target by inputting the preprocessed utterance into a first neural network, and generates a classification result for the geriatric disease of the classification target by inputting the acoustic feature into a second neural network.

[0009] The processor can generate the preprocessed utterance by performing the filtering based on whether the utterance is caused by an independent object, the degree of noise included in the utterance, and the amplitude of the utterance.

[0010] The processor can generate the preprocessed utterance by performing low-pass filtering on the utterance.

[0011] The processor can generate the preprocessed utterance by converting the utterance into a mel spectrogram and slicing the mel spectrogram.

[0012] The processor can generate three or more multiple classes to classify the health status of the subject based on the degree of the geriatric disease, and classify the health status of the subject into one of the multiple classes by inputting the acoustic features into the second neural network.

[0013] The above-mentioned geriatric disease includes hearing loss, and the plurality of classes may include healthy, mild, and severe.

[0014] The processor can train the first neural network and the second neural network based on predefined prior training data, and perform fine tuning on the second neural network based on acoustic features output from the trained first neural network.

[0015] The first neural network above includes a convolution layer, a maxpooling layer, and a fully connected layer, and the second neural network above may include a fully connected layer and an activation function.

[0016] In a device for classifying geriatric diseases using a neural network, a method for classifying geriatric diseases according to one embodiment comprises: receiving a speech generated from a classification target; generating a preprocessed speech by performing preprocessing including filtering on the speech; extracting an acoustic feature of the classification target by inputting the preprocessed speech into a first neural network; and generating a classification result for geriatric diseases of the classification target by inputting the acoustic feature into a second neural network.

[0017] The step of generating a preprocessed utterance may include the step of generating the preprocessed utterance by performing the filtering based on whether the utterance is caused by an independent object, the degree of noise included in the utterance, and the amplitude of the utterance.

[0018] The step of generating the preprocessed utterance may include the step of generating the preprocessed utterance by performing low-pass filtering on the utterance.

[0019] The step of generating the preprocessed utterance above can be performed by converting the utterance into a mel spectrogram and slicing the mel spectrogram to generate the preprocessed utterance.

[0020] The step of generating the above classification result may include the step of generating three or more multiple classes for classifying the health status of the subject to classification based on the degree of the above geriatric disease, and the step of classifying the health status of the subject to classification into one of the multiple classes by inputting the above acoustic feature into the second neural network.

[0021] The above-mentioned geriatric disease includes hearing loss, and the plurality of classes may include healthy, mild, and severe.

[0022] The step of generating the above classification result may include the step of training the first neural network and the second neural network based on predefined prior training data, and the step of performing fine tuning on the second neural network based on acoustic features output from the trained first neural network.

[0023] The first neural network above includes a convolution layer, a maxpooling layer, and a fully connected layer, and the second neural network above may include a fully connected layer and an activation function. Brief explanation of the drawing

[0025] FIG. 1 shows a schematic block diagram of a geriatric disease classification device according to one embodiment. Figure 2 is a diagram illustrating the operation of the geriatric disease classification device shown in Figure 1. Figure 3 shows an example of the implementation of the geriatric disease classification device illustrated in Figure 1. Figure 4a shows an example of an acoustic feature vector extracted from a normal group. Figure 4b shows an example of an acoustic feature vector extracted from a mild patient. Figure 4c shows an example of an acoustic feature vector extracted from a severe patient. Figure 5 shows an example of performance depending on whether preprocessing is performed on the geriatric disease classification device illustrated in Figure 1. Figure 6 shows a flowchart of the operation of the geriatric disease classification device illustrated in Figure 1. Specific details for implementing the invention

[0026] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, actual implementations are not limited to the specific embodiments disclosed, and the scope of this specification includes modifications, equivalents, or substitutions included in the technical concept described by the embodiments.

[0027] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0028] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0029] Singular expressions include plural expressions unless the context clearly indicates otherwise. In this document, phrases such as “A or B,” “at least one of A and B,” “at least one of A or B,” “A, B or C,” “at least one of A, B and C,” and “at least one of A, B, or C” may each include any one of the items listed together with the corresponding phrase, or all possible combinations thereof. In this specification, terms such as “comprising” or “having” are intended to designate the existence of the described feature, number, step, action, component, part, or combination thereof, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0030] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this specification.

[0031] As used herein, the term "module" may include a unit implemented in hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be a component formed integrally, or a minimum unit of said component or a part thereof that performs one or more functions. For example, according to one embodiment, a module may be implemented in the form of an application-specific integrated circuit (ASIC).

[0032] As used in this document, the term "part" refers to software or hardware components, such as FPGAs or ASICs, and the "part" performs certain roles. However, the meaning of "part" is not limited to software or hardware. The "part" may be configured to reside in an addressable storage medium or configured to operate one or more processors. For example, the "part" may include components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." Furthermore, the components and "parts" may be implemented to operate one or more CPUs within a device or secure multimedia card. Additionally, '~part' may include one or more processors.

[0033] Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the attached drawings, identical components are given the same reference numeral regardless of the drawing number, and redundant descriptions thereof will be omitted.

[0035] FIG. 1 shows a schematic block diagram of a geriatric disease classification device according to one embodiment.

[0036] Referring to FIG. 1, the geriatric disease classification device (10) can classify geriatric diseases. The geriatric disease classification device (10) can generate a classification result for geriatric diseases by processing utterances from a classification target. The classification result may include a diagnosis result of the geriatric disease. The classification result may include the type of geriatric disease, whether the geriatric disease has occurred, and / or the degree of progression of the geriatric disease.

[0037] Geriatric diseases may include conditions that appear as physical functions decline with age.

[0038] Geriatric diseases may include geriatric syndromes and diseases common in the elderly. Geriatric syndromes may include dementia, delirium, urinary incontinence, gait disturbances, falls, malnutrition, osteoporosis, pressure sores, sleep disorders, and / or terminal pain, the resulting outcome of a combination of disease and the aging process. Diseases common in the elderly may include hypertension, diabetes, stroke, parkinsonism, hypercholesterolemia, and / or heart failure, the prevalence of which increases with age.

[0039] The geriatric disease classification device (10) can classify geriatric diseases using artificial intelligence. Artificial intelligence may refer to a computer system equipped with functions such as learning, reasoning, or judgment. Artificial intelligence may be implemented using a neural network.

[0040] Neural networks (or artificial neural networks) in machine learning and cognitive science can include statistical learning algorithms that mimic biological neurons. A neural network can refer to a model in general that possesses problem-solving capabilities by having artificial neurons (nodes), which form a network through synaptic connections, change the strength of these connections through learning.

[0041] Neurons in a neural network may include a combination of weights or biases. A neural network may include one or more neurons or one or more layers composed of nodes. A neural network can infer a result to be predicted from an arbitrary input by changing the weights of the neurons through learning.

[0042] Neural networks can include deep neural networks. Neural networks include CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), perceptron, multilayer perceptron, FF (Feed Forward), RBF (Radial Basis Network), DFF (Deep Feed Forward), LSTM (Long Short Term Memory), GRU (Gated Recurrent Unit), AE (Auto Encoder), VAE (Variational Auto) Encoder), DAE (Denoising Auto Encoder), SAE (Sparse Auto Encoder), MC (Markov Chain), HN (Hopfield Network), BM (Boltzmann Machine), RBM (Restricted Boltzmann Machine), DBN (Depp Belief Network), DCN (Deep Convolutional Network), DN (Deconvolutional Network), DCIGN (Deep Convolutional Inverse Graphics Network), GAN (Generative Adversarial Network), LSM (Liquid State Machine), ELM (Extreme Learning Machine), ESN (Echo It may include State Network, DRN (Deep Residual Network), DNC (Differentiable Neural Computer), NTM (Neural Turning Machine), CN (Capsule Network), KN (Kohonen Network), VGG (Visual Geometry Group) network, and AN (Attention Network).

[0043] The geriatric disease classification device (10) can be implemented as a printed circuit board (PCB) such as a motherboard, an integrated circuit (IC), or a system on chip (SoC). The geriatric disease classification device (10) can be implemented as an application processor.

[0044] Additionally, the geriatric disease classification device (10) can be implemented in a PC (personal computer), a data server, or a portable device.

[0045] Portable devices can be implemented as laptop computers, mobile phones, smartphones, tablet PCs, mobile internet devices (MID), personal digital assistants (PDA), enterprise digital assistants (EDA), digital still cameras, digital video cameras, portable multimedia players (PMP), personal navigation devices (PND), handheld game consoles, e-books, or smart devices. Smart devices can be implemented as smart watches, smart bands, or smart rings.

[0046] The geriatric disease classification device (10) includes a receiver (100) and a processor (200). The geriatric disease classification device (10) may further include a memory (300).

[0047] The receiver (100) may include a receiving interface. The receiver (100) may receive data from an external source or from memory (300). The receiver (100) may receive utterances generated from a classification target. The receiver (100) may receive utterances from a microphone. The receiver (100) may output the received utterances to a processor (200).

[0048] The processor (200) can process data stored in memory (300). The processor (200) can execute computer-readable code (e.g., software) stored in memory (300) and instructions triggered by the processor (200).

[0049] The "processor (200)" may be a data processing device implemented in hardware having a circuit having a physical structure for executing desired operations. For example, the desired operations may include code or instructions included in a program.

[0050] For example, a data processing device implemented in hardware may include a microprocessor, a central processing unit, a processor core, a multi-core processor, a multiprocessor, an Application-Specific Integrated Circuit (ASIC), and a Field Programmable Gate Array (FPGA).

[0051] The processor (200) can classify geriatric diseases using a neural network. The neural network may include a first neural network and a second neural network.

[0052] The processor (200) can generate a preprocessed utterance by performing preprocessing including filtering on the utterance.

[0053] The processor (200) can generate a preprocessed utterance by performing filtering based on whether the utterance is by an independent object, the degree of noise included in the utterance, and the amplitude of the utterance.

[0054] The processor (200) can generate a preprocessed utterance by performing low-pass filtering on the utterance.

[0055] The processor (200) can convert the utterance into a mel spectrogram. The processor (200) can generate a preprocessed utterance by slicing the mel spectrogram.

[0056] The processor (200) can extract acoustic features of a classification target by inputting a preprocessed utterance into a first neural network. The processor (200) can generate a classification result for an age-related disease of the classification target by inputting the acoustic features into a second neural network. An age-related disease may include hearing loss. The acoustic features may take the form of a feature vector.

[0057] The processor (200) can generate three or more multiple classes to classify the health status of a subject based on the degree of geriatric disease. The multiple classes may include healthy, mild, and severe.

[0058] The processor (200) can classify the health status of a subject to classification into one of a plurality of classes by inputting acoustic features into a second neural network.

[0059] The processor (200) can train the first neural network and the second neural network. The processor (200) can train the first neural network and the second neural network based on predefined pre-training data.

[0060] The processor (200) can perform fine tuning on the second neural network based on acoustic features output from the first neural network that has been learned.

[0061] The first neural network may include a convolution layer, a maxpooling layer, and a fully connected layer.

[0062] The second neural network may include a fully connected layer and an activation function.

[0063] The memory (300) may store data for an operation or the result of an operation. The memory (300) may store instructions (or programs) executable by the processor. For example, the instructions may include instructions for executing the operation of the processor and / or the operation of each component of the processor.

[0064] The memory (300) can be implemented as a volatile memory device or a non-volatile memory device.

[0065] Volatile memory devices can be implemented as DRAM (dynamic random access memory), SRAM (static random access memory), T-RAM (thyristor RAM), Z-RAM (zero capacitor RAM), or TTRAM (Twin Transistor RAM).

[0066] Non-volatile memory devices can be implemented as EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, MRAM (Magnetic RAM), Spin-Transfer Torque (STT)-MRAM, Conductive Bridging RAM (CBRAM), FeRAM (Ferroelectric RAM), PRAM (Phase change RAM), Resistive RAM (RRAM), Nanotube RRAM, Polymer RAM (PoRAM), Nano Floating Gate Memory (NFGM), holographic memory, Molecular Electronic Memory Device, or Insulator Resistance Change Memory.

[0068] Figure 2 is a diagram illustrating the operation of the geriatric disease classification device shown in Figure 1.

[0069] Referring to FIG. 2, a processor (e.g., processor (200) of FIG. 1) can determine whether there is a hearing loss and / or dementia, including senile diseases, by extracting acoustic features from the speech of a subject to classification. The processor (200) can classify the type and degree of hearing loss and senile diseases.

[0070] The processor (200) can achieve a high recognition rate regardless of the language and user's environment by using intuitive acoustic features from the utterance of the classification target.

[0071] The processor (200) can define new acoustic features related to geriatric diseases from large-scale patient data obtained from a hospital. The processor (200) can diagnose geriatric diseases based on the newly defined acoustic features.

[0072] The processor (200) can extract acoustic features from speech. For example, the processor (200) can extract a feature embedding (210) according to hearing loss from speech.

[0073] The processor (200) can generate a classification result by processing the extracted acoustic features using a hearing loss recognition model (230). The classification result may include normal, mild, and severe.

[0074] A processor (e.g., processor (200) of FIG. 1) can store audio data containing the utterance of a classification target in memory (e.g., memory (300) of FIG. 1). The processor (200) can receive and store audio data from a patient suffering from varying degrees of geriatric disease over an arbitrary period.

[0075] The processor (200) can receive audio data based on a predefined data collection processor from a normal group, mild or severe patients and store it in memory (300).

[0076] The processor (200) can generate a preprocessed utterance by performing filtering based on whether the utterance is by an independent object, the degree of noise included in the utterance, and the amplitude of the utterance.

[0077] The processor (200) can generate a preprocessed speech by performing filtering on the stored audio data based on whether it is the speech of an independent speaker, the presence or absence of noise, or the amplitude of the audio data.

[0078] The processor (200) can perform preprocessing using various filtering methods. The processor (200) can generate a preprocessed utterance by performing low-pass filtering on the utterance.

[0079] The processor (200) can extract acoustic features from speech and classify geriatric diseases based on the extracted acoustic features. Since the processor (200) predicts geriatric diseases using the acoustic features of the speaker, it can predict geriatric diseases with a high recognition rate even for classification targets with different language and environment.

[0080] The processor (200) can generate generalized acoustic features by processing the speech of patients suffering from geriatric diseases and the speech of a normal group using CNN and DNN.

[0081] The processor (200) can extract generalized acoustic features using large-capacity training data. The processor (200) can train a neural network by using a predefined audioset as training data.

[0082] The processor (200) can define a feature vector using a first neural network. For example, the first neural network may include a VGG network.

[0084] Figure 3 shows an example of the implementation of the geriatric disease classification device illustrated in Figure 1.

[0085] Referring to FIG. 3, the processor (e.g., the processor (200) of FIG. 1) may include a preprocessor (310), a feature extraction network (330), and a classifier (350).

[0086] The preprocessor (310) can generate a preprocessed utterance by performing preprocessing that includes filtering on the utterance.

[0087] The preprocessor (310) can generate a preprocessed utterance by performing filtering based on whether the utterance is from an independent object, the degree of noise included in the utterance, and the amplitude of the utterance.

[0088] The preprocessor (310) can generate a preprocessed utterance by performing low-pass filtering on the utterance.

[0089] The preprocessor (310) can convert the utterance into a mel spectrogram. The preprocessor (310) can generate a preprocessed utterance by slicing the mel spectrogram.

[0090] The preprocessor (310) can convert raw audio data containing raw waveforms into a Mel spectrogram.

[0091] The processor (200) can extract acoustic features of a classification target by inputting a preprocessed utterance into a first neural network. The first neural network may include a feature extraction network (330).

[0092] The feature extraction network (330) can be implemented as a VGG network. The feature extraction network (330) can extract acoustic features from preprocessed speech. The processor (200) can define features optimized for hearing loss by defining new features using the VGG network and performing fine tuning by retraining the DNN.

[0093] The feature extraction network (330) may include a convolution layer, a max pooling layer, and a fully connected layer.

[0095] The processor (200) can generate a classification result for an age-related disease of a classification target by inputting acoustic features into a second neural network. The second neural network may include a classifier (350). The classifier (350) may include a DNN.

[0096] The classifier (350) can generate three or more multiple classes to classify the health status of the subject based on the degree of geriatric disease. The classifier (350) can divide the classification results into three classes. For example, the multiple classes may include healthy, mild, and severe.

[0097] The classifier (350) can be implemented as a Support Vector Machine (SVM) or a DNN.

[0098] The classifier (350) can classify the health status of a subject to classification into one of a plurality of classes based on acoustic features.

[0099] The classifier (350) may include a fully connected layer and an activation function. The activation function may include a sigmoid function, a hyperbolic tangent function, a Rectified Linear Unit (ReLU), Leaky ReLU, PreLU, Exponential Linear Unit (ELU), maxout and / or softmax function.

[0100] The processor (200) can train a feature extraction network (330) and a classifier (350). The processor (200) can train the feature extraction network (330) and the classifier (350) based on predefined pre-training data.

[0101] The processor (200) can perform fine-tuning of the classifier (350) based on acoustic features output from the learned feature extraction network (330).

[0103] Figure 4a shows an example of an acoustic feature vector extracted from a normal group, Figure 4b shows an example of an acoustic feature vector extracted from a mild patient, and Figure 4c shows an example of an acoustic feature vector extracted from a severe patient.

[0104] Referring to FIGS. 4a through 4c, a processor (e.g., processor (200) of FIG. 1) can extract acoustic features (e.g., acoustic feature vectors) from the utterance of a classification target. As can be seen in FIGS. 4a through 4c, the normal group, mild patients, and severe patients may have different weight values.

[0105] The processor (200) can improve classification performance by classifying geriatric diseases using different weight values ​​corresponding to normal groups, mild patients, and severe patients.

[0107] Figure 5 shows an example of performance depending on whether preprocessing is performed on the geriatric disease classification device illustrated in Figure 1.

[0108] Referring to FIG. 5, the performance of an age-related disease classification device (e.g., the age-related disease classification device (10) of FIG. 1) can be verified using a 5-fold cross-validation method.

[0109] In the example of FIG. 5, Chapter 1 may be a specific sound (e.g., Ah~) used to determine whether there is hearing loss. Chapter 7 may be audio data obtained during the process of reading a text given to the classification target.

[0110] The geriatric disease classification device (10) can identify geriatric diseases by processing voice audio markers through a machine learning method. The geriatric disease classification device (10) can classify geriatric diseases with higher accuracy compared to subjective judgments made by a person (e.g., an audiologist) by objectively classifying geriatric diseases.

[0111] As can be seen in Fig. 5, the geriatric disease classification device (10) can show an accuracy of 90 or higher for all data of various chapters.

[0113] Figure 6 shows a flowchart of the operation of the geriatric disease classification device illustrated in Figure 1.

[0114] Referring to FIG. 6, a receiver (e.g., receiver (100)) can receive an utterance generated from a classification target (610). A processor (e.g., processor (200)) can generate a preprocessed utterance by performing preprocessing including filtering on the utterance (630).

[0115] The processor (200) can generate a preprocessed utterance by performing filtering based on whether the utterance is by an independent object, the degree of noise included in the utterance, and the amplitude of the utterance.

[0116] The processor (200) can generate a preprocessed utterance by performing low-pass filtering on the utterance.

[0117] The processor (200) can convert the utterance into a mel spectrogram. The processor (200) can generate a preprocessed utterance by slicing the mel spectrogram.

[0118] The processor (200) can extract acoustic features of the classification target by inputting the preprocessed utterance into the first neural network (650). The processor (200) can generate a classification result for the geriatric disease of the classification target by inputting the acoustic features into the second neural network. Geriatric diseases may include hearing loss. The acoustic features may take the form of a feature vector.

[0119] The processor (200) can generate three or more multiple classes to classify the health status of a subject based on the degree of geriatric disease. The multiple classes may include healthy, mild, and severe.

[0120] The processor (200) can classify the health status of a subject to classification into one of a plurality of classes by inputting acoustic features into a second neural network (670).

[0121] The processor (200) can train the first neural network and the second neural network. The processor (200) can train the first neural network and the second neural network based on predefined pre-training data.

[0122] The processor (200) can perform fine tuning on the second neural network based on acoustic features output from the first neural network that has been learned.

[0123] The first neural network may include a convolution layer, a maxpooling layer, and a fully connected layer.

[0124] The second neural network may include a fully connected layer and an activation function.

[0126] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware and software components. For example, the devices, methods, and components described in the embodiments may be implemented using a general-purpose computer or a special-purpose computer, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. In addition, other processing configurations, such as parallel processors, are also possible.

[0127] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave in order to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on computer-readable recording media.

[0128] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may store program instructions, data files, data structures, etc., either individually or in combination, and the program instructions recorded on the medium may be those specifically designed and configured for the embodiment or those known and available to those skilled in the art of computer software. 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, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

[0129] The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0130] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based thereon. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0131] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 A geriatric disease classification device using a neural network, comprising: a receiver that receives an utterance generated from a classification target; and a processor that generates a preprocessed utterance by performing preprocessing including filtering on the utterance, extracts an acoustic feature of the classification target by inputting the preprocessed utterance into a first neural network, and generates a classification result for the geriatric disease of the classification target by inputting the acoustic feature into a second neural network, wherein the geriatric disease includes hearing loss, the acoustic feature includes an acoustic feature vector, and the processor generates a plurality of classes including healthy, mild, and severe to classify the health status of the classification target based on the degree of the geriatric disease, and classifies the geriatric disease using different weight values ​​corresponding to the acoustic feature vectors of patients corresponding to healthy, mild, and severe. Claim 2 A geriatric disease classification device according to claim 1, wherein the processor generates the preprocessed utterance by performing the filtering based on whether the utterance is by an independent object, the degree of noise included in the utterance, and the amplitude of the utterance. Claim 3 A geriatric disease classification device according to claim 1, wherein the processor generates the preprocessed utterance by performing low-pass filtering on the utterance. Claim 4 A geriatric disease classification device according to claim 1, wherein the processor converts the utterance into a mel spectrogram and generates the preprocessed utterance by slicing the mel spectrogram. Claim 5 In claim 1, the processor classifies the health status of the subject to classification into one of the plurality of classes by inputting the acoustic feature into the second neural network. Claim 6 delete Claim 7 A geriatric disease classification device according to claim 1, wherein the processor trains the first neural network and the second neural network based on predefined prior training data, and performs fine tuning on the second neural network based on acoustic features output from the trained first neural network. Claim 8 A geriatric disease classification device according to claim 7, wherein the first neural network comprises a convolution layer, a maxpooling layer, and a fully connected layer, and the second neural network comprises a fully connected layer and an activation function. Claim 9 A method for classifying geriatric diseases using a neural network, comprising: receiving an utterance generated from a classification target; generating a preprocessed utterance by performing preprocessing including filtering on the utterance; extracting an acoustic feature of the classification target by inputting the preprocessed utterance into a first neural network; and generating a classification result for the geriatric disease of the classification target by inputting the acoustic feature into a second neural network, wherein the geriatric disease includes hearing loss, the acoustic feature includes an acoustic feature vector, and the step of generating the classification result comprises generating a plurality of classes including healthy, mild, and severe to classify the health status of the classification target based on the degree of the geriatric disease, and classifying the geriatric disease using different weight values ​​corresponding to the acoustic feature vectors of patients corresponding to healthy, mild, and severe. Claim 10 A method for classifying geriatric diseases according to claim 9, wherein the step of generating a preprocessed utterance comprises the step of generating the preprocessed utterance by performing filtering based on whether the utterance is by an independent object, the degree of noise included in the utterance, and the amplitude of the utterance. Claim 11 A method for classifying geriatric diseases according to claim 9, wherein the step of generating the preprocessed utterance comprises the step of generating the preprocessed utterance by performing low-pass filtering on the utterance. Claim 12 A method for classifying geriatric diseases according to claim 9, wherein the step of generating the preprocessed utterance comprises converting the utterance into a mel spectrogram and generating the preprocessed utterance by slicing the mel spectrogram. Claim 13 In claim 9, the step of generating the classification result comprises the step of classifying the health status of the subject to classification into one of the plurality of classes by inputting the acoustic feature into the second neural network, a method for classifying geriatric diseases. Claim 14 delete Claim 15 A method for classifying geriatric diseases according to claim 9, wherein the step of generating the classification result comprises: a step of training the first neural network and the second neural network based on predefined prior training data; and a step of performing fine tuning on the second neural network based on acoustic features output from the trained first neural network. Claim 16 A method for classifying geriatric diseases according to claim 15, wherein the first neural network comprises a convolution layer, a maxpooling layer, and a fully connected layer, and the second neural network comprises a fully connected layer and an activation function.

Images