Device and method for managing swallowing disorder on basis of voice analysis

Abstract

A swallowing disorder management method by which a device manages a swallowing disorder on the basis of voice analysis, according to one embodiment of the present invention, comprises the steps of: (a) receiving a voice obtained using a first aspiration test of a subject; (b) inputting the voice into a voice analysis model so as to perform a first aspiration analysis for analyzing whether aspiration of the subject has occurred; and (c) guiding a swallowing-assisted electrical stimulation meal or a free meal by using an electrical stimulation module according to the result of the first aspiration analysis performed using the voice analysis model, wherein the voice obtained by the first aspiration test includes voices before and after the subject ingests a first type of food, and the step of performing the first aspiration analysis determines whether aspiration occurs on the basis of characteristic information of the voices before and after ingestion of the food.

Description

Device and method for managing swallowing disorders based on voice analysis

[0001] The present invention relates to a device and method for managing speech disorders based on voice analysis.

[0002] Dysphagia is a condition in which food cannot be properly swallowed through the oral cavity, pharynx, and esophagus. It can be divided into residual and aspiration states. Residual means that food remains in the oral cavity, pharynx, or esophagus, while aspiration means that food passes into the airway. Recently, video fluoroscopic swallowing studies (VFSS) have been used to diagnose dysphagia. However, these tests are only available in medical institutions equipped with fluoroscopy equipment, and their radiation exposure makes regular monitoring difficult. Therefore, a method that can diagnose dysphagia and classify the condition regardless of location is needed.

[0003] Dysphagia also makes swallowing difficult, making it difficult to consume food and drink. This limits nutritional intake and impacts quality of life (QOL). Dysphagia is also a major factor significantly increasing the risk of aspiration pneumonia in older adults. Rehabilitation therapy is one of the various treatments being implemented to improve dysphagia. However, rehabilitation therapy requires consistent, daily practice to be effective, and long-term patient participation and commitment are essential. Therefore, methods to improve accessibility to dysphagia rehabilitation training are needed.

[0004] A prior patent document related to this is Republic of Korea Patent No. 10-2591028 (Title of invention: Electrical stimulation treatment device and method for dysphagia).

[0005] Some embodiments of the present invention have been created to solve the above-described problems, and the purpose of the present invention is to provide a voice analysis-based swallowing disorder management device and method that applies sequential electrical stimulation suitable for a subject by using the results of analyzing the voice generated during an aspiration test.

[0006] However, the technical tasks that this embodiment seeks to accomplish are not limited to the technical tasks described above, and other technical tasks may exist.

[0007] As a technical means for achieving the above-described technical task, a method for managing dysphagia performed by a voice analysis-based dysphagia management device according to an embodiment of the present invention comprises the steps of (a) receiving a voice by a first suction test of a subject; (b) performing a first suction analysis for analyzing whether or not the subject has suctioned by inputting the voice into a voice analysis model; and (c) guiding a swallowing-assisting electrical stimulation meal or free-flow meal using an electrical stimulation module according to a result of the first suction analysis by the voice analysis model, wherein the voice by the first suction test includes a voice by the subject before and after ingestion of a first type of food, and the step of performing the first suction analysis determines whether or not suction has occurred based on characteristic information of the voice before and after ingestion of the food.

[0008] A method for providing a swallowing-assisting electrical stimulation meal mode performed by a voice analysis-based swallowing disorder management device according to another embodiment of the present invention includes the steps of: providing a swallowing-assisting electrical stimulation to a subject, allowing the subject to eat a meal based on a first food; performing a first voice analysis to determine whether the subject is normal without a swallowing disorder based on a first voice collected from the subject immediately after consuming the first food; continuing the swallowing-assisting electrical stimulation meal mode when the first voice analysis result is normal; and providing a swallowing-assisting electrical stimulation to the subject, when the first voice analysis result is abnormal, guiding the subject to eat a meal based on a second food having a higher viscosity than the first food.

[0009] According to another embodiment of the present invention, a voice analysis-based swallowing disorder management device includes a memory in which a swallowing disorder management program is stored; and a processor that executes the program stored in the memory, wherein the program receives a voice from a subject through a first suction test, inputs the voice into a voice analysis model, and performs a first suction analysis to analyze whether the subject has aspirated, and according to the result of the first suction analysis by the voice analysis model, guides a swallowing assistance electrical stimulation meal or free meal using an electrical stimulation module, wherein the voice from the suction test includes a voice from the subject before and after ingestion of a first type of food, and when the program performs the first suction analysis, determines whether aspiration has occurred based on characteristic information of the voice before and after ingestion of the food.

[0010] According to any one of the above-described problem solving means of the present invention, the present invention can provide a dysphagia rehabilitation training method suitable for the swallowing state of a subject based on the results of the dysphagia analysis using a voice analysis model.

[0011] In addition, the present invention can improve the accessibility of existing limited rehabilitation treatment by providing customized rehabilitation training for patients with dysphagia through an electrical stimulation module.

[0012] In addition, the present invention can provide swallowing-assisted electrical stimulation meals through an electrical stimulation module even when free eating is impossible.

[0013] In addition, the present invention can provide swallowing assistance electrical stimulation suitable for the eating type of the subject, and can control the application of swallowing assistance electrical stimulation by constantly monitoring the voice of the subject during eating.

[0014] FIG. 1 is a configuration diagram illustrating a speech analysis-based speech disorder management device according to one embodiment of the present invention.

[0015] FIG. 2 is a flowchart for explaining a method for managing swallowing disorders performed by a speech analysis-based swallowing disorder management device according to one embodiment of the present invention.

[0016] FIG. 3 is a drawing for explaining a method for providing electrical stimulation rehabilitation training depending on whether or not a subject is sucking according to one embodiment of the present invention.

[0017] FIG. 4 is a flowchart for explaining a swallowing assistance electrical stimulation eating motion performed by a speech analysis-based swallowing disorder management device according to one embodiment of the present invention.

[0018] FIG. 5 is a drawing for explaining a swallowing assistance electrical stimulation eating motion suitable for a subject's eating type according to one embodiment of the present invention.

[0019] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily implement them. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In the drawings, irrelevant parts have been omitted for clarity of description, and similar reference numerals have been used throughout the specification to indicate similar elements.

[0020] Throughout the specification, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected" but also the case where it is "electrically connected" with another element in between. Furthermore, when a part is said to "include" a component, this should be understood to mean that, unless specifically stated to the contrary, it may include other components rather than excluding them, and does not preclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0021] In this specification, a 'unit' includes a unit realized by hardware or software, and a unit realized using both, and one unit may be realized using two or more pieces of hardware, and two or more units may be realized by one piece of hardware. Meanwhile, the 'unit' is not limited to software or hardware, and the 'unit' may be configured to be on an addressable storage medium or may be configured to play one or more processors. Accordingly, as an example, the 'unit' includes 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 functionality provided within the components and 'units' may be combined into a smaller number of components and 'units', or further separated into additional components and 'units'. Additionally, components and '~parts' may be implemented to regenerate one or more CPUs within the device.

[0022] A network is a connection structure that enables information exchange between each node, such as terminals and devices, and includes a local area network (LAN), a wide area network (WAN), the Internet (WWW), wired and wireless data communication networks, telephone networks, and wired and wireless television communication networks. Examples of wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, Bluetooth communication, infrared communication, ultrasonic communication, visible light communication (VLC), LiFi, etc.

[0023] FIG. 1 is a configuration diagram illustrating a speech analysis-based speech disorder management device according to one embodiment of the present invention.

[0024] Referring to FIG. 1, a swallowing disorder management device (100) may include a communication module (110), a memory (120), a processor (130), a database (140), an electrical stimulation module (150), a sensor module (160), and a voice analysis model. Here, the voice analysis model is basically built into the swallowing disorder management device (100), but is not limited thereto, and a form in which the voice analysis model is installed in an external computer device or an external server (200) to process the results is also possible.

[0025] The communication module (110) may be a device including hardware and software necessary for transmitting and receiving signals, such as control signals or data signals, with other network devices via wired or wireless connections. For example, through an application installed on a user terminal, such as a smart phone, the operation of the swallowing disorder management device (100) may be controlled, or information regarding pulse sequences constituting electrical stimulation provided by the swallowing disorder management device (100) may be received through the communication module (110). In addition, various types of information regarding the usage history of the swallowing disorder management device (100) may be transmitted to the user terminal, an external computing device, or an external server through the communication module (110).

[0026] The memory (120) may record a swallowing disorder management program. The swallowing disorder management program receives the voice from the first suction test of the subject, inputs the voice into a voice analysis model to perform a first suction analysis to analyze whether the subject is sucking, and guides swallowing assistance electrical stimulation meal or free meal using the electrical stimulation module (150) based on the result of the first suction analysis by the voice analysis model. Here, the memory (120) may include a magnetic storage media or a flash storage media in addition to a volatile storage device requiring power to maintain stored information, but the scope of the present invention is not limited thereto.

[0027] The memory (120) may store a separate program, such as an operating system for processing and controlling the processor (130), or may perform a function for temporarily storing input or output data.

[0028] The processor (130) executes a speech disorder management program (hereinafter, “program”) stored in the memory (120) and provides a function of controlling the hardware of the speech disorder management device (100) according to the execution of the program. That is, the processor (130) can perform hardware control functions such as a necessary file system, memory allocation, network, basic library, timer, device control (display, media, input device, 3D, etc.), and other utilities according to the execution of the program.

[0029] The processor (130) executes a program to receive a voice from the subject's first suction test, input the received voice into a voice analysis model to perform a first suction analysis to analyze whether the subject is sucking, and guides swallowing assistance electrical stimulation meal or free meal using an electrical stimulation module according to the result of the first suction analysis by the voice analysis model.

[0030] In addition, the processor (130) may guide the subject to eat a meal based on a first food while providing a swallowing-assisting electrical stimulus to the subject, perform a first voice analysis to determine whether the subject is normal without a swallowing disorder based on a first voice collected from the subject immediately after consuming the first food, and if the result of the first voice analysis is normal, continue the swallowing-assisting electrical stimulus eating mode, and if the result of the first voice analysis is abnormal, provide the subject with a swallowing-assisting electrical stimulus while guiding the subject to eat a meal based on a second food having a higher viscosity than the first food. Meanwhile, a specific description of a swallowing disorder management method according to the execution of the program will be described later with reference to FIGS. 3 to 5.

[0031] The processor (130) may include any type of device capable of processing data. For example, it may refer to a hardware-embedded data processing device having a physically structured circuit to perform a function expressed by a code or command included in a program. Examples of such hardware-embedded data processing devices include processing devices such as a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated circuit (ASIC), and a field programmable gate array (FPGA), but the scope of the present invention is not limited thereto.

[0032] The database (140) stores or provides data required for the speech disorder management device (100) under the control of the processor (130). For example, the database (140) may store results generated during the execution of the speech disorder management program. This database (140) may be included as a separate component from the memory (120), or may be constructed in a portion of the memory (120).

[0033] The electrical stimulation module (150) outputs a bipolar or unipolar pulse signal, and the frequency of the pulse, pulse width, etc. can be controlled by a program, etc. The electrical stimulation module (150) can include an output terminal for outputting an electrical stimulation signal for each channel. In addition, a plurality of electrical stimulation pads that transmit electrical stimulation to the skin of the subject are attached around the swallowing muscles of the subject, and electrical stimulation can be provided to the subject through a wire connecting the output terminal and the electrical stimulation pad.

[0034] The electrical stimulation module (150) can provide electrical stimulation signals of multiple channels, such as 3 channels, 4 channels, 5 channels, or 6 channels. For example, a single channel can be configured to include a wire that transmits electrical stimulation supplied from the electrical stimulation module (150), a first electrical stimulation pad attached to a first point of a muscle, a second electrical stimulation pad attached to a second point of the muscle, and a muscle electrically connecting the first electrical stimulation pad and the second electrical stimulation pad.

[0035] For example, the electrical stimulation module (150) may be configured with 4 channels. For example, the 4-channel electrical stimulation module (150) may be configured with a first channel for stimulating the gastrocnemius muscle (gastric), a second channel for stimulating the stylohyoid muscle, a third channel for stimulating the thyrohyoid muscle (TH), and a fourth channel for stimulating the sternohyoid muscle, the scapula hyoid, and the sternothyroid muscle (StH). At this time, the electrical stimulation pads of the first channel may be respectively placed on the right suprahyoid muscle (SH), and the electrical stimulation pads of the second channel may be respectively placed on the left suprahyoid muscle (SH). Here, the electrical stimulation pads of each channel may be positioned above the hyoid bone at intervals of 1 cm and behind the mandible, but are not limited thereto, and may be positioned at intervals narrower than 1 cm depending on the shape of the person's face if the jaw is too narrow. Additionally, the electrical stimulation pads of the third channel are placed on each of the upper poles of the thyroid cartilage, and the electrical stimulation pads of the fourth channel are placed on the upper part of the sternocleidomastoid muscle and can be placed below the thyroid cartilage.

[0036] The electrical stimulation module (150) can sequentially stimulate all muscles necessary for the swallowing process (swallowing process) using electrical stimulation pads attached to each location of the aforementioned swallowing muscles. Accordingly, the program can control the electrical stimulation module (150) to provide a swallowing assistance electrical stimulation meal mode, a pre-test, and an electrical stimulation rehabilitation training mode based on the analysis results of the voice analysis model described below.

[0037] The sensor module (160) may include at least a microphone. For example, the sensor module (160) may be integrated into the swallowing disorder management device (100) in a built-in form, or may be externally integrated via a separate cable. The user may position the sensor module (160) close to the subject's oral cavity or airway to collect the subject's voice before and after swallowing food. This is not limited to this, and voices such as the subject's coughing sounds generated during electrical stimulation rehabilitation training or the entire meal process may also be collected.

[0038] The voice analysis model, when presented with a subject's voice from an aspiration test, outputs the results of the aspiration analysis. Specifically, when presented with the subject's voice before and after swallowing, it can output the results of a voice analysis that determines whether the voice is normal and free of swallowing disorders.

[0039] In the aspiration test, the subject is asked to utter a specific speech sound for a predetermined period of time before swallowing food, and the pre-swallowing sound is recorded. After swallowing the food, the subject is asked to utter the specific speech sound for a predetermined period of time, and the post-swallowing sound is recorded. For example, a speech sound such as "ah" is continuously uttered for 5 seconds before and after swallowing the food. In addition, the food can be swallowed with low viscosity, such as water or porridge, but this can be varied. In this way, when the recorded pre-swallowing sound and post-swallowing sound are input into the voice analysis model, the result regarding the presence or absence of aspiration is output based on the characteristic information of the two sounds.

[0040] In this way, the voice analysis model can output both the presence or absence of aspiration for the aspiration test voice and the normality of the swallowing voice, but is not limited thereto, and can be divided into a first voice analysis model that outputs the presence or absence of aspiration and a second voice analysis model that outputs the normality of the swallowing voice without a swallowing disorder.

[0041] For example, the first voice analysis model may be a model trained using a training data set labeled with correct answers and voice data of a predetermined time interval that matches identification information for the voices of patients with dysphagia or normal patients, and when the voice of the examinee through an aspiration test is input, an output indicating a normal or aspiration state can be inferred as a result of the aspiration analysis. In addition, the first voice analysis model may be trained using training data that collects multiple unit data sets including voices before and after swallowing food by having the examinee utter specific speech sounds for a predetermined time before and after swallowing food, and labeled with correct answers indicating whether or not aspiration occurred for each unit data set. According to this configuration, when voices before and after ingestion of food are input, the first voice analysis model can output whether or not aspiration occurred based on the feature information thereof. Here, for the convenience of the user, the voice before consuming food is recorded once and then stored and managed as a reference voice, and only the voice after consuming food is collected and used in the inference process.

[0042] As another example, the second voice analysis model is a model trained based on a learning data set labeled with voice data of a predetermined time interval that matches identification information for the voices of patients with dysphagia or normal patients and correct labels, and when the first voice or the second voice collected from the subject immediately after consuming the first food or the second food is input, an output indicating a normal or abnormal state can be inferred as a result of performing the first voice analysis or the second voice analysis.

[0043] In addition, the second voice analysis model may be trained using learning data that collects multiple unit data sets including the pre-swallowing sound of the first food or the second food and the post-swallowing sound of the first food or the second food after having the subject consume the first food or the second food, and labels the correct answer data indicating whether the condition of the subject is normal without a swallowing disorder for each unit data set. Here, each unit data set means a sound such as 'gulp' generated by the subject in the process of swallowing food, unlike the sound of the first voice analysis model making a specific speech sound for a predetermined period of time. According to this configuration, when the pre-swallowing sound of the first food or the second food and the post-swallowing sound of the first food or the second food are input, the second voice analysis model can output whether the subject is normal without a swallowing disorder based on the feature information.

[0044] For example, when the voice collected during the subject's aspiration test is input, the first voice analysis model can output a label including "normal," "residual," or "aspiration" as a result of whether there is aspiration. At this time, normal is a normal state without aspiration or residue, residual is a residual state in which food remains in the pharynx, and aspiration means an aspiration state in which food passes into the airway. In addition, when the second voice analysis model inputs the subject's voice before swallowing the first food or the second food and the voice after swallowing the first food or the second food collected during the operation of the swallowing assistance electrical stimulation eating mode, the second voice analysis model can output a label including "normal" or "abnormal" as a result of whether there is no swallowing disorder. At this time, normal is a normal state without swallowing disorder, and abnormal is an abnormal state in which there is swallowing disorder.

[0045] Meanwhile, the voice analysis model may be updated by an external learning device that preprocesses a voice dataset containing the voices of patients with dysphagia or normal patients, builds a learning dataset based on the preprocessed dataset, and trains the voice analysis model based on the learning dataset. A detailed description of the learning device that trains the voice analysis model will be provided below.

[0046] FIG. 2 is a flowchart for explaining a method for managing swallowing disorders performed by a speech analysis-based swallowing disorder management device according to one embodiment of the present invention.

[0047] Referring to FIG. 2, the processor (130) receives a voice from a subject's first suction test (S110), inputs the voice into a voice analysis model to perform a first suction analysis to analyze whether the subject is sucking (S120), and, based on the result of the first suction analysis by the voice analysis model, guides swallowing assistance electrical stimulation meal or free meal using an electrical stimulation module (150) (S130).

[0048] First, in the first suction test, the subject is asked to utter a specific speech sound for a predetermined period of time before swallowing the first type of food, so that the pre-swallowing sound is recorded, and after swallowing the food, the subject is asked to utter the same speech sound as the specific speech sound for a predetermined period of time, so that the post-swallowing sound is recorded. For example, the subject is asked to continuously utter a speech sound such as "ah" for 5 seconds before and after swallowing the food. In addition, the food may be water or low-viscosity food such as porridge, but this can be varied.

[0049] Next, the pre-swallowing and post-swallowing vocalizations are input into a voice analysis model to analyze whether the subject has aspirated. As previously explained, the voice analysis model can determine whether aspiration has occurred based on characteristic information from the pre-swallowing and post-swallowing vocalizations.

[0050] In addition, the processor (130) can record the history of the results of the suction analysis in the database (140). At this time, the first suction test is to collect a predetermined sound made by the subject while consuming low-viscosity food using the sensor module (160).

[0051] FIG. 3 is a drawing for explaining a method for providing electrical stimulation rehabilitation training depending on whether or not a subject is sucking according to one embodiment of the present invention.

[0052] Specifically, referring to FIG. 3, the program performs a first suction analysis on a voice from a first suction test (S10), and if no aspiration occurs as a result of the first suction analysis, guides a free meal (S11), guides the subject to perform a second suction test after the subject finishes the free meal, performs a second suction analysis that inputs the voice from the second suction test into a voice analysis model (S20), and if no aspiration occurs as a result of the second suction analysis, outputs a meal prescription including a free meal or a swallowing-assisted electrical stimulation meal (S21), and if aspiration occurs as a result of the second suction analysis, provides an electrical stimulation rehabilitation training mode (S22).

[0053] In addition, the program performs a first suction analysis on the voice by the first suction test (S10), and if aspiration occurs as a result of the first suction analysis, guides a pre-test for swallowing-assisted electrical stimulation eating using an electrical stimulation module (S12), guides a third suction test to be performed after the pre-test, performs a third suction analysis to input the voice by the third suction test into a voice analysis model (S30), and if aspiration does not occur as a result of the third suction analysis, provides a swallowing-assisted electrical stimulation eating mode (S31), and if aspiration occurs as a result of the third suction analysis, provides an electrical stimulation rehabilitation training mode (S32).

[0054] Next, the program can perform the second suction analysis (S20) described above after providing the swallowing assistance electrical stimulation eating mode (S31), and can perform the first suction analysis (S10) described above again after providing the electrical stimulation rehabilitation training mode (S32).

[0055] For example, when the subject performs the first suction test at step S10, the program may perform the first suction analysis using a voice analysis model. Here, as previously discussed, the suction test involves having the subject utter a specific speech sound before and after consuming water or low-viscosity food such as porridge. At this time, the program may collect voice data of the subject performing the suction test using the sensor module (160).

[0056] For example, if the results of the first suction analysis (S10) are normal, the program may use the user interface to guide the subject to free-feeding without electrical stimulation. After the subject has completed free-feeding, the program may guide the subject to a second suction test for the second suction analysis (S20).

[0057] For example, at step S20, if no aspiration occurs as a result of the second aspiration analysis, the program may record a meal prescription history for the next meal in the database (140) (S21). In this case, since no aspiration occurred in either the first or second aspiration tests, the meal prescription for the next meal may be recorded as a free meal.

[0058] In addition, in step S20, if aspiration occurs as a result of the second aspiration analysis, an electrical stimulation rehabilitation training mode may be provided (S22). In addition, guidance may be given that swallowing assistance electrical stimulation eating is possible with caution at the next meal. Here, the electrical stimulation rehabilitation training mode may include sequential electrical stimulation rehabilitation training and sequential electrical stimulation swallowing assistance training. In this case, the sequential electrical stimulation rehabilitation training means strengthening the swallowing muscles by repeatedly applying sequential electrical stimulation (SH → TH → StH) to the swallowing muscles for 30 to 60 minutes. The sequential electrical stimulation swallowing assistance training means training swallowing by sequentially applying electrical stimulation to the swallowing muscles while swallowing a single serving of water or porridge.

[0059] In addition, the program can record the history of the results of the aspiration analysis in the database (140). For example, the number of aspirations per individual subject, the training prescription history (the number of swallowing assistance training sessions or electrical stimulation rehabilitation training sessions), and the meal prescription history for the next meal (free eating or swallowing assistance electrical stimulation meal) can be managed. Accordingly, the swallowing disorder management device (100) of the present invention can provide customized rehabilitation training and swallowing disorder history management to patients with swallowing disorders by recording the history of the results of the aspiration analysis of the subject.

[0060] Meanwhile, in step (S10), if the first suction analysis (S10) determines that suction has occurred, a preliminary test for swallowing-assisted electrical stimulation can be provided to the subject using the electrical stimulation module (150) (S12). This is an optional step to test whether the subject can perform the swallowing-assisted electrical stimulation eating mode that will be provided in the future, and it is also possible to provide the swallowing-assisted electrical stimulation eating mode directly without performing this step. Thereafter, the program can guide the subject to a third suction test for the third suction analysis (S30).

[0061] At this time, the pre-test for swallowing assistance electrical stimulation means assisting swallowing by sequentially applying electrical stimulation to the swallowing muscles according to the subject's muscle contraction pattern when the subject swallows water (5 cc or 3 cc) or porridge once (not during a meal). For example, during the swallowing motion, the suprahyoid muscle (SH) contracts to lift the tongue and push the food in, the thyrohyoid muscle (TH) contracts to pull the thyroid bone down to widen the pharynx and secure space for the food to move, and the sternothyroid muscle (StH) contracts to pull the thyroid bone down further to widen the pharynx and allow the food to enter the esophagus. In other words, during a normal swallowing motion, the swallowing muscles contract in the order of the suprahyoid muscle (SH), thyrohyoid muscle (TH), and sternothyroid muscle (StH). Accordingly, a pre-test for swallowing-assisted electrical stimulation can be performed as a test performed before actually performing the swallowing-assisted electrical stimulation meal or as auxiliary training for the subject's adaptation, and provides the effect of enabling smooth swallowing by sequentially applying electrical stimulation to the swallowing muscles while swallowing a single serving of water or porridge, thereby sequentially contracting the muscles related to swallowing. Meanwhile, the electrical stimulation rehabilitation training mode includes training to strengthen the swallowing muscles by repeatedly applying sequential electrical stimulation (SH → TH → StH) to the above-described swallowing muscles for 30 to 60 minutes.

[0062] For example, at step S30, if the result of the third aspiration analysis shows that no aspiration occurred, the program may determine that swallowing-assisted electrical stimulation eating is possible, provide the subject with a swallowing-assisted electrical stimulation eating mode (S31), and guide the subject to perform a second aspiration analysis (S20) again after finishing the meal using the swallowing-assisted electrical stimulation eating mode. If the result of the third aspiration analysis shows that aspiration occurred, the program may determine that swallowing-assisted electrical stimulation eating is impossible, provide the subject with an electrical stimulation rehabilitation training mode (S32), and perform the first aspiration analysis (S10) again.

[0063] Hereinafter, a method for providing a swallowing-assisted electrical stimulation eating mode will be described.

[0064] FIG. 4 is a flowchart for explaining a swallowing assistance electrical stimulation eating motion performed by a speech analysis-based swallowing disorder management device according to one embodiment of the present invention, and FIG. 5 is a drawing for explaining a swallowing assistance electrical stimulation eating motion suitable for a subject's eating type according to one embodiment of the present invention.

[0065] Referring to FIG. 4, the program provides a swallowing-assisting electrical stimulus to the subject, guides the subject to eat a meal based on a first food, and performs a first voice analysis to determine whether the subject is normal without a swallowing disorder based on a first voice collected from the subject immediately after consuming the first food (S210). If the result of the first voice analysis is normal, the swallowing-assisting electrical stimulus eating mode is continued (S220). If the result of the first voice analysis is abnormal, the program provides a swallowing-assisting electrical stimulus to the subject while guiding the subject to eat a meal based on a second food having a higher viscosity than the first food (S230).

[0066] As described above, the voice analysis model may be constructed by collecting multiple unit data sets including the first voice or the second voice collected for a predetermined time before and after swallowing the first food or the second food, and using learning data labeled with correct answer data indicating whether or not the swallowing disorder is normal for each unit data set. For example, when the first voice or the second voice is input, the voice analysis model may output a correct answer label including normal and abnormal as a voice analysis result for determining whether or not the swallowing disorder is normal based on the feature information of the voice immediately after ingestion of the first food or the second food.

[0067] For example, the first food may be a food composed of a low-viscosity liquid such as water, and the second food may be a food composed of a low-viscosity semi-liquid such as porridge.

[0068] Additionally, the program can record the results of voice analysis for swallowing disorders in a database (140). Accordingly, the training prescription history (number of swallowing assistance training sessions or electrical stimulation rehabilitation training sessions) and the meal prescription history (e.g., free-range feeding recommendations) can be managed according to the subject's individual meal type (first food / second food).

[0069] Here, the program can continuously collect the sound of the subject before swallowing, after swallowing, or coughing sound that occurs while providing swallowing assistance electrical stimulation using the sensor module (160). Accordingly, the program can continuously monitor the cough sound of the subject throughout the entire meal process and urgently stop the swallowing assistance electrical stimulation applied through the electrical stimulation module (150). Here, the swallowing assistance electrical stimulation refers to an electrical stimulation that helps to swallow strongly and quickly by sequentially applying electrical stimulation to each swallowing muscle during the meal.

[0070] Referring to FIG. 5, in one embodiment, the program provides a swallowing-assisting electrical stimulus to a subject, guides the subject to eat a meal based on a first food, and performs a first voice analysis to determine whether the subject is normal without a swallowing disorder based on a first voice collected from the subject immediately after consuming the first food (S210), and if the result of the first voice analysis is normal, the swallowing-assisting electrical stimulus eating mode is continued (S220), and if the result of the first voice analysis is abnormal, the program provides a swallowing-assisting electrical stimulus to the subject, and guides the subject to eat a meal based on a second food having a higher viscosity than the first food (S230).

[0071] For example, step S210 can input the first voice into a voice analysis model to determine whether swallowing disorder is normal based on the characteristic information of the voice immediately after ingestion of the first food.

[0072] Hereinafter, if the result of the first voice analysis is normal, the process after the step (S220) of continuing the swallowing assistance electrical stimulation meal mode based on the first food will be described.

[0073] For example, after the step of continuing the swallowing assistance electrical stimulation meal mode (S220), the program outputs a user interface for receiving response input information including whether discomfort occurred while the subject is eating the first food-based meal, thereby receiving a response input from the user (S240), and if response input information indicating that discomfort was felt is input through the user interface, continuous voice analysis is performed to determine whether the first voice continuously collected from the subject is normal without swallowing disorder for a preset number of times (S250), and if response input information indicating that discomfort was not felt is not input, intermittent voice analysis is performed to determine whether the first voice intermittently collected from the subject is normal without swallowing disorder for a preset number of times (S260).

[0074] For example, step S240 can input discomfort experienced by the subject due to swallowing-assisted electrical stimulation during eating. For example, response input information regarding the presence or degree of discomfort can be received through a user interface displayed on the display.

[0075] For example, step S250 inputs the first voice into a voice analysis model to determine whether or not the swallowing disorder is normal based on the characteristic information of the voice immediately after ingestion of the first food. However, as the examinee swallows multiple times, multiple first voices collected are continuously input into the voice analysis model to output correct answer data indicating whether or not the swallowing disorder is normal a preset number of times.

[0076] In step S250, if all of the continuous voice analysis results are normal, free eating can be guided (S270), and if at least one of the continuous voice analysis results is abnormal, the electrical stimulation rehabilitation training mode can be guided (S280). For example, if the continuous voice analysis results are a preset number of 5 times, the program can continuously collect the first voice generated by the examinee's swallowing 5 times, and continuously input the collected 5 first voices into the voice analysis model to output 5 correct data (normal / abnormal). At this time, if all 5 correct data are output as normal, the program can guide free eating, and if at least one of the 5 correct data is output as abnormal, the program can guide the electrical stimulation rehabilitation training mode.

[0077] As another example, step S260 inputs the first voice into a voice analysis model to determine whether or not the swallowing disorder is normal based on the characteristic information of the voice immediately after ingestion of the first food. However, as the examinee swallows multiple times, multiple first voices collected are intermittently input into the voice analysis model to output correct answer data indicating whether or not the swallowing disorder is normal a preset number of times.

[0078] In step S260, if all intermittent voice analysis results are normal, free eating may be guided (S270), and if at least one intermittent voice analysis result is abnormal, the electrical stimulation rehabilitation training mode may be guided (S290). For example, if the intermittent voice analysis result is a preset number of three times, the program may intermittently collect three first voices generated by the examinee's swallowing, and intermittently input the three collected first voices into the voice analysis model to output three preset correct answer data (normal / abnormal). At this time, if all three correct answer data are output as normal, the program may guide free eating, and if at least one of the three correct answer data is output as abnormal, the program may guide the electrical stimulation rehabilitation training mode.

[0079] As illustrated in FIG. 5, the reason why the subject of the present invention first proceeds with a meal based on the first food (water type) is that if aspiration occurs, the food passes into the lungs, and water is safer than porridge as it causes less inflammation. Therefore, the subject proceeds with a meal based on the first food first.

[0080] Below, the process after the step (S230) of guiding the user to proceed with a meal based on the second food when the result of the first voice analysis is abnormal will be described.

[0081] For example, in step S230, the program may have the subject eat a meal based on the second food, and perform a second voice analysis to determine whether the subject is normal without a swallowing disorder based on the second voice collected from the subject immediately after consuming the second food (S310), and if the result of the second voice analysis is normal, the swallowing assistance electrical stimulation eating mode may be continued (S320), and if the result of the second voice analysis is abnormal, the program may guide the subject to an electrical stimulation rehabilitation training mode (S330).

[0082] For example, step S310 can input the second voice into a voice analysis model to determine whether or not the swallowing disorder is normal based on the characteristic information of the voice immediately after consuming the second food.

[0083] After step S320, the program may receive response input information including whether discomfort occurs while the subject is eating the second food-based meal (S340), and if discomfort is input, perform continuous voice analysis to determine whether the second voice continuously collected from the subject is normal without swallowing disorder for a preset number of times (S350), and if discomfort is not input, perform intermittent voice analysis to determine whether the second voice intermittently collected from the subject is normal without swallowing disorder for a preset number of times (S360).

[0084] For example, step S350 inputs a second voice into a voice analysis model to determine whether swallowing disorder is normal based on the characteristic information of the voice immediately after ingestion of the second food. As the subject swallows multiple times, multiple second voices collected are continuously input into the voice analysis model to output correct answer data indicating whether swallowing disorder is normal a preset number of times.

[0085] In step S350, if all the results of the continuous voice analysis are normal, free eating is guided (S370), and if at least one result of the continuous voice analysis is abnormal, the electrical stimulation rehabilitation training mode can be guided (S380). At this time, the process of outputting the correct answer data according to the results of the continuous voice analysis based on the second voice is identical to the process of outputting the correct answer data based on the first voice in step S250.

[0086] As another example, step S360 inputs a second voice into a voice analysis model to determine whether or not swallowing disorder is normal based on the characteristic information of the voice immediately after ingestion of the second food. However, as the examinee swallows multiple times, multiple second voices collected are intermittently input into the voice analysis model to output correct answer data indicating whether or not swallowing disorder is normal a preset number of times.

[0087] At step S360, if all intermittent voice analysis results are normal, free eating is guided (S370), and if at least one intermittent voice analysis result is abnormal, electrical stimulation rehabilitation training mode can be guided (S390). At this time, the process of outputting correct data based on the intermittent voice analysis results based on the second voice is identical to the process of outputting correct data based on the first voice in step S260.

[0088] Below, a learning method for training a voice analysis model is described.

[0089] In one embodiment, a learning method for training a voice analysis model may be performed by a computing device or server including a processor and memory.

[0090] The processor preprocesses a speech dataset containing speech sounds of patients with hearing impairment or normal patients.

[0091] Here, the speech dataset may include a first speech dataset including waveforms of sounds uttered by a patient with a dysphagia or a normal patient during a preset time period before and / or after a meal.

[0092] In other words, the voice dataset may include sounds uttered by a patient with a residual dysphagia, a patient with aspiration dysphagia, a normal patient diagnosed as normal, or a normal person for a preset period of time as the first voice dataset.

[0093] For example, the first voice dataset may include a voice dataset that records sounds made by a patient with a swallowing disorder or a normal patient after eating food, repeated twice or more for 5 seconds, so as to secure a certain length.

[0094] The speech dataset may include a second speech dataset. The second speech dataset is a large-scale acoustic dataset, which may include speech datasets collected in various acoustic environments and situations. The second speech dataset may include various speech features suitable for training a speech classification model. For example, the second speech dataset may include a large-scale crowdsourced acoustic dataset collected for training an artificial intelligence model.

[0095] Specifically, the processor can convert the speech dataset from a first format to a second format. The processor can convert the speech dataset from the first format to a second format suitable for training a speech analysis model. Here, the first format may refer to the format of the original speech dataset, and the second format may refer to the format of the preprocessed speech dataset.

[0096] For example, the processor can perform preprocessing to change the channels of a voice dataset. Specifically, the processor can perform preprocessing to combine information from the left and right channels of the voice dataset and integrate them into a single channel. In other words, the processor can change the channels of a stereoscopic voice dataset into a monoscopic voice dataset, taking into account factors such as file size, processing speed, recognition accuracy, and consistency.

[0097] For example, the processor may perform preprocessing to unify the file format of each data included in a voice dataset. Specifically, the processor may unify the various file formats of each data included in the voice dataset, such as wav, m4a, and mp3, into the mp3 format, taking into account factors such as file size, processing speed, ease of processing, and uniformity.

[0098] For example, the processor can encode the format of a speech dataset from a basic format to a hierarchical format. Specifically, the processor can structure the components included in the basic format of the speech dataset to express relationships between speech data. As a specific example, the processor can perform preprocessing to convert the initial format of the speech dataset to HDF5 format.

[0099] Here, a hierarchical format can be structured through multiple components. For example, a hierarchical format can be structured through three components. Specifically, a hierarchical format can be organized into groups, datasets, and attributes.

[0100] A group is a concept similar to a folder, representing a category that forms a hierarchical structure. A dataset represents actual valid data and can exist as a multidimensional array within a group. Attributes can be expressed as metadata, providing additional information that describes the dataset.

[0101] At this time, the attribute may include a first attribute, a second attribute, and a third attribute. For example, the attribute may include a first attribute representing the file name of the voice dataset so as to identify the voice dataset.

[0102] The attribute may include a second attribute indicating the file format of the audio dataset. For example, if the audio data in one of the audio datasets is in mp3 format, the second attribute may be described to indicate that it is in mp3 format.

[0103] The attribute may include a third attribute that represents the correct value of the speech dataset as a label. For example, the third attribute may include a correct class for speech classification training. Here, the label of the correct class may be expressed as a number encoded corresponding to "normal" or "abnormal," respectively. Alternatively, the label of the correct class may be expressed as a number encoded corresponding to "normal," "residual," and "aspiration," respectively.

[0104] As another example, a hierarchical format can be structured with four components. For example, a speech dataset might be structured with an identifier, pre-meal speech data, post-meal speech data, and a label.

[0105] Here, the identifier may refer to an anonymized number of the subject in the speech dataset. The pre-meal speech data may refer to speech data in the form of a number array recorded before a meal. The post-meal speech data may refer to speech data in the form of a number array recorded after a meal. The label may refer to the correct answer class of the speech dataset. In this case, the speech dataset may be composed of various combinations of pre-meal speech data and post-meal speech data. For example, the pre-meal speech data and post-meal speech data may be combined so that the channels are the same or different. The pre-meal speech data and post-meal speech data may be combined in segment units with the same or different time intervals.

[0106] The processor can transform the voice dataset from the time domain into a frequency domain spectrogram. This transformation into a spectrogram can be accomplished using the Fourier transform. In other words, the processor can visualize the voice dataset as a two-dimensional image called a spectrogram through the Fourier transform.

[0107] The processor can scale the spectrogram to resemble human hearing characteristics using a scaling filter modeled after human hearing characteristics. In other words, the processor can scale the spectrogram to have nonlinear characteristics.

[0108] The processor can use scaling filters to densely distribute filters in the low-frequency region and sparsely distribute filters in the high-frequency region to impart nonlinear features to the spectrogram.

[0109] At this time, the scaling filter includes multiple filters, and the frequency domain can be expressed by drawing a triangle based on the center values ​​of the previous filter and the filter. For example, the frequency band located inside the triangle can be set to 1, and the frequency band located outside the triangle can be set to 0.

[0110] The processor can reduce noise in the voice dataset by emphasizing high-frequency components included in the voice dataset and attenuating low-frequency components.

[0111] The processor can improve the quality of the speech dataset by taking the difference between the previous speech data and the current speech data and then amplifying the intensity of the high frequency components.

[0112] The processor builds a training dataset based on the preprocessed dataset.

[0113] To efficiently utilize a limited training dataset, the processor can construct training data by dividing the preprocessed dataset into multiple sub-datasets. Preferably, the multiple sub-datasets have the same size and contain no overlapping data.

[0114] The processor can divide the speech dataset into multiple sub-datasets by dividing the test data and training data into a preset ratio. For example, if the training data and test data are in an 8:2 ratio, the processor can divide the speech dataset into 5n sub-datasets.

[0115] The processor can construct a voice dataset as a learning dataset by designating at least one of the plurality of sub-datasets as test data and designating the remaining one or more sub-datasets as learning data.

[0116] For example, in the above ratio, the processor can divide the voice dataset into five sub-datasets and designate one of the five sub-datasets as a test dataset. The processor can repeat this process to designate another set as test data and the remaining four sub-datasets as training data, thereby constructing training data.

[0117] The processor can train a speech analysis model based on a training dataset. The processor can be trained to classify the status of a swallowing disorder based on the input speech dataset. Specifically, the processor can classify the presence or absence of a swallowing disorder based on the input speech dataset. For example, the processor can classify the status of a swallowing disorder based on the input speech dataset as normal or abnormal.

[0118] As another example, the processor can be trained to classify a speech dataset into a normal state with no dysphagia, a residual state with food remaining in the pharynx, or an aspiration state with food passing into the airway.

[0119] At this time, the processor can store and manage the pre-meal voice data and post-meal voice data in separate instances to independently process the pre-meal voice data and post-meal voice data.

[0120] The processor can determine the call order for pre-meal and post-meal voice data by sequentially calling separate instances. Furthermore, the processor can independently analyze pre-meal and post-meal voice data by propagating separate instances through the layer array defined by the sequential module.

[0121] At this time, the channels of the last layer of each layer array that processes pre-meal and post-meal voice data are connected to each other, so that the processor can merge and analyze the characteristics of pre-meal and post-meal voice data.

[0122] The processor can perform encoding on pre-meal and post-meal speech data. The processor can perform customized encoding on pre-meal and post-meal speech data.

[0123] At this time, the processor can perform customized encoding based on an encoding layer generated based on at least one of the inverted residual settings included in the voice analysis model and the factors of the last channel.

[0124] Meanwhile, the number of output channels of the last convolutional layer included in the voice analysis model can be set to single or multiple in terms of processing various types of input data.

[0125] The processor can train the speech analysis model using a combination of pairs of one or more sub-data sets designated as test data and one or more sub-data sets designated as training data among the plurality of sub-data sets.

[0126] For example, the processor can train a voice analysis model by designating one set of five sub-datasets as test data and the remaining four sub-datasets as training data, and then train the voice analysis model by designating another set as test data and the remaining four sub-datasets as training data. In this case, the processor can train the voice analysis model 5n times using the five pairs created from the test data and training data.

[0127] At this time, it is desirable for the sub-dataset to consist of training data generated from the same source. For example, it is desirable for the sub-dataset to consist of speech datasets obtained from the same speaker (either a patient with dysphagia or a normal patient).

[0128] Here, the speech analysis model may include a first model that extracts sequence features of a speech dataset and a second model that extracts temporal features of the speech dataset based on the sequence features.

[0129] The first model may include a transformer utilizing an attention mechanism. Specifically, the first model may divide input data into multiple attention heads and calculate attention for each head. Subsequently, the first model may perform a linear transformation on the attention of each head to obtain a final attention result. From this, the first model may extract sequence features of the input data based on the final attention result.

[0130] Here, the input data of the first model may mean data in which a spectrogram converted to correspond to a voice dataset is divided into multiple patches, and each of the divided patches is tagged with location information.

[0131] The second model may include a pre-trained Convolutional Neural Network (CNN)-based model. For example, the second model may include a pre-trained MobileNet V3-based model.

[0132] The second model can be trained by adjusting hyperparameters that determine the depth of the model. At this time, the parameters determined based on the hyperparameters in the second model may include at least one of the number of mel filters (n_mels), sampling frequency (sr), window function length (win_length), size of overlapping windows (hopsize), window size (n_fft), size of frequency masking used in the mel spectrogram (freqm), size of time masking used in the mel spectrogram (timem), minimum frequency (fmin) to be converted in the mel spectrogram, and maximum frequency (fmax).

[0133] The second model may include a first convolutional layer. Here, the first convolutional layer may be generated based on at least one of the number of corresponding input channels and the number of output channels of the inverted residual block. Here, the inverted residual block may be the first inverted residual block among one or more inverted residual blocks included in the second model.

[0134] The second model includes one or more inverted residual blocks. The one or more inverted residual blocks can be generated based on inverted residual settings, including the number of output channels, kernel size, and stride. The inverted residual settings can be provided when adding the configured inverted residual settings to the layer list.

[0135] The second model may include a second convolutional layer. In this case, the number of output channels of the second convolutional layer may be set to a preset multiple of the number of input channels (e.g., 6 times).

[0136] Below, we will explain the voice analysis model, and omit the explanation of components that overlap with the learning device described above.

[0137] The speech disorder management program inputs the subject's voice data into a pre-trained voice analysis model.

[0138] At this time, the input data may be preprocessed voice data. For example, the input data may be preprocessed by decoding compressed voice data into a multidimensional array.

[0139] Input data can be manipulated to a preset length for waveform identification. For example, the input data can be preprocessed to meet a predetermined length by truncating or padding some of the voice data.

[0140] The input data may be augmented speech data to facilitate analysis. Specifically, the input data may be augmented speech data to a degree that does not affect the accuracy of the speech analysis model. For example, the input data may be preprocessed to maintain a constant loudness by increasing or decreasing the volume of the speech data.

[0141] The input data may be data that is visualized as a spectrogram of speech data. The input data may be transformed into a two-dimensional spectrogram or mel-spectrogram to suit a speech analysis model trained to classify images.

[0142] The swallowing disorder management program uses a voice analysis model to output the aspiration state and swallowing disorder state corresponding to the subject's voice data.

[0143] Specifically, the swallowing disorder management program can provide the examinee with the aspiration analysis result by using a voice analysis model to output one of the classes including “normal,” “residual,” or “aspiration” along with the probability of belonging to the classified class, and can provide the examinee with the voice analysis result for swallowing disorder by using a voice analysis model to output one of the classes including “normal” or “abnormal” along with the probability of belonging to the classified class.

[0144] An embodiment of the present invention may also be implemented in the form of a recording medium containing computer-executable instructions, such as program modules, executed by a computer. Computer-readable media may be any available media that can be accessed by a computer, and includes both volatile and nonvolatile media, removable and non-removable media. Computer-readable media may also include computer storage media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0145] Although the methods and systems of the present invention have been described with respect to specific embodiments, some or all of their components or operations may be implemented using a computer system having a general-purpose hardware architecture.

[0146] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will readily appreciate that the present invention can be readily modified into other specific forms without altering the technical spirit or essential characteristics of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single entity may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined manner.

[0147] The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention.

Claims

1. In a method for managing dysphagia performed by a speech analysis-based dysphagia management device, (a) a step of receiving a voice from a subject through a first suction test; (b) a step of performing a first suction analysis to analyze whether the subject is suctioning by inputting the voice into a voice analysis model; and (c) a step of guiding swallowing assistance electrical stimulation eating or free eating using an electrical stimulation module according to the results of the first suction analysis by the above voice analysis model, The voice from the above first suction test includes the voice before and after the subject consumes the first type of food, The step of performing the above first suction analysis is: A method for managing dysphagia, wherein the occurrence of aspiration is determined based on characteristic information of the voice before and after ingestion of the above food.

2. In paragraph 1, The above voice analysis model is a method for managing dysphagia, wherein a plurality of unit data sets including voices before and after swallowing food are collected by making a specific speech sound for a predetermined period of time before and after swallowing food, and learning data labeled with correct answer data indicating whether aspiration occurred for each unit data set are collected.

3. In paragraph 1, A method for managing dysphagia, further comprising the step of recording the results of the aspiration analysis in a database.

4. In paragraph 1, A method for managing swallowing disorders, wherein the step (c) includes a step of guiding free eating when no aspiration occurs as a result of the first aspiration analysis.

5. In paragraph 1, The above step (c) is, if the result of the first suction analysis shows that no suction has occurred, Steps to guide free meals; Step 1: Instructing the patient to perform a second suction test after completing a free meal; A step of performing a second suction analysis by inputting the voice by the second suction test into the voice analysis model; If no aspiration occurs as a result of the second aspiration analysis, a step of outputting a meal prescription including free meals; and A method for managing dysphagia, comprising the step of providing an electrical stimulation rehabilitation training mode when aspiration occurs as a result of the second aspiration analysis.

6. In paragraph 1, In the above step (c), if aspiration occurs as a result of the first aspiration analysis, A method for managing dysphagia, comprising a step of guiding swallowing assistance electrical stimulation eating using an electrical stimulation module.

7. In paragraph 1, In the above step (c), if aspiration occurs as a result of the first aspiration analysis, Steps for guiding a pre-test for swallowing-assisted electrical stimulation meals using an electrical stimulation module; A step of guiding to perform a third suction test after the above pre-test; A step of performing a third suction analysis by inputting the voice by the third suction test into the voice analysis model; If no aspiration occurs as a result of the third aspiration analysis, a step of providing a swallowing-assisted electrical stimulation eating mode; and A method for managing swallowing disorders, comprising a step of providing an electrical stimulation rehabilitation training mode when aspiration occurs as a result of the above third aspiration analysis.

8. A method for providing a swallowing assistance electrical stimulation meal mode performed by a speech analysis-based swallowing disorder management device, A step of providing swallowing assistance electrical stimulation to a subject, having the subject eat a meal based on a first food, and performing a first voice analysis to determine whether the subject is normal and has no swallowing disorder based on a first voice collected from the subject immediately after consuming the first food; If the result of the above first voice analysis is normal, the step of continuing the swallowing assistance electrical stimulation eating mode; and A method for providing a swallowing-assisted electrical stimulation meal mode, comprising the step of providing a swallowing-assisted electrical stimulation to the subject and guiding the subject to eat a meal based on a second food having a higher viscosity than the first food, if the result of the first voice analysis is abnormal.

9. In paragraph 8, A method for providing a swallowing-assisting electrical stimulation meal mode, further comprising a step of monitoring the subject's voice generated while providing the swallowing-assisting electrical stimulation and stopping the electrical stimulation if a coughing sound is generated.

10. In paragraph 8, After the step of continuing the above swallowing-assisted electrical stimulation eating mode, A step for receiving response input information including whether discomfort occurs while the subject is eating the first food-based meal; If discomfort is input, a step of performing continuous voice analysis to determine whether the first voice continuously collected from the subject is normal without a swallowing disorder for a preset number of times, and A method for providing a swallowing assistance electrical stimulation eating mode, further comprising a step of performing intermittent voice analysis to determine whether the first voice intermittently collected from the subject is normal without swallowing disorder for a preset number of times when no discomfort is input.

11. In paragraph 8, The step of guiding you to proceed with a meal based on the above second food is: A step of having the subject eat a meal based on the second food, and performing a second voice analysis to determine whether the subject is normal and has no swallowing disorder based on a second voice collected from the subject immediately after consuming the second food; If the result of the above second voice analysis is normal, the step of continuing the swallowing assistance electrical stimulation eating mode and A method for providing a swallowing assistance electrical stimulation eating mode, comprising a step of guiding an electrical stimulation rehabilitation training mode when the result of the above second voice analysis is abnormal.

12. In paragraph 11, After the step of continuing the above swallowing-assisted electrical stimulation eating mode, A step for receiving response input information including whether discomfort occurs while the subject is eating a second food-based meal; If discomfort is input, a step of performing continuous voice analysis to determine whether the second voice continuously collected from the subject is normal without a swallowing disorder for a preset number of times, and A method for providing a swallowing assistance electrical stimulation eating mode, further comprising a step of performing intermittent voice analysis to determine whether the second voice intermittently collected from the subject is normal without swallowing disorder for a preset number of times when no discomfort is input.

13. In paragraph 10 or 11, The steps of performing the above voice analysis are: A step of inputting the first voice or the second voice into a voice analysis model and determining whether or not the swallowing disorder is normal based on the characteristic information of the voice immediately after ingestion of the first food or the second food, The above voice analysis model is constructed by collecting a plurality of unit data sets including a first voice or a second voice collected for a predetermined time before and after swallowing a first food or a second food, and using learning data labeled with correct answer data indicating whether or not swallowing disorder is normal for each unit data set, a method for providing a swallowing assistance electrical stimulation meal mode.

14. In paragraph 10 or 12, If all of the above continuous voice analysis results are normal, the step of guiding free eating and A method for providing a swallowing assistance electrical stimulation eating mode, comprising a step of guiding an electrical stimulation rehabilitation training mode when the result of performing the above continuous voice analysis is abnormal at least once.

15. In paragraph 10 or 12, If all of the above intermittent voice analysis results are normal, the step of guiding free eating, and A method for providing a swallowing assistance electrical stimulation eating mode, comprising a step of guiding an electrical stimulation rehabilitation training mode when the intermittent voice analysis result is abnormal at least once.

16. In a speech analysis-based speech disorder management device, Memory where the disability management program is stored; and Including a processor for executing a program stored in the above memory, The above program receives the voice from the subject's first suction test, inputs the voice into a voice analysis model to perform a first suction analysis to analyze whether the subject is sucking, and guides swallowing assistance electrical stimulation meal or free meal using an electrical stimulation module according to the result of the first suction analysis by the voice analysis model. The voice from the above suction test includes the voice of the subject before and after ingestion of the first type of food, A swallowing disorder management device, wherein the program determines whether aspiration occurs based on characteristic information of the voice before and after ingestion of the food when performing the first aspiration analysis.

17. In paragraph 16, The above-mentioned voice analysis model is a swallowing disorder management device constructed by collecting multiple unit data sets including voices before and after swallowing food by making specific speech sounds for a predetermined period of time before and after swallowing food, and using learning data labeled with correct answer data indicating whether aspiration occurred for each unit data set.

18. In paragraph 16, The above program is a device for managing dysphagia, which records the history of the results of suction analysis in a database.

19. In paragraph 16, The above program is a device for managing swallowing disorders, which guides free eating when no aspiration occurs as a result of the first aspiration analysis.

20. In paragraph 16, The above program, if no suction occurs as a result of the first suction analysis, A swallowing disorder management device that guides a user to perform a second suction test after completing a free meal, performs a second suction analysis by inputting the voice from the second suction test into the voice analysis model, outputs a meal prescription including a free meal when no suction occurs as a result of the second suction analysis, and provides an electrical stimulation rehabilitation training mode when a suction occurs as a result of the second suction analysis.

21. In paragraph 16, The above program, if suction occurs as a result of the first suction analysis, A swallowing disorder management device that guides swallowing assistance electrical stimulation eating using an electrical stimulation module.

22. In paragraph 16, The above program, if suction occurs as a result of the first suction analysis, A swallowing disorder management device that guides a pre-test for swallowing-assisted electrical stimulation eating using an electrical stimulation module, guides a third suction test to be performed after the pre-test, performs a third suction analysis by inputting the voice from the third suction test into the voice analysis model, and provides a swallowing-assisted electrical stimulation eating mode when no suction occurs as a result of the third suction analysis, and provides an electrical stimulation rehabilitation training mode when a suction occurs as a result of the third suction analysis.

23. A computer-readable recording medium recording a computer program for executing a method for managing a hearing impairment according to any one of paragraphs 1 to 7.

24. A computer-readable recording medium recording a computer program for executing a method for providing a swallowing assistance electrical stimulation meal mode according to any one of claims 8 to 15.

Images