KR20210017195A - System of estimating industrial information based on DEEP LEARNING with using electroencephalography and operating method thereof - Google Patents

Abstract

The present invention relates to a system for predicting trend information on a specific industrial field based on deep learning and using a brain wave signal and a method for operating the same. According to the present invention, a brain wave is detected that is generated by a shape thought or a word vocalized by a person and then a cloud server is accessed and deep learning-based search is performed. Accordingly, an optimal result is output and shown as to which shape was thought and which word was vocalized. In addition, multiple cloud servers are accessed by user selection, deep learning search is performed, and thus an optimal result is predicted, output, and shown with regard to corresponding industrial information. The present invention includes: a brain wave electrode unit coming into dry contact with the scalp, directly detecting a brain wave signal of several microvolts to hundreds of microvolts, performing primary amplification on a voltage level, and visually outputting result information received during conversion into a wireless signal and transmission; a brain wave management unit filtering the noise signal component of the brain wave signal wirelessly received from the brain wave electrode unit, performing secondary amplification and output, and sending the input result information to the brain wave electrode unit; and a brain wave analysis unit performing control and monitoring such that a signal received from the brain wave management unit is input, a cloud server accessed through a first communication module is searched for based on deep learning, the shape or word indicated by the brain wave signal is confirmed, and the confirmed result signal is transmitted to the brain wave electrode unit.

Description

[System of estimating industrial information based on DEEP LEARNING with using electroencephalography and operating method thereof]

The present invention relates to a system for predicting trend information in a specific industry using a deep learning-based EEG signal and a method of operating the same, and more specifically, to detect an electroencephalography generated by a shape or a word spoken by a person, and Deep learning search by connecting to the server, so that the meaning of the thought shape or spoken word is accurately checked, converted into a corresponding command, and deep learning search by connecting to a number of multi-layered cloud servers, so the optimal result is selected from the searched multiple result information And, it relates to a specific industry trend information prediction system using deep learning-based brainwave signals that accurately predict the trends of a designated specific industry field and a method of operation thereof.

Deep learning (DEEP LEARNING) is a field of machine learning and includes research to transform data into a matrix form that can be processed by a computer and build a model that learns it with an artificial neural network. For a specific learning goal that recognizes objects or faces, deep learning is known to perform better in recognition rate than existing machine learning techniques. This deep learning is based on an artificial neural network created with inspiration from neuroscience such as neurons, and is composed of several layers of hierarchical structure similar to human recognition methods.

EEG refers to the potential fluctuations occurring in the cerebrum of a person or animal (hereinafter, described with reference to'people'). Alpha (α) waves appearing when a person is awake and lost during sleep, mental Beta (β) waves that appear when active or when the nervous system is active, theta (θ) waves that appear well in childhood and when there is emotional stress after adulthood, delta (δ) waves that appear during sleep in adults and appear well in patients with brain diseases. ) There are four types of waves, and an example of each type of brain wave of a person according to an embodiment is well illustrated in FIG. 1.

2 is a functional block diagram of a deep learning-based prediction system according to an embodiment of the prior art.

Hereinafter, a detailed description with reference to the accompanying drawings is a configuration including a measuring device 10, a pre-processing device 20, a learning device 30, and a verification device 40.

The measuring device 10 measures the electrocardiogram and the photoelectric pulse wave provided from each sensor, and the preprocessing device 20 removes the noise of the measured signal, divides it into a predetermined window size, removes data outside the threshold range, and performs sampling and holder reel. After performing the preprocessing by this data shuffling, it is applied to the learning device 30. The learning device 30 predicts the relaxation value and contraction value of blood pressure by performing deep learning through the set learning model, and applies it to the verification device 40. The verification device 40 performs verification by comparing the predicted value of the blood pressure and the measured value.

The prior art provides a system configuration of a technical idea that predicts human blood pressure through deep learning using waveforms measured by electrocardiogram, but it is a problem that the brain wave signal is deep-learned to detect the shape or words that a person thinks. There is.

3 is a functional configuration diagram of an electrode for measuring brain waves according to an embodiment of the prior art.

Hereinafter, referring to the accompanying drawings, an electrode for measuring (detecting) an EEG is a configuration including a circuit board protection member 30, a circuit board 40, a conductive needle rod 50, and a case.

The circuit board protection member 30 has a U shape in which the upper plate 31 and the lower plate 32 are connected as a support part 33 and a space part 34 is formed, and a vertical through hole 35 is perforated in the center of the lower plate 32 Has been. The circuit board 40 is fitted into the space 34, and the brain wave amplifying means 41 is installed on the upper surface, and the contact part 42 is provided on the bottom surface. The needle rod 50 is inserted into the vertical through hole 35 and processed so that the upper end contacts the contact portion 42 and the lower end is sharp. The case is configured to surround the entire surface of the circuit board protection member 30 in a state in which the circuit board 40 is embedded in the space part 34 and the needle rod 50 is assembled in the through hole 35.

Although such a conventional technique is a structure suitable for detecting EEG by dry contact, there is a problem in that the EEG detection signal level is weakened due to contact resistance with the skin, thereby causing a detection error of EEG.

On the other hand, in the prior art, there is a problem in that it is not possible to determine what shape a person is thinking (imagining) or what word he is speaking using the detected brain waves.

Therefore, it is necessary to develop a technology that accurately predicts the trend information of the necessary industrial field by checking the shape of thinking (imagining), speaking words or sentences, and accurately predicting the necessary trend information of the industrial field by selecting it, as it accurately detects brain waves without errors and analyzes deep learning using the detected brain waves. There is.

Korean Patent Registration No. 10-1937018 (2019. 01. 03.)'A method and apparatus for automatically dividing the inner and outer diameters of intravascular ultrasound images using deep learning' Korean Patent Registration No. 10-1953019 (February 21, 2019)'How to perform translation and provide content using deep learning artificial intelligence translation software' Korean Patent Application No. 10-2001-0037196 (2001. 06. 27)'electrode for measuring brain waves' Korean Patent Application No. 10-2017-0154208 (2017. 11. 17.)'Deep learning-based blood pressure prediction system and method'

The present invention, conceived to solve the problems and necessity of the prior art as described above, is a trend in a specific industrial field using deep learning-based EEG signals that accurately detect EEG signals by dry contact by a plurality of electrodes arranged in a headset or band shape. Its purpose is to provide an information prediction system and its operation method.

In addition, the present invention provides a system for predicting trend information in a specific industry field using a deep learning-based brain wave signal that accurately identifies a shape or a word (sentence) spoken by a person by deep learning processing of the detected brain wave signal, and a method of operation thereof. That is its purpose.

On the other hand, the present invention identifies the shape or words (sentences) that the user thinks with brain waves and converts them into commands to control multiple multi-layered cloud servers and search for deep learning, so the trend of specific industrial fields corresponding to commands using brain waves Its purpose is to provide a system for predicting trends in specific industrial fields using brainwave signals based on deep learning that simply and accurately investigates and predicts information, and a method of operating the same.

The specific industry trend information prediction system using the deep learning-based EEG signal of the present invention devised to achieve the above object directly detects EEG signals of several to hundreds of microvolts by dry contact with the scalp and determines the voltage level. A brainwave electrode unit for first amplifying, converting into a wireless signal, transmitting, and visually outputting the received result information; An EEG management unit for filtering the noise signal component of the EEG signal wirelessly received from the EEG electrode unit, second amplifying and outputting it, and transmitting the input result information to the EEG electrode unit; The signal received from the EEG management unit is input, and the cloud server connected through the first communication module is searched in a deep learning method to confirm the shape or word meaning of the EEG signal, and the confirmed result signal is transmitted to the EEG electrode unit. EEG analysis unit for controlling and monitoring; It may include.

The EEG analysis unit is selected and accessed through the first communication module, and the shape or word that the detected EEG signal means is searched and provided in a deep learning method, and the shape or word input through the second communication module refers to a command. A cloud server that searches and provides trend information of an industrial field in a deep learning method and consists of one or more; It may further include.

The EEG electrode unit comprises a plurality of electrode elements including three or more needle rods and containing fine particle carbon powder to form conductivity and flexibility; An electrode management unit that connects to the plurality of electrode elements, allocates and manages corresponding serial numbers, respectively, and detects EEG signals from the pair of electrode elements selected by corresponding control signals; A first connected to the electrode management unit, including both an infrared signal transmission/reception unit and a Wi-Fi signal transmission/reception unit, activating one or more by selection, transmitting each detected brain skin signal, receiving the corresponding control signal, and transmitting it to the electrode management unit. EEG radio unit; It may include.

Activated the second communication with any one or more or all connected cloud servers selected by random check operation among a plurality of cloud servers by selecting and connecting any one of the multiple cloud servers connected by activating the first communication module It may be configured to be connected through a module.

In order to achieve the above object, the method of operating a trend information prediction system for a specific industry using a deep learning-based EEG signal of the present invention is characterized by using a deep learning-based EEG signal including an EEG electrode unit, an EEG management unit, and an EEG analysis unit. A method of operating an industrial trend information prediction system, comprising: an EEG detection process of detecting and amplifying EEG signals in a state imagined or spoken by a user when it is determined that a command signal for detecting a user's EEG is input by the EEG electrode unit; The EEG analysis unit receives the EEG signal detected through the EEG management unit, selects and connects to a specific cloud server from among a plurality of cloud servers, and translates the words and sentences that the EEG signal means by a deep learning method; A deep learning search process of searching for multiple result information by accessing a plurality of multi-layered cloud servers by selection and searching the translated command in a deep learning method when the EEG analysis unit determines that the user has approved the translated command; An optimal result output process of selecting and outputting the optimal result information from the plurality of result information by the brain wave analysis unit and proceeding to an end when a signal of satisfaction is inputted from the user; It may include.

In the deep learning search process, if the user determines that the translated command is not the same as the intended command, and therefore, if a signal that does not approve is determined, it is fed back to the brainwave detection process to detect and amplify the EEG signal in a state that the user imagines or speaks. EEG repetitive detection process of repeating; It may further include.

In the process of outputting the optimum result, if a signal to be selected by the user is satisfied with the optimum result information is not input, it is fed back to the deep learning search process, and the user approves the translated command, and then deep learning and selection of the optimum result are performed. Iterative deep learning repetition process; It may further include.

The present invention having the configuration as described above has the advantage of accurately detecting an EEG signal through dry contact using a headset or a band in which a plurality of electrodes are uniformly arranged.

In addition, the present invention has the advantage of accurately identifying a shape or a word (sentence) spoken (pronounced) by a user because the detected brainwave signal is deep-learned.

On the other hand, the present invention identifies the shape or word (sentence) thought by the user with EEG, converts it into a command, and controls a plurality of multi-layered cloud servers to search for deep learning, so the trend of a specific industry field corresponding to the command converted using EEG It has the advantage of simple and accurate investigation and prediction of information.

1 is an exemplary diagram for each type of human brainwave according to an embodiment;
2 is a functional block diagram of a deep learning-based prediction system according to an embodiment of the prior art;
3 is a functional configuration diagram of an electrode for measuring brain waves according to an embodiment of the prior art,
4 is a functional configuration diagram of a system for predicting trend information in a specific industry using a deep learning-based brain wave signal according to an embodiment of the present invention;
5 is a configuration diagram of a headset-type EEG detection unit according to an embodiment of the present invention,
6 is a detailed functional configuration diagram of an EEG detection unit according to an embodiment of the present invention,
7 is a cross-sectional view of an electrode for detecting brain waves according to an embodiment of the present invention,
8 is a flow chart illustrating a method of operating a trend information prediction system for a specific industry using a deep learning-based brain wave signal according to an embodiment of the present invention.
And
9 is a diagram illustrating a method of searching for multiple multi-layered cloud servers using a deep learning method according to an embodiment of the present invention.

Since the present invention may apply various transformations and may have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to a specific embodiment, it should be understood to include all conversions, equivalents, or substitutes included in the spirit and scope of the present invention. In describing the present invention, when it is determined that a detailed description of a related known technology may obscure the subject matter of the present invention, a detailed description thereof will be omitted.

In the following description, the indications and meanings are to be used in the same meaning, and can be selectively written appropriate to the context, and words and sentences are to be used in the same meaning and can be selectively written appropriate to the context. do.

Deep Learning refers to a technology used to classify or cluster objects or data. It is a kind of machine learning that trains computers to classify things as the human brain classifies things.

In 2006, the term deep learning was first used in a dissertation by Professor Geoffrey Hinton at the University of Toronto, Canada. The basic concept of deep learning is similar to an artificial neural network (ANN). An artificial neural network is an algorithm that processes information in a manner similar to that of a human brain. It learns information by repeating the process of combining and classifying data of various elements such as the face or shape of an object. In his thesis, Hinton suggested a method to overcome the shortcomings of the existing artificial neural network model. In addition, with the development of hardware that can analyze a large amount of data and the advent of big data, artificial neural networks show more excellent results, which is called deep learning.

One of the features of deep learning is a data classification method through unsupervised learning. In general, computer data classification methods can be divided into supervised learning and unsupervised learning. Supervised learning is a method of first inputting classification criteria into a computer, and existing machine learning algorithms have usually classified data using supervised learning. In unsupervised learning, information is input without classification criteria and the computer automatically classifies it. Computers find similar clusters by themselves and classify data, which requires a high level of computational power. Deep learning performs artificial neural network optimization by trimming data through pre-training using an unsupervised learning method. The feature of deep learning is that the algorithm includes from feature extraction to learning.

A number of companies, including Google and Facebook, are using deep learning techniques. Google has succeeded in recognizing cats in YouTube videos by using Deep Neural Network (DNN) technology, which is one of the deep learning algorithms, and Facebook has also developed DeepFace, a facial recognition technology using deep learning. did. These deep learning technologies are now generally well-known technologies, so further detailed descriptions will be omitted.

Brain wave (electroencephalography) is a measurement of a set of brain cell action potentials with electrodes on the scalp, and the origin of the brain wave (EEG) was experimental in 1875 by British physiologist Richard Caton. Used voltage-sensitive primitive devices to record electrical activity from the brain surfaces of dogs and rabbits. It is known that human brain waves were first observed by Berger (H.) of Germany in 1929. Berger observed that the brainwaves of working people and those of sleeping people were clearly different.

It is known that the size of the EEG of a person (human) is in the range of several to several hundred microvolts (μV), and the frequency is in the range of 0.5 to 60 Hertz (Hz), and the waveform or size of the EEG changes according to the brain activity state.

For example, when a person closes his eyes and rests, an EEG of 8 to 13 Hz and amplitude of 10 to 50 μV, which is significantly induced in the occipital region, is called an alpha (α) wave (basic wave), and 14 when the eyes are opened. A small amplitude EEG wave of 2 to 20 microvolts of from 60 Hz is referred to as a beta (β) wave (fast wave), and an EEG of 0.5 to 3.5 Hz and tens to 300 μV that appears during sleep is referred to as a delta (δ) wave (slow wave). In addition, the brain waves of children or waves of 4 to 13 Hz and 20 to 100 microvolts that appear when an adult is excited are largely classified as theta (θ) waves, and in some cases, they are further classified. The criteria for classifying them may be slightly different.

These brain waves are induced according to sensations, stimulation, and thoughts (triggered brain waves), and are attracting attention as various diagnostic tools for interpretation of brain functions and body changes. In other words, the electroencephalogram signal by the electric potential measured by using the electrode to the microscopic signals on the surface of the brain, which is a result of the synthesis of electrical signals generated by numerous nerves in the brain, changes spatiotemporally according to the brain activity, the state at the time of measurement, and the brain function EEG measurement is an essential tool for diagnosing brain function and disorders.

EEG measurement is a method of attaching about 24 electrodes at a standard position on the scalp and detecting a weak amplitude voltage change between the pair of electrodes in the front, rear, left and right, and the number of electrodes is added or subtracted as necessary, and the position to be attached. It is quite natural that is subject to change.

Electrodes for EEG detection are classified into a wet method using a liquid and a dry method using a dry state to increase conductivity, and each is a commonly known method.

4 is a functional configuration diagram of a trend information prediction system for a specific industry using a deep learning-based EEG signal according to an embodiment of the present invention, and FIG. 5 is a configuration diagram of a headset-type EEG detection unit according to an embodiment of the present invention. 6 is a detailed functional configuration diagram of an EEG detection unit according to an embodiment of the present invention, and FIG. 7 is a cross-sectional view of an electrode for EEG detection according to an embodiment of the present invention.

Hereinafter, a system for predicting trends in specific industrial fields using deep learning-based EEG signals will be described with reference to all accompanying drawings. The EEG electrode unit 1000, the EEG management unit 2000, the EEG analysis unit 3000, and one or more It is a configuration including a cloud server (4000).

The EEG electrode unit 1000 includes a plurality of electrode elements 1100, an electrode management unit 1200, a first EEG radio unit 1300, and a visual display unit 1400.

It is quite natural that the electrode element 1100 is made up of one or more pluralities, and usually may be made up of 24, and can be added or subtracted if necessary. In the case of the EEG electrode unit 1000 in the shape of a headset (hat), it may be composed of 24 electrode elements 1100 or may have more electrode elements 1100 according to needs and purposes. In the case of the EEG electrode unit 1000 having a band (headband) shape, it is quite natural that at least two electrode elements 1100 are provided, and the number can be further increased according to the need and purpose. On the other hand, the shape of the headset and the shape of the band are not determined, and can have a wide variety of shapes or geometric shapes depending on the type of brain waves to be detected.

The EEG electrode unit 1000 is in the form of a headset or a headband in the shape of a hat, and when it is covered on the head, it is not peeled off until sufficient external force is applied from the outside, and when the EEG electrode unit 1000 is covered on the head, each The electrode element 1100 will be described as a configuration in which the electrode element 1100 is in close contact with a designated position to maintain an electrically conductive state.

The electrode elements 1100 are preferably arranged with the same number on the left and right sides of the head and on the upper and lower sides of the head, and if necessary, the electrode elements 1100 may be intensively arranged at the frontal lobe position or at randomly uniform intervals or at an unspecifiedly selected position. I can.

The electrode element 1100 in FIG. 5 schematically illustrates a state in which the corresponding electrode is connected to the terminal unit 1120 and is installed without being exposed to the outside.

Each electrode element 1100 has a terminal portion 1120 for connecting an electric wire to a portion of the upper side of the base 1110 having a disk shape, and at the lower edge portion of the electrode element 1100, three or more electrodes having a straight bar shape 1 The protrusions 1130 are equally spaced on a concentric circumference and form an angle according to any one value selected from a range of 10 to 20 degrees based on a corresponding vertical line, and an end portion is formed to be inclined in an outward direction. At this time, it is relatively preferable to configure the first protrusion 1130 in a shape that has an angle of 15 degrees and an end portion spreads along the outer direction in order to improve contact with the scalp. Since the electrode element 1100 has a plurality of protrusions, the contact area is increased and the contact resistance is reduced, so that the EEG signal is accurately detected.

A second protrusion 1140 in the shape of a circular rod is formed in the center of the circumference formed by the first protrusion 1130, but the length of the second protrusion 1140 in the vertical line direction is shorter and thicker than the first protrusion 1130 Is thicker.

The base 1110 has a diameter (a) of 10 millimeters (mm) and a thickness of 5 millimeters, and the first protrusion 1130 has a diameter (e) of 2.5 millimeters and a vertical length (b) of The shape of a circular rod of 10 millimeters, and the end in contact with the scalp has a round shape, and the second protrusion 1140 has a diameter (d) of 3.5 millimeters and a vertical length (c) of 8 millimeters. And the end portion in contact with the scalp has a round shape, and the first protrusion 1130 and the second protrusion 1140 form a needle rod. The terminal portion 1120 is not limited in size and shape, and may have a groove shape if necessary.

The base 1110, the terminal 1120, the first protrusion 1130 and the second protrusion 1140 are made of the same material of elastic rubber or plastic, respectively, and carbon powder of fine particles of 0.5 to 1 micrometer or less in diameter It is included in the range of 75 to 95% by weight to form conductivity. The smaller the fine particle size of the carbon powder is, the more uniformly the electrical conductivity is formed and the electrical conductivity is improved, and a configuration including the fine-particle carbon powder in the range of 90% of the total weight of each electrode element 1100 is preferable.

The electrode management unit 1200 electrically connects to the terminal unit 1120 constituting the plurality of electrode elements 1100 in a separate state, assigns a corresponding serial number to each electrode element 1100, and controls the corresponding EEG signals are respectively detected from the pair of electrode elements 1100 selected by the signal. Here, it is very natural that each electrode element 1100 forming a pair may be selected indefinitely if necessary, but a certain position may be selected.

The first EEG wireless unit 1300 is connected to the electrode management unit 1200 and has both an infrared signal transceiving unit and a WiFi signal transceiving unit to simultaneously activate the operation state and simultaneously transmit each detected EEG signal as an infrared signal and a WiFi signal. It receives the corresponding control signal and transmits it to the electrode management unit 1200.

On the other hand, if necessary, only one selected from the infrared signal transmitting and receiving unit and the WiFi signal transmitting and receiving unit may be operated. That is, if there is a medical device affected by the Wi-Fi wireless signal nearby, only the infrared signal transceiving unit is operated in an active state. Otherwise, both the infrared signal transceiving unit and the WiFi signal transceiving unit are activated to transmit the same signal. Since it transmits and receives simultaneously through the path, transmission errors of the corresponding signal can be reduced. Of course, it will be described that the counterpart side receiving the transmission signal has a function unit for detecting the minimum transmission error signal.

The visual display unit 1400 is a configuration that connects to the electrode management unit 1200 and outputs and displays the result information applied from the EEG analysis unit 3000 as visual information, and the front is visible from the position in front of the user's eyes as shown in the drawing. It is installed with transparent glasses, and the corresponding result information is output and displayed as a semi-transparent color signal. It is very preferable that the visual display unit 1400 is positioned in front of the eye or can be temporarily adjusted to another position as necessary.

The EEG management unit 2000 includes a second EEG wireless unit 2100, wirelessly connects to the first EEG wireless unit 1300 of the EEG electrode unit 1000, and selects any one or any one or more of an infrared signal and a Wi-Fi signal. Both the infrared signal transceiving unit and the Wi-Fi signal transceiving unit are provided to receive or transmit at the same time. In addition, it is possible to activate only one selected as needed and operate to transmit and receive corresponding signals.

The first EEG wireless unit 1300 of the EEG electrode unit 1000 and the second EEG wireless unit 2100 of the EEG management unit 2000 are provided with an infrared signal transceiving unit and a Wi-Fi signal according to the corresponding control signal of the EEG analysis unit 3000. Any one or one or more selected from the transmission/reception unit is operated in an active state and transmitted to other connected components as necessary.

The method of reducing the transmission error of the received signal is a method of selecting the signal with the least transmission error among the signals received through each path, and calculating the transmission error by calculating the signal received through each path using a well-known arithmetic average method. Any one or any one or more of the reduction methods can be selected and used. On the other hand, a method of selecting a signal having a small transmission error among each path can be divided into intervals of seconds, and the received original signal and the final processed signal are recorded and stored in respective corresponding memory areas.

The EEG management unit 2000 filters the noise signal component of the EEG signal received wirelessly, amplifies and outputs the second amplification, and transmits the result information input from the EEG analysis unit 3000 to the EEG electrode unit 1000. Here, the second amplified signal can be amplified to a level of 5 volts suitable for digital signal processing. The noise signal component filtering method excluding the EEG signal of the EEG management unit 2000 uses a low-pass filter to pass only the corresponding frequency signal of the EEG signal because the frequency of the EEG signal is relatively low. This is a method of filtering signal components. For such a low-pass filter, a known LC filter is used, and a configuration in which the capacity of a coil or L (L) having a core is relatively large is used, which is made of an inductor.

The EEG management unit 2000 is connected to the EEG analysis unit 3000 by wire, and each of the corresponding communication functional units is provided. Since the wired communication function is a general configuration, detailed descriptions will be omitted.

Each component of the entire system including the EEG analysis unit 3000 will be described as having all the corresponding programs and functions required for operation in a built-in state.

The EEG analysis unit 3000 inputs the signal received from the EEG management unit 2000 and searches the cloud server 4000 selectively connected through the first communication module 3110 and through the operation of random checks in a deep learning method, The shape or word (sentence) displayed by the signal is checked, and the resultant signal is transmitted to the brainwave electrode unit 1000 to be output from the visual display unit 1400 and monitored.

Here, it is relatively very desirable that the cloud server 4000 is configured to have 40 or more pieces of information to be searched in order to secure the reliability of the search result information. In addition, it is highly desirable to indicate that the search result is selected from among the total number of stored information.

That is, the EEG analysis unit 3000 accurately checks the command meaning of the EEG signal received from the EEG electrode unit 1000 via the EEG management unit 2000 in a deep learning method, and the user's confirmation through the visual display unit 1400 Receive.

On the other hand, the EEG analysis unit 3000 determines when a confirmation signal of a state selected by the user is input for a command indicated by a shape or word (sentence) output to the visual display unit 1400 of the EEG electrode unit 1000, and , The user's confirmation signal is applied as an EEG signal detected by the EEG electrode unit 1000.

The user's confirmation signal again searches the cloud server 4000 selectively connected through a random number table in a deep learning method and analyzes the search result, so that the EEG analysis unit 3000 determines the final confirmation. The EEG analysis unit 3000 includes a personal computer (PC).

When the EEG analysis unit 3000 is finally confirmed, analyzed, and determined as a specific command signal, the command signal is connected to one or more cloud servers 4000 connected through the second communication module 3120 and deep learning In this way, the information requested by the command signal is searched.

Here, one or more cloud servers 4000 connected through the second communication module 3120 may include a cloud server 4000 connected through the first communication module 3110, and all clouds subject to deep learning Any one selected from the servers 4000, any one or more, or all cloud servers 4000 may be included.

Since the first communication module 3110 and the second communication module 3120 have the same function, if a failure occurs in any one of the configurations, any one normally operated by the corresponding program is replaced and can be operated without errors. Alternatively, it may be provided with only one and operated in a separated state.

The cloud server 4000 searches for a command corresponding to a shape or word that is connected through the first communication module 3110 and selected through the operation of the random number table by the EEG analysis unit 3000 and detected in a deep learning method. To provide. The process of confirming the command may be described as a translation process, and it is natural that the C-language may be included again in the translation process, and the cloud server 4000 includes a corresponding processor.

In addition, the cloud server 4000 searches and provides trend information of a specific industry field by selection by operation of a random number table by the brain wave analysis unit 3000 and a command input through the second communication module 3120 in a deep learning method. It consists of one or more.

It will be described that the cloud server 4000 includes all cloud servers that can be searched by a deep learning method. The cloud server 4000 may include all commonly known servers.

If necessary, each cloud server 4000 connects to another cloud server 4000 and duplicates the result information in a deep learning method, such as an input layer, a hidden layer, and an output layer. It is an artificial neural network with a multi-layered structure of and can constitute a deep neural network of a deep neural network, and any one of them can be constructed.

The hidden layer can consist of one or two or more layers, the input layer is the layer in which data is initially set, the hidden layer is the layer in which the data is hidden and hidden, and the output layer is the layer in which the learned data to be obtained is output. Can mean layer. A portion to which each cloud server 4000 is connected may be expressed as a separate node. The nodes of each layer can generate an output value obtained by multiplying the input value by the weight and pass it to the next layer. For example, the nodes of the input layer may be input to each node of the first hidden layer as an input value. The cloud server 4000 can learn a nonlinear relationship by repeating as many layers as the number of layers, and the characteristics of different layers can be learned for each layer, so that the cloud server can grasp the potential structure of Gungbo. In addition, if the weight of each node is properly adjusted, the user can generate or search for a desired learned result value.

Each cloud server 4000 is a deep learning algorithm that provides not only information on sentences and orders, but also stock price fluctuations based on economic conditions and stock information, public behavior patterns, corporate sales forecasts, information on local commercial districts, and information on terrorism. It can be input through the input layer of or a corresponding node.

Values input to the input layer can be input to each node of the hidden layer, and the value input to the hidden layer passes through a specific layer to deliver the result information desired by the user to the corresponding node of the output layer. The cloud server 4000 generates at least one or more result information by applying a deep learning algorithm. The result information generated as a result of applying the deep learning algorithm is, for example, based on one or more results, and may include information on stock price fluctuation prediction, vehicle collision threat, and sales performance prediction of a company.

The corresponding result information or prediction information is output to the user's EEG analysis unit 3000, and the user may request re-run of the search if it is determined that it is not necessary information by checking and reviewing it.

Since the brainwave analysis unit 3000 is a configuration including a personal computer, the optimal result information is selected from the result information retrieved from a plurality of cloud servers 4000 and displayed visually on an output unit provided by itself, or a visual display unit 1400 It can be controlled and monitored so that it is output and displayed in ).

Meanwhile, the EEG analysis unit 3000 and the cloud server 4000 may be connected through a public communication network including an Internet network, a wired communication network, and a wireless communication network, or may be connected through a dedicated communication network, overlapped, or directly connected.

8 is a flow chart illustrating a method of operating a trend information prediction system for a specific industry using a deep learning-based brain wave signal according to an embodiment of the present invention, and FIG. 9 is a deep learning method according to an embodiment of the present invention. This is a diagram explaining how to search for multiple layers of cloud servers.

Hereinafter, referring to all the accompanying drawings, a method of operating a trend information prediction system for specific industries using deep learning-based EEG signals including an EEG electrode unit 1000, an EEG management unit 2000, and an EEG analysis unit 3000 If it is determined that the command signal for detecting the user's EEG is inputted by the EEG electrode unit (S100), the EEG signal in the state that the user imagines or speaks is detected, and the detected EEG signal is amplified to a usable level. The high-frequency noise signal is blocked by using a low-pass filter and thus is removed (S110).

The EEG analysis unit 3000 receives the EEG signal detected through the EEG management unit 2000, selects and connects to a specific cloud server selected by driving a random number table among a plurality of cloud servers 4000 (S120), and uses a deep learning method. The meaning of the word or sentence that the EEG signal means is searched and provided to the user, and when the user checks and selects the search result, it converts or translates into commands that can easily operate various electronic devices such as computers or cloud servers. (S130). These commands include C-language.

For example, when the user imagines or pronounces the word'mom' and when the shape of a mother is imagined, a plurality of EEG signals detected through the EEG electrode unit made of various types of headsets including a hat shape and a headband shape It connects to any one of the cloud servers selected by random check operation and searches for the meaning of the detected brain waves using a deep learning method.

That is, when a user imagines a specific word, a specific shape, or pronounces a specific word (sentence), it detects the brainwave signal generated from the brain, connects to a specific cloud server selected by operating a random number table, and By searching in a running method, it precisely checks what words are imaginable, what shapes are imaginable, or what words are pronounced. In this case, the cloud server is relatively highly desirable to record, manage, or collect and manage result information of corresponding EEG signals of at least 40 or more users.

On the other hand, in the case of pronouncing a word, for example, '1','ㅂ','ㅜ','ㄴ','ㄱ','ㅣ','space', each consisting of consonants, vowels, and numbers. Pronouncing each word in an arrangement such as'ㅇ','ㅖ','ㅅ','ㅏ','ㅇ','ㅁ','ㅐ','ㅊ','ㅜ', and'ㄹ' sequentially or In the case of imagining, each result information retrieved from the cloud server is sequentially output to the visual display unit, and when the user selects sequentially according to the corresponding order, the EEG analysis unit rearranges the searched consonants and vowels in the corresponding order, and'estimated sales for the first quarter' It will be described as being a recombination construct with the words of. At this time, it is very natural that the corresponding shape (object or figure image, pattern, text image, etc.) can be included in the result information displayed on the visual display unit, and the user simply selects or expresses yes or similar opinion and rejects or You can select or reject the result information displayed as a no or similar expression. In addition, in the case of each selection, all consonants and vowels selected so far can be sequentially displayed or combined in a complete word form.

If the EEG analysis unit confirms that the command translated by the user is the same as the command intended by the user, and determines that the approved command signal is input (S140), accesses multiple multi-layered cloud servers (S150), and deep learning the translated command. By searching in a method, multiple result information is searched (S160). Since a number of multi-layered cloud servers are searched, the searched result information becomes a large number.

In this case, when one cloud server is displayed as a single dot, a state or method of searching for multiple multi-layered cloud servers is as shown in FIG. 9.

As shown in the accompanying drawings, a cloud server constituting an input layer, a plurality of cloud servers, and first to n-th hidden layers consisting of a multi-layered cloud server and an output layer It is quite natural that each number of cloud servers constituting the (output layer) can be added or decreased according to need or choice.

In other words, the larger the number of each of the cloud servers constituting the input layer, the first to nth hidden layers 1, 2, 3, and n, the output layer, the larger the big data. Since it is a search method, the reliability of search results can be further increased.

Therefore, there are a number of results retrieved by the deep learning method from a plurality of multi-layered cloud servers, and any one of them can be selected again as a deep learning method, a random number table method, or a selection method that is generally expected or known.

The optimal result information is selected and output from the plurality of result information searched by the EEG analysis unit (S170), and when it is determined that a signal of satisfaction for selecting the corresponding result information is input by the user's confirmation, the process proceeds to the end (S180).

In the deep learning search process by the EEG analysis unit, if it is determined that a signal that the user does not approve or select is input because the translated command is not the same as the intended command (S140), the user is fed back to the EEG detection process (S100). It repeats the detection and amplification of the EEG signal in the state of imagining or speaking.

In addition, if it is determined by the brainwave analysis unit that a signal to be selected by satisfying the search result information from the user is not input in the process of outputting the optimal result (S180), it is fed back to the deep learning search process (S150), and the user translates the command. After approval, the process is repeated to repeat deep learning and optimal result selection.

With the recent rapid increase and active operation of the Internet, a variety of information is being searched and collected through Internet media. In particular, it provides a variety of related information to predict stock price fluctuations, traffic volume (navigation), economic conditions, etc. For the convenience of providing various types of data and information such as management, management support, and decision to enter a new business, the number of companies or systems that provide the service is increasing.

In the past, users had to directly search for, collect and analyze various related information they need, or pay a fee to the expert and request the results to be provided.If the information is collected directly, there is a loss of time and a request to the expert. In this case, there were inconveniences such as relatively excessive cost.

In addition, if information such as economic trends, population, temperature, etc. directly collected by the user is old information, information on the current environment is not included, so accurate information covering from the past to the present cannot be provided and analyzed. .

On the other hand, in the case of company sales forecasting, when old consumption pattern information of past consumers is retrieved and collected, consumption pattern information at the present time cannot be provided. Therefore, it is possible to provide optimal information on sales prediction and marketing according to the current consumption pattern. Since it cannot be analyzed, accurate statistics and situations according to the current situation cannot be predicted or analyzed.

However, since the configuration described and described above searches a number of multi-layered cloud servers in a deep learning method, it is possible to search various situations in the past, present and future with multiple result information, and select the optimal result information from among the plurality of searched result information. There is this. That is, for example, it has the advantage of being able to accurately predict and analyze stock price fluctuations, traffic volume (navigation), economy, corporate sales, etc., and at the same time, the shape of the user's thinking or imagining and the state of pronunciation are checked by the user's brainwave detection. There are advantages such as being able to quickly and easily control various electronic devices including cloud servers.

In the above, the present invention has been described in detail with respect to the described specific examples, but it is obvious to those skilled in the art that various modifications and modifications are possible within the scope of the technical idea of the present invention, and it is natural that such modifications and modifications belong to the appended claims.

1000: EEG electrode unit 1100: electrode element
1200: electrode management unit 1300: first EEG wireless unit
1400: visual display unit 2000: brain wave management unit
2100: second EEG wireless unit 3000: EEG analysis unit
3110: first communication module 3120: second communication module
4000: Cloud server

Claims (7)

An EEG electrode unit that directly detects an EEG signal having a size of several to hundreds of microvolts by dry contact with the scalp, amplifies the voltage level first, converts it into a wireless signal, transmits it, and visually outputs the received result information;
An EEG management unit for filtering the noise signal component of the EEG signal wirelessly received from the EEG electrode unit, second amplifying and outputting it, and transmitting the input result information to the EEG electrode unit;
Input the signal received from the EEG management unit, search the cloud server connected through the first communication module in a deep learning method to check the shape or word displayed by the EEG signal, and transmit the confirmed result signal to the EEG electrode unit. EEG analysis unit for controlling and monitoring; A system for predicting trend information in a specific industry field using a deep learning-based brain wave signal comprising a. The method of claim 1,
The EEG analysis unit is selected and accessed through the first communication module, and the shape or word that the detected EEG signal means is searched and provided in a deep learning method, and the shape or word input through the second communication module refers to a command. A cloud server that searches and provides trend information of an industrial field in a deep learning method and consists of one or more; A system for predicting trend information in a specific industry field using a deep learning-based brain wave signal further comprising a. The method according to claim 1 or 2,
The EEG electrode part
A plurality of electrode elements composed of three or more needle rods and containing fine particle carbon powder to form conductivity and to have flexibility;
An electrode management unit that connects to the plurality of electrode elements, allocates and manages corresponding serial numbers, respectively, and detects EEG signals from the pair of electrode elements selected by corresponding control signals;
It is connected to the electrode management unit, has both an infrared signal transmission/reception unit and a Wi-Fi signal transmission/reception unit, activating any one or more by a corresponding control signal, transmitting each detected brain skin signal and receiving the corresponding control signal, and the electrode management unit A first EEG radio unit that transmits to; A system for predicting trends in specific industry sectors using brainwave signals based on deep learning, characterized by a configuration comprising a. The method of claim 3,
The EEG analysis unit
Activated the second communication with any one or more or all connected cloud servers selected by random check operation among a plurality of cloud servers by selecting and connecting any one of the multiple cloud servers connected by activating the first communication module A system for predicting trends in specific industry sectors using brainwave signals based on deep learning, characterized in that consisting of a configuration connected through a module. In a method of operating a trend information prediction system for specific industries using deep learning based brain wave signals including an EEG electrode unit, an EEG management unit, and an EEG analysis unit,
An EEG detection process of detecting and amplifying the EEG signal in a state that the user imagines or speaks when it is determined that a command signal for detecting a user's EEG is input by the EEG electrode unit;
The EEG analysis unit receives the EEG signal detected through the EEG management unit, selects and connects to a specific cloud server from among a plurality of cloud servers, and checks the words and sentences that the EEG signals mean by deep learning, and translates them into commands. ;
A deep learning search process of searching for multiple result information by accessing a plurality of multi-layered cloud servers by selection and searching the translated command in a deep learning method when the EEG analysis unit determines that the user has approved the translated command;
An optimal result output process of selecting and outputting the optimal result information from the plurality of result information by the brain wave analysis unit and proceeding to an end when a signal of satisfaction is inputted from the user; A method of operating a trend information prediction system for a specific industry using a deep learning-based brain wave signal comprising a. The method of claim 5,
In the deep learning search process, if the user determines that the translated command is not the same as the intended command, so that a signal that does not approve is input, it is fed back to the EEG detection process to detect and amplify the EEG signal in the state of imagination or speech EEG repetitive detection process of repeating; A method of operating a trend information prediction system for a specific industry using a deep learning-based brain wave signal further comprising a. The method of claim 5 or 6,
In the process of outputting the optimum result, if a signal to be selected by the user is satisfied with the optimum result information is not input, it is fed back to the deep learning search process, and the user approves the translated command, and then deep learning and selection of the optimum result are performed. Iterative deep learning repetition process; A method of operating a trend information prediction system for a specific industry using a deep learning-based brain wave signal further comprising a.

Images