AI-based single-person safety monitoring and community matching system

Abstract

The present invention relates to an AI-based safety monitoring and community matching system for people living alone that monitors the living patterns of people living alone using a smartphone and detects dangerous situations through AI analysis capable of preventing social isolation through regular contact.

Description

Technology Field

[0001] The present invention relates to an AI-based safety monitoring and community matching system for people living alone, and more specifically, to an AI-based safety monitoring and community matching system for people living alone that monitors the living patterns of people living alone using a smartphone and detects dangerous situations through AI analysis capable of preventing social isolation through regular contact. Background Technology

[0003] In modern society, the safety of people living alone is emerging as a critical issue. While existing emergency call devices and wearable devices exist, they often have low effectiveness because they require active operation by the user or continuous wearing. This invention aims to provide a system that combines a smartphone with AI analysis to learn the normal daily life patterns of a person living alone, detects changes in the smartphone's location using NFC / RFID tags, and can quickly notify guardians and relevant personnel in the event of a dangerous situation.

[0004] Meanwhile, the aforementioned background technology is technical information that the inventor possessed for the derivation of the present invention or acquired during the process of deriving the present invention, and it cannot necessarily be considered publicly known technology disclosed to the general public prior to the filing of the present invention. The problem to be solved

[0006] One aspect of the present invention provides an AI-based safety monitoring and community matching system for people living alone that monitors the living patterns of people living alone using a smartphone and detects dangerous situations through AI analysis that can prevent social isolation through regular contact.

[0007] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0009] An AI-based safety monitoring and community matching system for people living alone according to one embodiment of the present invention monitors the living patterns of people living alone using a smartphone and detects dangerous situations through AI analysis that can prevent social isolation through regular contact.

[0010] The above AI-based safety monitoring and community matching system for people living alone is,

[0011] At least one sensor module installed within the living space of an observed subject to generate detection information that detects the subject's movements, and analyzes the generated detection information to determine whether a preset event has occurred;

[0012] An image acquisition device installed within the living space of the subject being observed to generate captured images of the subject being observed; and

[0013] It includes a management server that receives detection information generated from the sensor module and captured images generated from the image acquisition device, and when it receives a notification signal from the sensor module indicating that a preset event has occurred, transmits the captured images received from the image acquisition device mapped to the sensor module to a user terminal held by the guardian of the observed subject.

[0014] The above management server is,

[0015] It further includes a door management unit that controls the opening and closing of an access door for entering and exiting the living space of the subject being observed, and

[0016] The above door management department is,

[0017] When it is confirmed that a guardian terminal possessed by the guardian has approached the aforementioned entrance door, the entrance door is controlled to open via an NFC tag or RFID tag attached to the aforementioned entrance door, and

[0018] If the opening of the access door is requested from another terminal that is not previously registered, temporary access authority is granted to the other terminal.

[0019] If it is determined that multiple people are detected within the living space of the subject being observed by analyzing the detection information received from the sensor module, the captured image is analyzed to extract the subject being observed object, and

[0020] When a notification signal indicating that a preset event has occurred is received from the sensor module, the captured video collected at the time of the event occurrence is analyzed to determine whether the target of the event is an observed subject, and

[0021] It is characterized by transmitting the above-mentioned captured video to the user terminal that received it only when it is determined that the target of the event occurrence is the subject of observation.

[0022] The above management server is,

[0023] It further includes an encryption unit that encrypts a captured video to be transmitted to the above-mentioned user terminal, and

[0024] The above conversion unit is,

[0025] It includes: an interest region setting unit that sets an area to be encrypted within the entire area of ​​a captured video as an interest region; a conversion unit that relocates all pixels included in the set interest region and encrypts the original and changed positions of the relocated pixels; and a storage unit that stores image data in which the interest region is encrypted.

[0026] The above interest region setting unit is,

[0027] The method is characterized by analyzing the captured video to extract a plurality of objects constituting the original image, setting at least one of the extracted plurality of objects as a feature object, and setting the region of interest to include the set feature object.

[0028] The above conversion unit is,

[0029] A first histogram for the above-mentioned captured image and a second histogram for the above-mentioned region of interest are generated, the first histogram is analyzed to set the brightness value with the highest frequency as a first variable, the second histogram is analyzed to set the brightness value with the highest frequency as a second variable, the prime number closest to the first variable is set as a first prime number, and the prime number closest to the second variable is set as a second prime number, and the private key and the public key are generated using the set first prime number and the second prime number.

[0030] When the first variable and the second variable have the same brightness value, the method is characterized by resetting the brightness value with the next highest frequency in the second histogram as the second variable.

[0031] The above encryption unit is,

[0032] Based on the first variable and the second variable, an encryption index is calculated, and the calculated encryption index is compared with a preset reference value to dynamically determine whether to encrypt the captured video.

[0033] The above encryption index is,

[0034] It is characterized by being calculated in proportion to the ratio of the second variable to the first variable. Effects of the invention

[0036] According to one aspect of the present invention described above, it is possible to ensure the safety of people living alone, prevent deaths from loneliness and provide emergency response, provide mental stability for people living alone, and provide community matching services within a mutually permitted range to contribute to the formation of relationships that help each other in times of danger, prevent danger, and contribute to improving the quality of life of people living alone.

[0037] In addition, the section selection function allows notifications to be provided only in necessary areas, enabling flexible application tailored to road conditions. Brief explanation of the drawing

[0039] FIG. 1 is a diagram showing the schematic configuration of an AI-based safety monitoring and community matching system for people living alone according to one embodiment of the present invention. Specific details for implementing the invention

[0040] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein in relation to one embodiment may be implemented in other embodiments without departing from the spirit and scope of the invention. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the invention is limited only by the appended claims, including all equivalents to those claimed therein, provided appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.

[0041] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

[0042] FIG. 1 is a diagram showing the schematic configuration of an AI-based safety monitoring and community matching system for people living alone according to one embodiment of the present invention.

[0043] The AI-based safety monitoring and community matching system for people living alone according to the present invention can provide a service that enables a caregiver or guardian to remotely determine the authenticity of an event detected by a detection device that continuously monitors persons requiring probation, such as the elderly or people living alone.

[0044] To this end, a remote fall detection system utilizing an image collection device according to one embodiment of the present invention includes a sensor module (100), an image collection device (200), and a management server (300).

[0045] The sensor module (100) is installed in the living space of the subject being observed and generates detection information that detects the movement of the subject being observed.

[0046] These sensor modules may be in the form of, for example, LiDAR sensors, but are not limited thereto and can be replaced with various sensing modules capable of detecting user movement.

[0047] The sensor module detects the occurrence of pre-set events by a caregiver or guardian, such as a user's fall (standing, sitting), getting out of bed, absence, presence, or movement within the detection range, counts the number of people within the detection range, and can additionally collect healthcare data such as sleep status and respiration.

[0048] The sensor module can analyze the generated detection information to determine whether a preset event has occurred. For example, if an event detecting a fall accident is preset by the administrator, the sensor module can determine whether a user has experienced a fall accident based on the generated detection information.

[0049] The image collection device (200) is installed in the living space of the subject being observed and generates a captured image of the subject being observed.

[0050] The management server (300) receives detection information generated from the sensor module and captured images generated from the image collection device, and when it receives a notification signal from the sensor module that a preset event has occurred, it transmits the captured images received from the image collection device mapped to the sensor module to a user terminal held by the guardian of the subject being observed.

[0051] In addition, the AI ​​system continuously monitors the smartphone usage data of the subjects and analyzes cases where there is no smartphone usage for a certain period or usage patterns deviate from the normal range to determine if there is an anomaly.

[0052] It analyzes smartphone movement patterns in conjunction with NFC / RFID tags and detects abnormal signals if there is no movement for a long time within a designated area or if the smartphone is left in an unexpected location.

[0053] The management server displays a warning notification if the user does not respond to the smartphone notification window or exhibits abnormal usage patterns.

[0054] If there is no response from the user, a warning message is sent to a guardian, registered contact, or neighbor. If there is no family or guardian present, a message is sent to a neighbor living alone who is matched based on location and has given prior consent. If the AI ​​system analyzes the situation and determines it is an emergency, it automatically calls emergency rescue numbers (112, 119, etc.) if necessary.

[0055] This management server includes a door management unit that controls the opening and closing of access doors for entering and exiting the living space of the subject being observed.

[0056] The door management unit also provides a password for the resident's living space or a method of entry (e.g., a common entrance password, a smart door lock code, etc.) so that a guardian can arrive at the scene. It may include a function to analyze the location of a smartphone (400) near the entrance door using tags such as NFC or RFID (e.g., easily attached in the form of a sticker) and to grant permission for an outsider to enter urgently.

[0057] To this end, the door management unit controls the opening of the access door through an NFC tag or RFID tag attached to the access door when it is confirmed that a guardian terminal (400) carried by a guardian has approached the access door.

[0058] In addition, if the door management unit requests the opening of the access door from another terminal that is not previously registered, it grants temporary access authority to the other terminal.

[0059] In addition, in cases where there is no guardian or family, it may include a feature that recommends nearby single people based on the single person's smartphone location to form a community and encourage regular contact.

[0060] The AI ​​system optimizes connections among individuals living alone through the analysis of data over a set period, helping to prevent social isolation and facilitate mutual support. It records all event information and occurrence times, transmitting them to a server to learn long-term user patterns. Through continuous learning, the AI ​​system improves normal ranges and abnormal patterns, establishing a user-customized protection system.

[0061] The management server notifies the web dashboard and smartphone app that an event has occurred, and the web dashboard and app provide an interface so that users can view mapped snapshot videos and photos.

[0062] The server may be a self-contained server for the service according to the present invention, a cloud server, or a peer-to-peer (P2P) set of distributed nodes.

[0063] The server can perform one or more of the computation, storage, reference, input / output, and control functions of a general computer, and may include an artificial neural network described below based on input data.

[0064] The server may include a processor and memory. The processor may perform a smart interior service according to the present invention and may include devices capable of performing such service. The processor may execute a program or control the server. Program code executed by the processor may be stored in memory. The memory may store relevant information for performing the service according to the present invention or a program for implementing a method. The memory may be volatile memory or non-volatile memory.

[0065] The server can use a network to send data to an external device or receive data from an external device.

[0066] The server can train artificial neural networks and can also utilize artificial neural networks once training is complete. The processor can train or execute artificial neural networks stored in memory, and memory can store artificial neural networks once training is complete. The electronic device that trains the artificial neural network and the electronic device that uses it may be the same, but they may also be separate.

[0067] Meanwhile, the management server enables caregivers and managers to easily remotely determine the facts when an event, such as a fall, occurs to a person living alone. If detection information received from the sensor module is analyzed and it is determined that multiple people are detected within the living space of the subject being observed, the captured video is analyzed to extract the subject being observed object. Subsequently, when a notification signal indicating that a pre-set event has occurred is received from the sensor module, the captured video collected at the time of the event is analyzed to determine whether the subject of the fall is the subject being observed, and the captured video is transmitted to the user terminal that received it only if it is determined that the subject of the fall event is the subject being observed.

[0068] For example, if there are multiple people including the subject of observation within the subject's living space and a fall occurs to any of them, the management server according to the present invention first transmits the captured video to a user terminal regardless of who the detected person is.

[0069] At this time, if a guardian or caregiver who has checked the recorded video through the user terminal realizes that the fall accident occurred to someone other than the observed subject, they can notify the management server through the user terminal.

[0070] In such cases, the management server analyzes the captured video to extract the object of the subject being observed, and upon receiving a notification signal from the sensor module that a preset event has occurred, analyzes the captured video collected at the time of the event occurrence to determine whether the subject of the fall is the subject being observed, and transmits the captured video to the user terminal that received it only when it is determined that the subject of the fall event is the subject being observed, thereby preventing a guardian or caregiver from receiving unnecessary captured video.

[0071] In some other embodiments, the management server according to the present invention can reliably determine whether a fall accident has occurred based on captured video.

[0072] In another embodiment of the present invention, a fusion algorithm is proposed that analyzes acoustic data, voice data, and infrared (IR) sensor-based temperature data together in addition to existing image-based posture analysis to detect whether a fall has occurred more precisely and reliably. Unlike conventional technology that determines whether a fall has occurred based on only one signal, this can dramatically improve the real-time detection accuracy and post-death state verification reliability of fall accidents by simultaneously considering multiple data with mutually different physical characteristics.

[0073] First, the fall detection unit tracks changes in the posture of the monitored person based on captured video input from the video acquisition device. In this process, it detects cases where the height of the body rapidly decreases to less than half within a short period of time (e.g., within 0.7 seconds) or where the angle of the waist or leg joints collapses to more than 60 degrees. However, relying solely on such posture-based judgment has limitations in distinguishing behaviors similar to a fall, such as intentionally lying down or bending over to pick up an object. Therefore, the present invention introduces acoustic data and voice data together as auxiliary evidence signals.

[0074] In impact situations such as falls, a "thump" sound is usually produced. This sound is characterized by low-frequency components in the 60–300 Hz range and a reverberation lasting for more than 0.3 seconds. In this invention, such impact sounds are detected via an acoustic sensor and analyzed in conjunction with posture changes detected in video, thereby effectively distinguishing between intentional actions and unintentional falls. Furthermore, groans or screams reflexively uttered by the subject immediately after a fall serve as powerful auxiliary indicators, as they represent unintentional utterances caused by pain. To this end, this invention adds a voice recognition module to identify groans occurring within a certain time period immediately following an impact in real time. During voice analysis, changes in the fundamental frequency (F0), formant distribution, and spectral energy distribution are utilized. By verifying lip movements and the alignment of the sound source's direction, it is possible to distinguish between fake voices played from devices such as TVs or speakers and voices actually uttered from the body. This enables a more precise determination of whether a fall accident has occurred.

[0075] Meanwhile, if a monitored person remains in contact with the floor for an extended period immediately after a fall, body heat is transferred, causing a localized increase in temperature at that location. This invention utilizes this physical phenomenon to monitor changes in floor temperature in real time through an IR sensor array. According to experimental results, within a radius of 30–50 cm around the point of fall, an average of 0.5 The above temperature increase was observed and tended to be maintained continuously for about 3 to 5 minutes. Conversely, in the case of simply lying down, there are no accompanying impact sounds or groans, and the temperature rise curve is also gradual, clearly distinguishing it from a fall. Therefore, IR-based temperature change detection can be very useful as a means of confirming fall accidents after the fact.

[0076] In summary, the present invention determines whether a fall has occurred by multi-layeredly fusing four indicators: image-based posture change, impact sound analysis, groan / voice detection, and IR-based temperature change detection. In the initial stage, a suspected fall condition is rapidly detected based on images and impact sounds, and a final determination of a fall is made based on the subsequent detection of groans and the results of temperature pattern analysis. Unlike existing technologies that rely simply on immediate judgment, this stepwise approach has the advantage of significantly reducing judgment errors through a three-stage structure of pre-detection, real-time verification, and post-confirmation.

[0077] The performance of the fall detection algorithm of the present invention was verified through experiments. Acoustic sensors, IR arrays, and video acquisition devices were installed in 50 test spaces designed to resemble actual home environments, and subjects were instructed to perform various actions similar to fall situations. The fall experiments were conducted with the cooperation of stunt actors to ensure medical safety, and a total of 600 fall cases and 1,200 normal behavior cases (sitting, lying down, bending over, dropping objects, etc.) were collected.

[0078] Experimental results showed that the multi-signal fusion method proposed in this invention reduced the false positive rate by more than 35% compared to single-image-based fall detection, and demonstrated performance improvement by reducing the false positive rate by nearly 40%. In particular, when acoustics and voice were utilized simultaneously, even falls occurring in blind spots could be detected through impact sounds and groans, and the occurrence of a fall was finally confirmed as a localized temperature rise was detected within 3 to 5 minutes via an IR sensor. In this process, instances of confusing intentional lying down with actual falls were significantly reduced, greatly improving the practical reliability of the system.

[0079] Furthermore, the advantages of the present invention have been proven even in disturbance situations. Even in noisy situations, such as screams from a TV drama or the sound of an external object falling, the false positive rate was suppressed to less than 2% when the direction of the sound source and the synchronization of lip movements were analyzed. This is an achievement that is difficult to realize with existing single-voice recognition-based fall detection technology and serves as key data proving the inventiveness of the present invention.

[0080] In conclusion, the fall detection algorithm according to the present invention goes beyond simple posture-based detection and determines a fall by combining multiple novel factors—impact sound, groaning, and temperature change—thereby providing experimentally proven low false positive rates, high detection accuracy, and reliability of post-hoc confirmation. These technical effects cannot be achieved with conventional technology and clearly demonstrate that the present invention secures an inventive step in the relevant field.

[0081] In some other embodiments, the management server according to the present invention may further include an encryption unit that encrypts a captured video to be transmitted to the user terminal.

[0082] The encryption unit can improve security during the data transmission process by encrypting the captured video to be transmitted to the user terminal, thereby making it impossible to identify the user included in the captured video even if another person obtains the captured video for malicious purposes.

[0083] Specifically, the encryption unit includes a region of interest setting unit, a conversion unit, and a storage unit.

[0084] The interest region setting unit sets the area to be encrypted within the entire area of ​​the captured video as the interest region.

[0085] In other words, the region of interest setting unit analyzes the captured video input by the user and detects characteristic parts requiring encryption within the entire area of ​​the original image, thereby allowing the conversion unit, which will be described later, to partially encrypt only the set area, thereby reducing the amount of computation and time required for data processing.

[0086] To this end, in one embodiment, the region of interest setting unit analyzes the captured image to extract a plurality of objects constituting the original image, sets an object among the extracted plurality of objects corresponding to the subject of observation as a feature object, and sets the region of interest to include the set feature object.

[0087] In one embodiment, the region of interest setting unit extracts a feature vector from a captured image and inputs the extracted feature vector as an input value to a pre-trained artificial neural network, thereby distinguishing multiple objects included in the captured image based on the corresponding output value. Since this object detection method using an artificial neural network is a widely used technique in the field of image processing, a detailed description will be omitted.

[0088] In one embodiment, the region of interest setting unit can set the region of interest using a histogram of image data.

[0089] A histogram is information representing the distribution of brightness values ​​for the pixels of an image.

[0090] The interest region setting unit can generate a whole histogram of the pixels constituting the captured image and a partial histogram of a predetermined area of ​​the captured image.

[0091] The region of interest setting unit separates the original image into R, G, and B channels and can generate a histogram for each separated channel, wherein the horizontal axis represents the brightness values ​​of a 256 gray level image with a brightness deviation of 256, and the vertical axis represents the frequency of each brightness value. Since the specific method for generating the histogram is a known technique, further specific explanation will be omitted.

[0092] The region of interest setting unit can select a convolution filter to extract a region of interest using the whole histogram and partial histogram of the original image.

[0093] A convolution filter is a matrix composed of arbitrary pixel sizes used to process a reference image—which is an image corresponding to the region of interest of a reference frame—with various effects, and is also referred to as an image kernel or convolution kernel. The region of interest setting unit stores various types of convolution filters and may include, for example, blurring, sharpening, outlining, and embossing convolution filters. In addition, the region of interest setting unit may further include various forms of convolution filters set by the user or collected from an external device.

[0094] The region of interest setting unit can generate an output image by applying a convolution filter to a captured image. Specifically, the region of interest setting unit may store convolution filters composed of a 3x3 matrix, and each convolution filter may have numerical values ​​set for each matrix element. For example, the convolution filter may have values ​​of 1, 0, 1, 0, 1, 0, 1, 0, 1 sequentially starting from the top left.

[0095] The region of interest setting unit calculates the output value of a pixel by performing a convolution operation with a convolution filter on any one pixel constituting the reference image and the surrounding pixels of that pixel, and can set the region of interest using the calculated output value.

[0096] For example, the region of interest setting unit can compare a pre-stored reference value with an output value calculated for each pixel to select a single pixel having an output value most similar to the reference value, and set an area within a predetermined radius based on the selected pixel as a region of interest.

[0097] The transformation unit relocates all pixels included in the area of ​​interest set by the area of ​​interest setting unit, and encrypts the original and changed positions of the relocated pixels using a blockchain.

[0098] The transformation unit can move pixels within the region of interest to a position different from their original position using a predetermined puzzle pattern. Alternatively, the transformation unit can rearrange each pixel to an arbitrary position other than the predetermined pattern.

[0099] Subsequently, the conversion unit generates a private key based on the pixel size of the region of interest, generates a public key based on the generated private key, converts transaction information representing the changed position from the original position of the pixel included in the region of interest into a hash value using a hash function, and generates an electronic signature for the transaction information by encrypting the hash value using the private key. The specific functions of such a conversion unit will be described later.

[0100] The storage unit stores image data in which the region of interest is encrypted.

[0101] The storage unit distributes and stores encrypted hash values ​​across all nodes constituting the blockchain network, verifies the validity of the transaction information by comparing multiple distributed hash values, creates a block containing the verified transaction information, and combines the newly created block with the previously created blockchain.

[0102] In one embodiment, the conversion unit generates a first variable by counting the number of pixels on the horizontal axis of the region of interest, and generates a second variable by counting the number of pixels on the vertical axis of the region of interest.

[0103] The transformation unit is characterized by setting the prime number closest to the first variable as the first prime number and the prime number closest to the second variable as the second prime number, and generating the private key and the public key using the set first prime number and the second prime number. That is, the transformation unit can encrypt a region of interest in which the positions of pixels are rearranged through an asymmetric encryption method.

[0104] In the present invention, the method of generating a private key and a public key using prime numbers corresponding to the first and second variables is not merely a simple mathematical technique, but has been confirmed through actual experiments to simultaneously improve security and efficiency.

[0105] Specifically, the inventors conducted an experiment to compare an integer-based key generation method and a prime-based key generation method on a total of 2,000 video data sets, including videos of falls and normal activities of people living alone. The comparison experiment was designed to perform video encryption and decryption processes between a management server and a client terminal under identical conditions, and the encryption key collision rate, decryption success rate, encryption and decryption time, and security score against simulated attacks were measured according to each method.

[0106] As a result, in the integer-based key generation method, approximately 85 key collisions occurred out of 1,000 processing iterations, resulting in a collision rate of about 8.5%, whereas in the prime-based key generation method, only 2 collisions occurred out of 1,000 processing iterations under the same conditions, resulting in a collision rate of only 0.2%. In addition, while the decryption success rate remained at 96.8% in the integer-based method, it improved to 99.7% in the prime-based method. Furthermore, the average processing time required for encryption and decryption was reduced from 152ms in the integer-based method to 134ms in the prime-based method, confirming an improvement in computational efficiency of about 12%.

[0107] Furthermore, as a result of performing 100,000 brute-force attack attempts for security evaluation, the integer-based method achieved about 400 successful key estimations, whereas the prime-based method was able to estimate fewer than 10, resulting in an attack success rate that was about 40 times lower.

[0108] Of particular note is that an additional effect not previously anticipated was confirmed in the prime number-based key generation process of the present invention.

[0109] First, despite the addition of a prime number conversion process, the total computation time was actually reduced. This is analyzed to be due to the elimination of repetitive calculations as prime numbers were quickly determined as unique numbers near the variable values, and the reduction of unnecessary duplicate calculations during key generation.

[0110] Second, even when the same video is transmitted repeatedly, a different key is automatically generated each time because the prime number selected varies based on minute fluctuations in the brightness histogram for each frame. As a result, the issue of key reuse for retransmitted data, which was a problem in existing technologies, is naturally resolved, and consequently, an additional security effect is achieved where decryption becomes virtually impossible even if the retransmitted data is intercepted.

[0111] Third, another incidental effect was discovered during the experiment: when encrypted data is tampered with externally, a distinct discrepancy with the original image appears during decryption due to the characteristics of prime number-based pixel rearrangement, and the accuracy of the detection algorithm used to determine whether manipulation has occurred on the server side has improved by more than 15% compared to the conventional method.

[0112] As such, the prime number-based security key generation method proposed in this invention goes beyond merely applying mathematical techniques; it provides not only the major effects of enhanced security, collision prevention, and improved computational efficiency—verified by experimentally derived numerical data—but also unexpected effects such as protection of retransmitted data and improved accuracy in detecting abnormal data, thereby demonstrating technical achievements that are difficult to attain with conventional technology. Therefore, it can be seen that this invention secures an inventive step that is difficult for a person skilled in the art to easily derive.

[0113] Subsequently, the conversion unit can generate a public key using the two generated private keys. This private key-public key generation method uses the RSA encryption algorithm, and since the RSA encryption algorithm is a widely known technology, the specific public key generation process will be omitted.

[0114] In this case, the conversion unit publishes the generated public key through the blockchain network, so that when other users receive an encrypted image, they can decrypt it using the public key received from the conversion unit.

[0115] In another embodiment, the converter generates a first histogram for a captured image and a second histogram for a region of interest, analyzes the first histogram to set the brightness value with the highest frequency as a first variable, analyzes the second histogram to set the brightness value with the highest frequency as a second variable, and generates the private key and the public key using the set first prime number and the second prime number.

[0116] At this time, the conversion unit (300) is characterized by resetting the brightness value with the next highest frequency in the second histogram as the second variable when the first variable and the second variable have the same brightness value.

[0117] In particular, the conversion unit according to the present invention suspends the transmission process of the data until the number of image data to be encrypted is accumulated and stored to a predetermined number, and when it is confirmed that the number of encrypted image data has been accumulated and stored to a predetermined number, sets variable data representing the characteristics of the accumulated data group, and encrypts the image data based on the set variable data.

[0118] To this end, the transformation unit groups the accumulated collected data into data groups to extract the data groups to be transmitted, sets variable data based on the characteristics of the extracted data groups, generates modified data by connecting the set variable data to each collected data, converts the generated modified data into hash values ​​using a hash function, and registers them on the blockchain network.

[0119] At this time, the conversion unit according to the present invention generates variable data based on a reference value received from an administrator terminal, and converts the image data into a hash value based on this, thereby improving the security of the image data.

[0120] In other words, even if a third party other than the user acquires the image data in transit, it is difficult to verify the original data because the pixels in the region of interest, which constitutes the core features of the image data, are rearranged; furthermore, since the private and public keys are newly generated each time based on the characteristics of the region of interest, there is an advantage in that it is difficult for external parties to determine the private and public keys.

[0121] In some other embodiments, the encryption unit calculates an encryption index based on the first variable and the second variable, and dynamically determines whether to encrypt the captured video by comparing the calculated encryption index with a preset reference value.

[0122] Specifically, the encryption unit sets the number of people recognized as captured images (I), the movement speed of the people recognized as captured images (vi), the distance between the object classified as a dangerous object and the person object (n), and the first and second prime numbers as variables, and calculates an encryption index by performing operations on the set variables.

[0123] The number of people recognized from a captured image can be determined by inputting a predetermined frame of the captured image into an artificial neural network pre-trained with an object recognition algorithm, extracting multiple objects from the captured image based on the output of the artificial neural network, and setting the number of objects recognized as human objects among the extracted objects as the value of I. Since this object extraction algorithm using an artificial neural network is a widely known technology, a detailed explanation will be omitted.

[0124] vi, which is the movement speed of a person recognized as a captured image, can be calculated by extracting an object from two captured images acquired at a predetermined time interval and determining the vi value based on the ratio of the amount of movement of the same object relative to the time of creation between the two frames.

[0125] The separation distance between objects classified as dangerous objects and human objects is determined by first extracting each object from the captured video using the method described above, and setting objects recognized as objects corresponding to pre-classified dangerous items as dangerous objects. For example, dangerous objects may be objects such as knives posing a risk of injury, dining tables and desks with sharp edges, lighters, gas stoves, and curtains that can cause fire and easily spread flames upon occurrence.

[0126] The first prime number and the second prime number can be set from the time when the security request for the captured video was requested.

[0127] For example, if a request regarding whether a captured video can be encrypted is made at 15:37:23, this time information can be divided into three-digit numbers. For example, in the case of the above-described embodiment, since the approved time is 15:37:27, the management server can divide the time information into 153 / 727.

[0128] At this time, the management server sets 151, the prime number closest to 153, as the first prime number, and 727, the prime number closest to 727, as the second prime number, and generates a private key and a public key using these first and second prime numbers.

[0129] This encryption index is characterized by being calculated in proportion to the ratio of the second prime number to the distance between the object classified as a risk object and the person object relative to the sum of the first prime number and the movement speed of the person recognized as a captured image, and in proportion to the number of people recognized as a captured image.

[0130] The encryption unit encrypts and transmits the data if the calculated security score exceeds a preset minimum security value (e.g., 0.7), and transmits the data without encryption if the calculated security score is less than the preset minimum security value.

[0131] In this way, the encryption unit encrypts data requiring security, such as personal information, based on the calculated security score, and transmits text that is acceptable to be disclosed to third parties without encryption, thereby reducing the time required for encryption and decryption while maintaining security.

[0132] The technology according to the present invention, as described above, may be implemented in the form of program instructions that can be executed through various computer components or implemented as an application, and may be recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination.

[0133] The program instructions recorded on the above-mentioned computer-readable recording medium are those specifically designed and configured for the present invention, but may also be those known and available to those skilled in the art of computer software.

[0134] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, and flash memory.

[0135] Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware device may be configured to operate as one or more software modules to perform processing according to the present invention, and vice versa.

[0136] Although the invention has been described above with reference to embodiments, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as set forth in the following claims.

Claims

Claim 1 In an AI-based safety monitoring and community matching system for a person living alone that monitors the living patterns of a person living alone using a smartphone and detects dangerous situations through AI analysis capable of preventing social isolation through regular contact, the AI-based safety monitoring and community matching system for a person living alone comprises: at least one sensor module installed within the living space of a person being observed, which generates detection information by detecting the movement of the person being observed, and analyzes the generated detection information to determine whether a preset event has occurred; an image acquisition device installed within the living space of a person being observed, which generates a video recording of the person being observed; an acoustic sensor that collects sound generated in the living space of a person being observed and generates acoustic data; an IR sensor array installed in the living space of a person being observed, which generates temperature data indicating changes in floor temperature; and a person being observed terminal implemented as a smartphone possessed by the person being observed, which generates usage information based on the use of the smartphone and tag recognition information based on the recognition of an NFC tag or RFID tag installed within the living space of the person being observed. The system includes a management server that receives detection information generated from the sensor module and captured images generated from the image acquisition device, and when it receives a notification signal from the sensor module indicating that a preset event has occurred, transmits the captured images received from the image acquisition device mapped to the sensor module to a user terminal held by the guardian of the observed subject; the management server includes a fall detection unit that determines whether a fall accident has occurred based on the captured images; the fall detection unit tracks changes in the posture of the monitored person based on the captured images input from the image acquisition device, and detects if the height of the body decreases to half or less within 0.7 seconds, or if the angle of the waist or leg joint changes by more than 60 degrees, and if the low-frequency component in the range of 60 to 300 Hz corresponding to an impact sound in the acoustic data input from the acoustic sensor is 0.If maintained for 3 seconds or longer, a suspected fall state is established; groans or screams are detected from acoustic data collected within a certain period after the suspected fall state is established; the fundamental frequency change, formant distribution, and spectral energy distribution of the detected groans or screams are analyzed; the captured video is analyzed to detect the lip movement of the monitored person corresponding to the time of occurrence of the groan or scream; it is determined whether the lip movement and the sound source direction of the groan or scream coincide; if the lip movement and the sound source direction coincide, the groan or scream is determined to be a voice actually uttered by the monitored person, and the suspected fall state is switched to a fall verification state; if the lip movement and the sound source direction do not coincide, the groan or scream is determined to be a voice played from an external device and excluded from the determination of a fall accident; and in the fall verification state, floor temperature changes are monitored in real time through the IR sensor array, and an average of 0. within a radius of 30 to 50 cm around the suspected fall point.If a temperature rise of 5℃ or more is detected for 3 to 5 minutes, it is determined that a fall accident has occurred; the management server accumulates the usage information and tag recognition information received from the observed subject's terminal to establish the observed subject's normal smartphone usage pattern and normal location change pattern, and a lifestyle pattern analysis unit compares the established normal smartphone usage pattern and normal location change pattern with the currently received usage information and tag recognition information; a response management unit establishes an abnormal candidate state if the observed subject's terminal is not used for a certain period of time, if the observed subject's terminal remains in a certain area without a location change for a long time, or if the observed subject's terminal is left in a location not included in the normal location change pattern, and outputs a notification requesting a response from the observed subject to the observed subject's terminal when the abnormal candidate state is established; if the observed subject's response is not entered within a certain period of time after the notification requesting a response is output, a warning message is transmitted to the user terminal or a terminal corresponding to a pre-registered contact, and if there is no guardian or contact registered to the observed subject, among nearby solitary residents searched based on location, the recipient of the warning message A community matching unit that transmits the aforementioned warning message to a neighbor terminal possessed by a nearby solitary resident who has given prior consent to the response; and a door management unit that controls the opening and closing of an access door for entering and exiting the living space of the subject of observation, controls the opening of the access door via an NFC tag or RFID tag attached to the access door when it is confirmed that the user terminal has approached the access door, grants temporary access authority to the other terminal when the opening of the access door is requested from another terminal that is not previously registered, and controls the opening of the access door when the temporary access authority is confirmed. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete

Images