CAMERA-BASED MOTOR DEVELOPMENT TOLERANT SIGNAL MOTION DATA PROCESSING AND ADAPTIVE IMAGE OUTPUT CONTROL SYSTEM AND METHOD
Patent Information
- Application Number
- TR202612079
- Authority / Receiving Office
- TR · TR
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2026-07-20
- Publication Date
- 2026-09-21
Smart Images
Figure 00000021_0000
Abstract
Description
1 TARIFF CAMERA-BASED MOTOR DEVELOPMENT TOLERANT SIGNALING MOTION DATA PROCESSING AND ADAPTIVE IMAGE OUTPUT CONTROL SYSTEM AND THE METHOD TECHNICAL FIELD The invention uses computer vision to reconstruct human motion from successive image frames. feature extraction, temporal image data classification, and imaging devices adaptively according to processed camera data. The invention relates to the technical aspects of controlling a standard web 10. proximity of hand, face and body from successive image frames taken from the camera extracting the points of proximity; a temporal proximity from these proximity points. the creation of a point sequence; the long-short temporal proximity point sequence Turkish Sign Language sign class, hand, processed by a long-memory-based iterative neural network Shape accuracy value, movement width deviation value and temporal flow deviation 15 It relates to the production of value. The invention also includes the accuracy value of the hand shape produced, and the deviation value of the range of motion. and the temporal flow deviation value, stored in memory and with user age data Hand shape accuracy threshold selected from associated numerical threshold sets, motion comparison of width, deviation tolerance and temporal flow flexibility; word 20 the target marker displayed on the screen according to the comparison result the display time between successive animation frames of the animation, image The repeat counter and the target marker data to be displayed are automatically generated. It is related to a system and method for making changes. STATE OF THE ART In the known state of the art, sign language recognition systems include data gloves and motion recognition devices. hand, face and body movements obtained using a sensor or camera It is based on matching with predefined signal classes. Camera In systems based on sequential image frames, hand, face, or body features are extracted. is being extracted; the temporal motion correlation between image frames is determined by recurrent neural wavelength. using networks, long-term and short-term memory networks, or convolutional neural networks These systems are processed. A significant portion of these systems are performed using image data. 2 the classification of the signal or the classified signal data as text or sound It is aimed at converting it into output. The document with publication number EP 3467707 B1 contains a first-person opinion. Three-dimensional dynamic hand movements in the images can be analyzed using deep learning. The identification is explained. The document in question shows hand 5 in the image frames. classification of postures and movement according to the temporal change of the posture sequence Although the class is determined; hand is calculated separately from image data. shape accuracy value, range of motion deviation value, and temporal flow deviation its value with three different numerical limit values associated with user age data. The comparison is not explained. The document also states that the comparison in question is 10. the representation of the result between successive frames of the animation displayed on the screen converting it into control data that changes the duration and image repetition counter It is not available. In document number CN 116453224A, one of the sign language images. Obtaining temporal proximity point sequences using a position detection tool, 15 training a long- and short-term memory-based sign language model with the sequences in question and converting recognized sign words into a sentence through natural language processing It is explained. The document in question shows the proximity point from the image frames. being close in terms of extraction and classification of temporal signal data together; hand shape accuracy value obtained from the classification process, movement 20 width deviation value and temporal flow deviation value, stored in memory, and Hand shape accuracy threshold, range of motion deviation associated with user age data. It does not compare tolerance and temporal flow flexibility separately. In the document also the display time of the comparison result between animation frames the number of repetitions of the target signal image and 25 in the next image processing cycle Converting the digital limit values to be used into control outputs that change the outputs It is not explained. The document with publication number CN 105956529 A describes the characteristics of sign language in education. used as data, and a type of long-term or short-term memory is formed with that data. The recurrent neural network is trained and the probability value of the signal to be recognized at the network output is 30. A sign language recognition method determined according to CN 108171198 A is described. The document with publication number [publication number] states that the continuous sign language video is asymmetrical and multi-layered. The automatic translation using long-term and short-term memory structures is being addressed. This 3 documents, matching temporal signal sequences with signal classes focusing on; the hand shape and spatial range of motion extracted from the same image data. and three different numerical limit values for temporal motion flow characteristics It does not include an evaluation. In the documents in question, this trio The evaluation result changes the screen display parameters and the new 5 a closed system that allows images to be reprocessed with updated limiting values It is also not explained how it is connected to the technical control loop. In the document with publication number WO 2019157344 A1, the posture is shown in the image frames. The signals are realized through neural network architecture by extracting estimated and optical flow information. Its timely recognition and use in sign language interpretation are explained. 10 The document also includes image data generation and modeling for training the neural network. There is an evaluation of his performance. However, the aforementioned The element modified in the document is the neural network determined according to the training dataset. These are the parameters. The control is created from the user's in-session visual data. According to their outputs, hand shape accuracy threshold, movement width deviation tolerance and 15 the temporal flow flexibility needs to be redefined and re-evaluated during the study. The determined values are then used in processing subsequent camera images. It is not explained. In current camera-based recognition systems, a fixed standard exists for all users. or a single acceptance threshold of 20 determined according to adult user images If used, the hand shape, range of motion, and temporal aspect of the target signal User actions that approach the flow to a certain extent are incorrectly flagged as invalid. They can be classified as such. Furthermore, in current systems, image data... The spatial and temporal deviation values calculated and displayed on the screen Among the image outputs, 25 numerical threshold sets selected according to user age data. and readjust the threshold values based on in-session control outputs. There is no closed technical control loop that determines this. Therefore, users with different range of motion and duration of movement The images are processed with the same fixed limit values; the animation frames on the screen display time between them, image repetition count, and subsequent image processing 30 The digital threshold values to be used in the cycle are obtained from the user's image. It cannot be automatically changed according to the measurement results. 4 THE TECHNICAL PROBLEM THAT IS INTENDED TO BE SOLVED The technical problem that the invention aims to solve is a standard web. hand, face and body extracted from successive image frames taken from the camera proximity points, hand shape, spatial range of motion, and 5 false rejections when using a fixed acceptance threshold due to variations in movement time. The goal is to reduce its output. The invention also provides an accuracy value for the hand shape calculated from image data. The range of motion deviation value and temporal flow deviation value are stored in memory. three distinct numerical limits stored and associated with user age data. The aim is to compare their values individually. 10 Another technical problem that the invention aims to solve is; the comparison result in the target marker animation displayed on the screen the display time between consecutive animation frames, the image of the target marker repeat counter and the numerical limit to be used in the next image processing cycle It is the conversion of the values into control outputs that automatically change. 15 With this invention, it is possible to capture a camera image and extract the image frames. Extraction of proximity points, calculation of temporal motion characteristics, Comparing the calculated values with the selected numerical limit values, screen display. Controlling display parameters, readjusting numerical limit values Determining and reprocessing new camera images with updated values 20 Establishing a closed technical control loop that includes the stages that is intended. A BRIEF DESCRIPTION OF THE INVENTION The system described in the invention consists of a standard webcam, the aforementioned web 25 at least one processor that processes successive image frames taken from its camera, processing commands, numerical threshold sets, calibration intervals, and image sequences Different displays based on processed image data stored in memory. It includes a screen that produces output based on its parameters. The system also includes a visual capture and motion 30 performed by the processor. The feature extraction unit displays an image with a motor development tolerance evaluation unit. The output includes a control and recalibration unit. Image capture and motion inference unit, standard web hand, face and body proximity in every image frame captured by the camera. extracting the points simultaneously and the proximity of successive image frames It forms a series of temporal proximity points from these points. The aforementioned temporal proximity point sequence, long-short memory-based recurrent neural network 5 processed through a network, a Turkish Sign Language sign class, a hand shape accuracy value, a A range of motion deviation value and a temporal flow deviation value are generated. The motor development tolerance assessment unit has multiple functions stored in memory. a set of thresholds associated with the user's age data from a set of numerical thresholds It selects the following threshold set: hand shape accuracy threshold, range of motion deviation 10. It includes tolerance and temporal flow flexibility. The motor development tolerance assessment unit evaluates the hand shape accuracy value. with the accuracy threshold, the range of motion deviation value range of motion deviation with its tolerance and temporal flow deviation value separately with temporal flow flexibility. It compares the hand shape accuracy value to the hand shape accuracy threshold, which is equal to or 15. high, range of motion deviation value equal to range of motion deviation tolerance or a low temporal flow deviation value equal to the temporal flow elasticity or If the first control output is low, then at least one of the aforementioned conditions If this is not provided, a second control output is generated. The video output control and recalibration unit has 20 on the first control output. This activates the first image sequence on the corresponding screen; the second control output. the sequential animation in the target marker animation shown on the corresponding screen It increases the display time between frames and the repetition counter for the target signal. It is changing. An in-session performance value of 25 from sequential control outputs. is calculated. The in-session performance value is a numerical value stored in memory. If the success threshold is reached, the next target signal image sequence is selected; If the value in question does not reach the success threshold, the hand shape accuracy threshold, At least one of the following is stored in memory: range of motion, deviation tolerance, and temporal flow flexibility. It is redefined within the defined calibration range. 30 By using redefined thresholds, tolerances and / or flexibilities, the web New sequential image frames are being captured from the camera, and the image in question... 6 The frames are reprocessed. This allows the camera image to be captured and the movement to be recorded. extracting features, comparing them with numerical limit values, displaying the screen. Controlling display parameters and updating new images a closed technical control loop including reprocessing stages with limit values is being created. 5 DESCRIPTION OF THE FIGURES Figure 1 shows the process of capturing successive image frames from a standard webcam. starting with; extracting the proximity points of hands, face and body from the image frames, Creation of a temporal proximity point sequence, movement with signal class 10 Generating numerical values for the properties, and assigning those values to the selected numerical values. Comparison with limit values, screen display based on the comparison result. control of parameters and recalibration of digital limit values It is a process flowchart showing the stages of the process. DETAILED DESCRIPTION OF THE INVENTION The invention concerns camera-based motor development tolerance signal motion data. processing and adaptive image output control system; a standard webcam, at least one that processes consecutive image frames received from the webcam in question The processor stores processing instructions and data used during image processing. at least one screen that generates image output based on image data processed with a small amount of memory. It is being implemented on [the site]. Within memory, proximity points can be extracted from image frames. the creation of a sequence of temporal proximity points, the sequence in question classification, motion feature values from image data 25 calculation, with the numerical limit values of the values in question comparison, modification of screen display parameters and numerical The commands for redefining the limit values are stored. Memory also contains targets for previously defined Turkish Sign Language sign classes. signal data, hand shape patterns corresponding to the target signal data, spatial 30 movement intervals, temporal movement flows, user age data associated numerical threshold sets, calibration of those threshold sets 7 ranges, imaging parameters and target signal image sequences It is hiding. At least one processor executes the processing instructions stored in memory, thus providing standard web functionality. image output created on the screen from image data received from the camera It creates a technical data processing and control cycle between them. 5 TAKING IMAGES A standard webcam shows the user a Turkish Sign Language sign. It produces sequential image frames during the process. The image frames target 10 within a specific sequence of squares from the beginning to the end of the signal movement It is transmitted to the processor. Thanks to the image data being obtained from a standard webcam, Data glove placed on the user: accelerometer, gyroscope, electromagnetic sensor without using a motion sensor or similar separate motion detection equipment Visual data related to signal movement is obtained. 15 Sequential image frames show the order in which the image frames were acquired. It is processed along with time information. Thus, the hand and face in each image frame are processed accordingly. and not just the placement of body positions at a single moment, but the image frames The spatial and temporal changes that occur throughout are also determined. EXTRACTION OF PROXIMITY POINTS The processor takes each image frame from the webcam and integrates it into a holistic system. The process involves extracting the proximity point. The result of this operation... proximity to the hand, face, and body parts within the image frame The points are determined simultaneously. 25 In one application, there are twenty-one proximity points for each hand, and four for the face. Sixty-eight proximity points are derived for the body and thirty-three for the female body. Hand proximity points; fingertips, finger joints, palm area and includes coordinate data corresponding to the wrist. Body proximity. The coordinates 30 correspond to the positions of the shoulder, elbow, wrist, and upper body. It includes data on facial proximity points such as eyebrows, eyes, nose, mouth, and the area around the face. It includes coordinate data corresponding to the regions. 8 The proximity points extracted from each image frame are two-dimensional and / or The three-dimensional coordinates are formed as a square coordinate set. Sequential The coordinate sets of the image frames, the chronological order of the image frames by bringing them together in sequence to form a series of temporal proximity points It is being converted. 5 The relative positions of the hand's proximity points to each other, the finger placement of the target index finger. and is used in determining palm placement. Hand and body proximity. The positional changes of the points between successive image frames indicate the movement of the signal. It is used in determining the spatial range of motion. Proximity points The sequence of changes and the duration of changes occurring throughout the image frames are indicated by symbol 10. It is used in determining the temporal flow of movement. Face proximity points are face-to-face interactions performed with the target marker. It is used to include their movements in the image data. Thus not just hand movement, but a holistic approach formed by the combined data of the hand, face, and body. Temporal data related to signal movement is obtained. 15 PROCESSING THE TEMPORAL PROXIMITY POINT SERIES The processor stores the generated temporal proximity point sequence in long-term and short-term memory. It transmits input data to a recurrent neural network based on long-term and short-term memory. a recurrent neural network based on 20 determines the motion correlation between successive image frames. by preserving the sequence of temporal proximity points stored in memory as a predefined Turkish Sign Language matches it with one of the sign classes. This process does not simply produce one class of signal; it targets According to signal data, a hand shape accuracy value, a range of motion deviation value and a temporal flow deviation value is also generated. 25 The hand shape accuracy value refers to the proximity of the user's hand points to the target marker. The hand is a numerical value representing the level of conformity to the proximity point arrangement. The shape accuracy value is determined by considering the fingertips, finger joints, palm area, and according to the correspondences of wrist proximity points in the target signal data Location suitability is being evaluated. 30 9 The range of motion deviation value is determined by the proximity of the user's hand and body to the points of contact. target marker with the spatial range of movement it covers across successive image frames It is a numerical value representing the difference between the spatial range of motion of the data. The temporal flow deviation value is the signal generated by the user. The transition times between image frames in the movement and the total movement time are the target 5. a numerical value representing the difference between the temporal flow of signal data. It is valuable. Hand shape accuracy value, range of motion deviation value, and temporal flow. The deviation value represents different physical motion characteristics. This Therefore, these values cannot be combined under a single general recognition possibility. 10 They are processed separately. SELECTION OF THE DIGITAL THRESHOLD SET In memory, multiple data are associated with different user age ranges. An additional set of numerical thresholds is stored. The user's age data, specifically the 15-year-old, is included. It is used as input data for selecting one of the numerical threshold sets. Age data is used for the purpose of conducting a psychological or medical assessment. not used; only pre-stored numerical threshold sets in memory It enables the selection of the relevant one. Each set of numerical thresholds; 20 a hand shape accuracy threshold, a range of motion deviation tolerance and It includes a temporal flow flexibility. The hand shape accuracy threshold is the hand shape accuracy threshold for target mark acceptance. It specifies the minimum numerical fitness value that the value must meet. 25 The range of motion deviation tolerance refers to the spatial extent of the movement performed by the user. the maximum allowed range between the range of motion and the spatial range of motion of the target marker It indicates a high numerical difference. Temporal flow flexibility, the gesture movement performed by the user The transition time between frames and the total movement time, along with the temporal 30 of the target marker. It specifies the maximum numerical difference allowed between the flow of motion. Thus, there is no single generally accepted threshold for evaluating image data. not used; for hand shape, spatial range of motion and temporal motion flow Three different numerical limit values are used. COMPARISON OF MOTION CHARACTERISTIC VALUES 5 The processor selects the numerical accuracy value of the hand shape generated from the image data. The hand shape in the threshold set is compared to the accuracy threshold. The processor also moves width deviation value, movement width deviation tolerance, and temporal flow It compares the deviation value to the temporal flow elasticity. The hand shape accuracy value must be equal to or higher than the hand shape accuracy threshold, 10 the range of motion deviation value is equal to or equal to the range of motion deviation tolerance its low and temporal flow deviation value equal to temporal flow elasticity If the voltage is low or not, the first control output is generated. If at least one of the three comparison conditions is not met A second control output is generated. The control output generated as a result of the comparison is 15. solely for the purpose of labeling a sign as correct or incorrect. not in use; the display parameters applied on the screen and numerical limit values to be used in the next image processing cycle It is used as numerical control data that enables modification. The first and second control outputs indicate which comparison conditions are met. 20 or can be produced with data showing that it is not provided. Thus, the hand shape accuracy value, range of motion deviation value, or temporal flow deviation which of the values falls outside the relevant numerical limit value It can be determined. IMAGE OUTPUT CONTROL ACCORDING TO THE FIRST CONTROL OUTPUT TO BE DONE When the first control output is generated, the processor displays the first image on the screen. It activates the sequence. The first image sequence shows the selected movement of the target marker. Visual reinforcement showing that it was performed within numerical limit values 30 the image and / or the next target marker image stored sequentially in memory It can include the series. 11 The first control output also includes the repeat counter for the current target signal. It enables the image data to be stopped or reset. Thus, the image data is divided into three separate parts. The first control output obtained by comparing it with the numerical limit value, the sequence of images to be displayed on the screen and the operating status of the repeat counter It is converted into a technical control data that changes the data. 5 CONTROLLING THE IMAGE OUTPUT BASED ON THE SECOND CONTROL OUTPUT. If the second control output is generated, the processor displays the target on the screen. It modifies at least one of the display parameters of the signal animation. In one application, the successive animation frames in the target marker animation are 10. The display time between consecutive animation frames is being increased. Increasing the display time results in the target marker animation's playback speed. is being reduced. The second control output also includes the repeat counter for the target signal. It increases the playback speed of the same target signal image sequence. It is being displayed again. 15 Voice guidance output is also connected to the second control output. It can be enabled. However, the basic technique in the display parameter The change is to alter the display time between consecutive animation frames. Hand shape accuracy value, range of motion deviation value, or temporal flow. If at least one of the deviation values falls outside the relevant numerical limit value... 20 As a result, a second control output is generated, obtained from the camera image. spatial and / or temporal difference display time and repetition counter on the screen It is converted into technical control data that directly changes the system. CREATING IN-SESSION PERFORMANCE EVALUATION 25 The processor processes the first and second control values generated in successive image processing cycles. their outputs, target signal repetition counts, and processed signal classes within the session. It stores performance data in memory. A numerical in-session performance value from in-session performance data. is being calculated. 30 12 In-session performance metric; from the ratio of the primary control outputs to the total control outputs, from the number of the second control outputs, from the number of repetitions per target signal, from the number of signal classes processed or 5 It can be created by bringing together the data in question. In memory, a numerical value is stored for comparison with the in-session performance value. The threshold for success is hidden. A success threshold is an abstract measure of a user's pedagogical or psychological success. This is not an evaluation; it's the result of a specified number of image processing cycles. a limit used to evaluate the numerical distribution of control outputs It is valuable. If the in-session performance value reaches the numerical success threshold The next target beacon image sequence stored in memory is selected. In-session If the performance value does not reach the numerical success threshold, the numerical threshold is 15. The calibration process for re-assigning the team is being initiated. RECALIZING DIGITAL LIMIT VALUES In memory, there is a calibration interval for each set of numerical thresholds. It is stored. The calibration range is the hand shape accuracy threshold, with a range of motion of 20°. The lower and upper limits where deviation tolerance and temporal flow flexibility can be changed. It includes numerical limits. If the in-session performance value does not reach the numerical success threshold processor, hand shape accuracy threshold, movement width deviation tolerance, and temporal flow. It redefines at least one of its flexibility features within the relevant calibration range. 25 During the redefinition process, which comparison condition Control data showing that it is not provided can be used. Hand shape accuracy If the value falls below the hand shape accuracy threshold, the hand shape accuracy threshold, It can be reset within the calibration range defined in memory. The range of motion deviation value is 30 times the range of motion deviation tolerance. If exceeded, the range of motion deviation tolerance, as defined in the calibration memory, It can be redefined within this range. 13 If the temporal flow deviation value exceeds the temporal flow elasticity temporal flow flexibility, re-calibration within the calibration range defined in memory. It can be determined. These three numerical limit values can be used independently or together. It can be redefined. 5 The recalibration process involves long- and short-term memory-based iterative neural circuitry. Recalibration differs from changing the training parameters of the network. During this process, the weights of the classification model are not retrained; image The numerical limit used in accepting the three motion characteristics obtained from the data. The values are changed during the process. 10 CLOSED TECHNICAL CONTROL LOOP After the numerical limit values were redefined, the standard web New sequential image frames of the user are being captured from the camera. New The proximity points of hands, face, and body are extracted again from the image frames, creating a new 15 A sequence of temporal proximity points is created, and this sequence includes long- and short-term points. It is reprocessed using a memory-based iterative neural network. The hand shape accuracy value and range of motion obtained from the new image data. Deviation value and temporal flow deviation value, redefined hand shape accuracy threshold, range of motion deviation tolerance and / or temporal flow flexibility 20 They are being compared. According to the comparison results, the screen display parameters were repeated. It is being controlled and a new control output is being generated. Like this; Capturing image frames from a standard webcam, 25 Extracting proximity points of hands, face, and body from image frames, Creation of a sequence of temporal proximity points, Generating three separate motion characteristic values from image data, Comparing these values with three different numerical limit values, According to the comparison results, the screen display parameters are 30 changing, Calculation of in-session performance value, 14 redefining the numerical limit values and Reprocessing new camera images with updated limit values A closed technical control loop is created, including the stages. In this closed technical control loop, the camera's physical image data... It produces spatial and temporal motion characteristics from image data by the processor. 5 extracting, storing, and performing memory comparisons of numerical limit values used in the comparison. Image output with different display parameters according to screen control outputs. It constitutes. Thanks to this structure, the evaluation of image data is based solely on a fixed approach. It does not adhere to the classification threshold; hand 10 is calculated separately from the image data. The shape, range of motion, and temporal flow characteristics of the numerical limits differ from one another. It is evaluated based on its values, and the evaluation result shows the screen's operation. parameters and the limiting values to be used in the next image processing cycle It is changing. EXAMPLE APPLICATION In an application, the user's age data is collected as system input data. and the set of numerical thresholds associated with that age data is retrieved from memory. The target Turkish Sign Language image sequence is selected. This is shown. During the user's target execution, standard web 20 Sequential image frames are captured from the camera. Points of proximity between the hand, face, and body are extracted from each image frame. and a temporal sequence of proximity points from those proximity points It is being created. The temporal proximity point sequence is based on long-term and short-term memory. The signal class, hand shape accuracy value, and other parameters of the target signal are processed using a recurrent neural network. The range of motion deviation value and temporal flow deviation value are generated. Hand shape accuracy value is the hand shape accuracy threshold, and the range of motion deviation. value of movement width deviation tolerance and temporal flow deviation value This is compared to temporal flow elasticity. If all three comparison conditions are met, the first control output is 30. is being generated, the first image sequence is activated on the screen, and the current target signal is displayed. The corresponding repeat counter is being terminated. A second check will be conducted if at least one of the comparison conditions is not met. The output is being generated, sequential animation frames in the target marker animation. The display time between them is increased, and the repeat counter for the target signal is also increased. is being increased. In-session performance value from sequential control outputs. is calculated. The in-session performance value is a numerical value stored in memory. If the success threshold is reached, the next target signal image sequence is selected. If the in-session performance value does not reach the numerical success threshold hand shape accuracy threshold, range of motion deviation tolerance, and temporal flow. At least one of its flexibility is that it can be re-defined within the calibration range defined in memory. It is determined. New consecutive image frames from a standard webcam are 10. is being obtained and new image data is being redefined according to numerical limit values. It is reprocessed using... 20 30
Claims
16 REQUESTS 1. Camera-based Turkish Sign Language movement data in a computer system. It is a processing and image output control method, and its feature is; - Target 5 achieved by a user from a standard webcam. Capturing sequential image frames of Turkish Sign Language sign movements, - The proximity points of the hand, face, and body are matched in each image frame. extraction in time and proximity of successive image frames Creating a temporal proximity point sequence from these points, - 10 long- and short-term memory-based temporal proximity point sequences A class of signals, a hand shape accuracy, is processed using a recurrent neural network. the value, a range of motion deviation value and a temporal flow deviation the production of value, - User's age from multiple sets of numerical thresholds stored in memory Selecting a set of numerical thresholds associated with the data and the selected 15 a hand shape accuracy threshold from a numerical threshold set, a range of motion deviation tolerance and achieving temporal flow flexibility, - hand shape accuracy value with hand shape accuracy threshold, range of motion deviation value, range of motion, deviation tolerance, and temporal flow. Comparison of the deviation value with the temporal flow elasticity, 20 - the hand shape accuracy value is equal to or higher than the hand shape accuracy threshold, the range of motion deviation value the range of motion deviation tolerance equal to or lower than the temporal flow deviation value of the temporal flow If the first control output is equal to or lower than its elasticity, then... If at least one of the comparison conditions is not met, 25 Generating the second control output, - the first image sequence on the screen corresponding to the first control output activation and repeat counter for the current target signal termination, - Target signal 30 displayed on the screen corresponding to the second control output. the representation between successive animation frames in the animation Increasing the duration and increasing the repeat counter for the target signal, 17 - an intra-session performance value from sequential control outputs calculation, - the in-session performance value is stored in memory as a numerical success threshold. If it does not reach the hand shape accuracy threshold, range of motion deviation tolerance and temporal flow flexibility, at least one of which is present in memory 5 re-determination within the defined calibration range and - new sequential image frames taken from a standard webcam redefined hand shape accuracy threshold, range of motion deviation reprocessing using tolerance and / or temporal flow flexibility It is characterized by including procedural steps. 10 2. The method according to Claim 1, its characteristic is that each image frame includes the hand, face, and generating two-dimensional and / or three-dimensional proximity points of body regions It is characterized by being processed through a holistic proximity point extraction operation.
3. The method according to claim 1 or 2, characterized by its ability to extract each image frame from each Twenty-one proximity points for one hand, four hundred and sixty-eight proximity points for one face. 15 It is characterized by the extraction of thirty-three proximity points for the point and the body.
4. It is a method according to claim 1, and its characteristic is; - hand shape accuracy value, targeting the user's hand proximity points. from the conformity of the hand proximity point arrangement to the sign, - the range of motion deviation value of the hand and 20 in successive image frames the spatial range of motion covered by the body's proximity points and the target from the difference between the spatial range of motion of the signal and - temporal flow deviation value, transition times between image frames and Total movement time and temporal movement flow of the target marker. It is characterized by being calculated from the difference between them. 25 5. This method, according to claim 1, is characterized by its ability to store the user's age data in memory. one of the stored first digital threshold sets or second digital threshold sets It is characterized by being used as index data for selection purposes.
6. This method, according to claim 1, is characterized by the following: the second control output; - 30 between consecutive animation frames in the target marker animation increasing the screening time, - Increase the repeat counter for the target signal and 18 - the same target signal image sequence again with increased display time It is characterized by its ability to enable visualization.
7. This method, according to claim 1, is characterized by its in-session performance value; - Ratio of primary control outputs to total control outputs, - Number of second control outputs, 5 - Number of repetitions per target signal and - calculated using at least one of the processed signal classes It is characteristic.
8. This method, according to Claim 1, is characterized by its numerical value of in-session performance. if it does not reach the threshold of success; 10 - the hand shape accuracy value remains below the hand shape accuracy threshold corresponding hand shape accuracy threshold, - the range of motion deviation value of the range of motion deviation tolerance In contrast to exceeding the range of motion deviation tolerance, - the temporal flow deviation value exceeding the temporal flow elasticity 15 corresponding temporal flow flexibility defined in memory for the relevant calibration. within the range, independently or together again It is characterized by its determination.
9. A camera-based Turkish Sign Language gesture data processing and adaptive system. It is a video output control system, and its feature is; 20 - a user's target Turkish Sign Language sign movement a standard webcam that produces sequential image frames, - operation commands, target signs belonging to Turkish Sign Language sign classes data, numerical threshold sets, calibration intervals, and target markers. at least one memory location that stores image sequences, 25 - Hands, face, and body from each image frame captured by the webcam. extracting proximity points simultaneously, the proximity in question forming a sequence of temporal proximity points from points and temporal proximity point sequence long- and short-term memory-based iterative neural network By processing it through a network, a signal class, a hand shape accuracy value, a movement 30 minimum that produces a width deviation value and a temporal flow deviation value a processor, 19 - User's age from multiple sets of numerical thresholds stored in memory selects a set of numerical thresholds associated with the data, and the selected numerical a hand shape accuracy threshold from the threshold set, a range of motion deviation at least one of these that achieves tolerance and a temporal flow flexibility processor, 5 - hand shape accuracy value with hand shape accuracy threshold, range of motion deviation value, movement width deviation tolerance, and temporal flow Comparison by comparing the deviation value with temporal flow elasticity. If both conditions are met, the first control output and If at least one of the comparison conditions is not met, the second 10 At least one of the processors that generates the control output, - the second activates the first image sequence in response to the first control output. sequential animation in the target marker animation corresponding to the control output display time between frames and repetition counter for the target signal a screen that changes, 15 - Calculates an in-session performance value from sequential control outputs and in-session performance value, numerical success stored in memory. If it does not reach the threshold, the hand shape accuracy threshold, range of motion at least one of the following is required in memory: deviation tolerance and temporal flow flexibility. redefining within the defined calibration range, the aforementioned EN 20 It is characterized by containing a small number of processors.
10. The system is defined according to claim 9, and its characteristic is that at least one processor displays each image. twenty-one proximity points for each hand from the square, four hundred and sixty for the hundred. eight proximity points and thirty-three proximity points for the body It is characterized by its structure. 25 11. The system is defined according to claim 9, and its characteristic is that the screen corresponds to the second control output; - between successive animation frames in the target marker animation will increase the screening time, - Playback of target marker animation according to increased display time will reduce its speed and 30 - at least one of the same target signal image sequences will be displayed repeatedly. It is characterized by being controlled by a processor.
12. A system according to claim 9, characterized by having at least one processor; - the hand shape accuracy value remains below the hand shape accuracy threshold corresponding hand shape accuracy threshold, - the range of motion deviation value of the range of motion deviation tolerance Despite exceeding the range of motion deviation tolerance, 5 - the temporal flow deviation value exceeding the temporal flow elasticity corresponding temporal flow flexibility defined in memory for the relevant calibration. within the range, independently or together again will determine and new sequential images taken from a standard webcam. The squares are redefined with a hand shape accuracy threshold and a range of motion of 10. by using deviation tolerance and / or temporal flow flexibility again It is characterized by being structured in a way that allows it to function. 20 30