TR202022823A2 - STIMULATOR DEVICE WORKING WITH ARTIFICIAL INTELLIGENCE SUPPORTED SENSORY REPLACEMENT FOR PROSOPAGNOSIA PATIENTS - Google Patents

Abstract

The invention allows prosopagnosia (face blindness) patients to distinguish faces, recognizing the image taken by the camera with an artificial intelligence-supported face recognition algorithm, creating separate signals for each face and distinguishing the faces through the patient's skin with electrical stimulation, and also causing neural plasticity by using it for rehabilitation. It is about a support and rehabilitation device consisting of camera, microprocessor system, communication unit, oscillator unit that creates the necessary waveform for skin stimulation and conductor array components, which aims to create new pathways in the brain and thus restore facial recognition function.

Description

DESCRIPTION ARTIFICIAL INTELLIGENCE ASSISTED SENSORY SUPPLEMENT FOR PROSOPAGNOSIA PATIENTS STIMULATOR DEVICE WORKING WITH THE METHOD TECHNICAL FIELD: The invention is designed for patients with prosopagnosia (face blindness) to distinguish faces. With the artificial intelligence supported face recognition algorithm of the image taken with the camera, recognition and electrical stimulation by creating separate signals for each face. Rehabilitation, which enables the differentiation of faces through the skin of the patients. new pathways in the brain by causing neural plasticity The camera, which aims to restore the face recognition function, microprocessor system, communication unit, necessary waveform for skin stimulation a support consisting of the oscillator unit and the conductor array components, and relates to the rehabilitation device. PRIOR ART Prosopagnosia, also called face blindness, in Greek prosopon (face) and agnosia (not knowing). Faces, including one's own face It is a cognitive disorder that he cannot recognize. Intellectual abilities of people with this disorder levels do not show any problems. The term originally came from the acute brain. refers to a condition that follows injury (acquired face blindness), but The disorder also has a congenital or developmental form and is present in all cases. prevalence rate is 2.5% (Grüter, Grüter, Carbon, 2008). Usually with prosopagnosia associated specific brain area, particularly in the temporal lobe that is activated in response to faces It is the fusiform gyrus. Thanks to the functionality of the fusiform gyrus, the faces are complex. details are recognized in more detail than inanimate objects. the aforementioned brain The fusiform gyrus located in the right hemisphere of the hemispheres is more in the face recognition process. is active. The fusiform gyrus is specific only for the recognition of human faces. is not yet known. Prosopagnosia as a result of the lesion, occipito- It is caused by damage to the temporal lobe and is mostly found in adults. Improvement Although there have been several attempts for treatment, no therapy has produced a definitive and permanent solution.

In the current literature, the name of the person to whom the face belongs to the screens with face recognition systems. authored applications or sound notification devices are available, but these systems such as regaining the facial recognition ability by creating new pathways in the brain. have no abilities.

BRIEF DESCRIPTION OF THE INVENTION: The flow of life in prosopagnosia patients with face recognition problems thanks to the invention While continuing, they will both be able to distinguish faces and be rehabilitated. Like this they will be able to alleviate the problems they face in maintaining their daily lives. your device Because it is portable and wearable, instant data flow will be possible. instead of electrical stimulations and additionally mobile phone messages if desired Using it will also prevent possible interference with the environment.

In the later stages of rehabilitation, the brain faces facial stimulation by electrical stimulation. The patient will begin to analyze the details, until the device is no longer needed. Daily life will continue with the device by the device.

An artificial intelligence supported algorithm embedded in the images taken from the camera The system with a microprocessor, which recognizes by means of It delivers the electrical stimulation taught to the patient to the patient's skin. image and electrical stimulation association works in the visual cortex region of the brain. increases, creates urges to distinguish, and the treatment process begins. The proposed invention is the first developed for the treatment of patients with prosopagnosia. is the device. Unlike the existing ones, it is both supportive and therapeutic. is the system. No other device has the ability to rehabilitate.

LIST OF FIGURES: Figure 1. Artificial Intelligence Assisted Sensory Substitution for Patients with Prosopagnosia Contact Unit of the Stimulator Device Working with the Method Figure 2. Artificial Intelligence Assisted Sensory Substitution for Patients with Prosopagnosia Block Diagram of the Stimulator Device Working with the Method Figure 3. Artificial Intelligence Assisted Sensory Substitution for Patients with Prosopagnosia Flow Chart of the Stimulator Device Working with the Method EQUIVALENT OF THE NUMBERS USED IN THE FIGURES: 1. Communication Unit 1.1. USB 1.2. Ethernet 1.3. Bluetooth 2. Micro Processor 3. Camera 4. Stimulator 4.1. Sign Generator 4.2. Power Amplifiers 4.3.DEMUX . Contact Unit DETAILED DESCRIPTION OF THE INVENTION The invention is a sensory replacement device with facial recognition. Sensory substitution devices; three consists of the main component. These are the camera (3) that detects the image from the environment, contact and visual information with electrical stimuli and user skin contact unit. (5) providing communication between the transmitting stimulator (4) and camera (3) and the stimulator (4), appropriate stimulation, in which image processing and face recognition algorithms are performed. that creates the pattern and interacts with a mobile phone or personal computer. It is a communication unit (1) with a microprocessor (2). Communication unit (1), USB (1.1).

It includes ethernet (1.2) and bluetooth (1.3). Images detected by the camera (3) in the invention transferred to the microprocessor (2) unit. Face written on the microprocessor (2) unit The faces detected by the recognition algorithm are recognized, named and for each face. A separate signal pattern is created. This generated signal is the patient contact unit. It is given to the patient via the stimulator (4), which is in contact via (5), and the incoming It is ensured that they can distinguish faces by means of signals. Stimulator (4); sign generator ( contains. In-kind at the same time, via bluetooth (1.3) to a mobile phone or personal computer. Necessary notifications for the patient's habituation process by sending facial name tags has been provided.

First of all, a stimulator (4) circuit is designed for this device. designed circuit human contact unit (5) is implemented on a suitable organic pcb. has been produced. Our stimulator (4), originally designed as a 10 x 10” matrix It provides an image with a resolution of 100 pixels. in higher resolution possible to do. The contact unit (5) produced is double-layered, with a length of 104 mm and a width of 76 mm. has. PCB is a very flexible product with a thickness of 0.6 mm. After production, It comes in the form of a 10 x 10 matrix that will provide stimulation to the user.

This device stores a 100-bit long pattern for known persons. It applies the pattern of the people to the patient as long as the person concerned is in the image.

In addition, the system does not use the stimulator for 0 people, even if it comes from photo and video sections. It ensures that the patient always gives the same patterns. In this way, the patient is together with the device. After a short study, with the photos of his family and those they want, introduced to the system. distinguishes faces detected by the device camera during normal life flow can recognize.

Facial recognition systems have become very popular as a result of the advancement of artificial intelligence methods. They are advanced and very useful systems. Convolutional artificial neural Even if their networks are particularly successful in image recognition, they still need systems with powerful processors. and there are implementations that require large memories. Recommended portable It is not possible for them to be realized with high performance for the system. Because face detection, segmentation and recognition processes with embedded hardware with artificial intelligence and face recognition algorithms that can be applied has been carried out.

Face recognition systems work in two stages. In the first stage it is determined whether there is a face and if there is, where this face is. detect later segmented by this face recognition algorithm and found in the database. It is decided to whom it belongs by comparing it with the labeled faces.

Face detection is the first step. so much to make it happen Haar classifier, which is an efficient object detection method, is used. Machine It is a learning-based method.

Initially training the algorithm and for this training there are many positives (hundreds of (containing) and negative (without faces) images are needed. Later feature subtraction is done. 24 x if all haar characteristics such as location, scale, species are taken into account For a 24 image we would have over 160,000 attributes, but these processors This is too large a number and needs to be reduced. from extreme quality AdaBoost to avoid and extract enough attributes that work for us. algorithm is used.

AdaBoost algorithm has many other learning methods to improve performance. can be used with the algorithm type. Other learning algorithms (weak learners) output is weighted, representing the final output of the amplified classifier. combined into a total. AdaBoost is designed to remove unnecessary features and only Used to select features. For example; relevant when detecting a vertical edge nose it is a feature but it is useless for lips. Thanks to AdaBoost this is irrelevant in a 24 x 24 image with parameters from properties disabled The AdaBoost algorithm uses the face and face in the image to detect face. It tries to separate non-existent objects, that is, it performs a classification. For this first It starts by assigning equal weight values to the points in the iteration and creates the classifier. In the second iteration, this first iteration was misclassified. the weights of the points (total correct weight/total wrong weight) x initial weight formula are changed to be equal, and thus the total correct weights and the total incorrect their weights are equalized. Each of these iterations creates weak classifiers. Final the sum of these is taken for the result. The higher the number of iterations, the higher the classifier the final success rate will be higher. Let's explain this algorithm with an example.

Let there be an environment with five red and five blue dots. AdaBoost algorithm In the first iteration, it starts the classification process by giving a weight value of 1 to each point.

As a result of the first iteration, three blue dots were classified as red.

Total weight of misclassified points and correctly classified points we need to re-equalize their total weights so the initial weights of the blue dots the total number of points classified correctly and the total points classified incorrectly Multiply the ratios of the numbers. The initial weight value is 1, the total weight of the right points is 7, The total weight of the wrong spots is 3. The value of 2.33 obtained as a result of 1*(7/3) operation It is assigned as the new weight value of the misclassified points. correctly classified the weights of the points are not changed and the second iteration is started with these weight values.

As a result of the second iteration, the classification process takes the form shown in Table 1. This Once a red and two blue dots are in the wrong classes. The above-described formula is again is applied and the new weight values of the three misclassified points are set as 3.66. is calculated. Three classified points at the end of the third and final iteration We get the set. These three are the weak learners mentioned above. Now this is your turn there is a consolidation of weak learners. 3 iterations while doing this merge operation All resulting classes are taken into account. While doing this calculation, each class log(d0correct class/wrong class) values are used. ”41° 01 30 +13; 71.35 +13CI +1.35 +03; '0-35 _0.55 .130 '1-30 _1.30 +1.&5 ”'35 *1-85 1.40 *MU *4-00 Table 1 - Consolidation Coefficients for Each Region The methods used for face recognition are divided into two as traditional and modern methods. We can examine it in class. Detecting the geometric features of faces and Traditional methods of comparing faces include EigenFaces (Essential Faces), FisherFaces and LBPH (Local Binary Paterns Histograms) algorithms are the most common are the ones.

The modern methods used in face recognition are machine learning and deep Includes learning techniques. In this method, learning algorithm fed with new images They go through the process of getting to know faces better over time. A deep learning Once the algorithm has learned enough patterns or features from the data, have the ability to extract features from a digital image they have not seen before and this example or previous images whose properties are stored in the database. will be able to recognize who the picture belongs to by comparing it.

One of the face recognition methods mentioned above is LBPH (Local Binary) in this device. Patterns Histograms) were used. Deep learning methods, convolutional need high processing power, such as deep learning neural networks Therefore, the LBPH method was used. LBPH algorithm histogram values using classification. Like other classification algorithms, in LBPH should be trained with visual data. LBPH because our faces are made of micro patterns suitable for classification. LBPH analyzes images in 3*3 matrices and attaches special importance to the value at the center of this matrix. Neighboring pixel of this center The pixel values in the center are compared with the values of the center pixel. Greater than center value pixels are set to 1*e, pixels smaller to 0. Then the binary values of each blog It is converted to a histogram by converting it to a decimal number. Finally. we are interested in This block is used to generate a single feature vector for an image containing all features. histograms are combined.

Table 2 - Pixel Values to Binary Formata and Histogram Value in LBPH Conversion The block diagram of the device is shown in Figure 2. Here it is detected by the camera (3) Face recognition algorithm embedded in the microprocessor (2) unit. recognizable and recognizable facial stimulation is applied to the stimulator to meet the patient's lecture. (4) is transferred. Differentiating faces with the stimulator (4) requires a process in the first place. available in the communication unit (1) for the recognition of faces that the user cannot distinguish Whose face belongs to the user's mobile phone via bluetooth (1.3) It is also possible to send information in text form. In addition, the standard for the device All settings except the operation of the mobile phone via bluetooth (1.3) connection. or with a personal computer. The device is suitable for patients with prosopagnosia. It is the first system developed for the treatment of

Claims (3)

REQUESTS 1. It is a stimulator device that works with artificial intelligence supported sensory substitution method for patients with prosopagnosia. The communication unit (1), which enables the camera (3) to communicate with the stimulator (4), where image processing and face recognition algorithms are performed, which creates the appropriate stimulation pattern and interacts with the mobile phone or personal computer, recognizes the faces detected by the face recognition algorithm, recognizes the named and named faces. The unit with a microprocessor (2) that creates a separate signal pattern for each face, the camera (3) that detects the image from the environment, the stimulator (4) that enables the patient to distinguish faces through these signals given and delivered, the visual information transmitted by the stimulator (4) to the skin of the user. It is characterized by a contact unit (5) that sends electrical impulses. 2. It is the communication unit (1) mentioned in claim 1, and its feature is; It is characterized by including USB (1.1), Ethernet (1.2) and bluetooth (1.3). 3. It is the stimulator (4) mentioned in claim 1, its feature is; The signal generator (4.1) is characterized in that it contains power amplifiers ( ).