Robot Control with Speech Processing
Advisor: Asst. Prof. Serap Kırbız, Asst. Prof. Yusuf Aydın
Team: Baran Yiğit Aladoğan, Ali İhsan İlhan
Keywords: Robot, robotic, control, distance sensor, line sensor, ultrasonic distance sensor, proximity sensor, Arduino, sound processing, machine learning, neural networks, convolutional neural networks, artificial neural networks, artificial intelligence, dynamic time warping
Abstract: The aim of this project is to design a robot that can pick one-word commands via speech and executes them correctly. This report focuses on speech recognition and robotics parts. For speech recognition, deep learning methods are used. Convolutional Neural Networks (CNN) and Mel-frequency Cepstrum (MFCC) techniques are used to train models for speech recognition. Results have shown great success in terms of word recognition. In some trials, the program was able to recognize the word with 99% accuracy. According to these results, the speech recognition part was successfully completed, and the car constructed in the second part of the project, speech recognition system was implemented to the robotic car, the car executes the simple speech commands recognized by trained speech recognition model.
Autonomous Carry Robot
Advisor: Asst. Prof. Yusuf Aydın
Team: Yiğit Atik, Ziya Oğuzhan Çevik
Keywords: SLAM, robot, autonomous, sensor
Abstract: Today, we live in a world where we need to reduce one-to-one contact and attach importance to social distance in many areas. In addition, the employment of personnel in many industries makes human labor impossible due to both dimensional and spatial hazards and radioactive or toxic components in the service area. Able to avoid all static obstacles and the robot will be the solution to these problems. Arduino-based robots detect effective obstacles with the help of sensors. It is very practical and effective as it does not need to be connected to another device and can be operated for a long time with only a 9V battery. It has a structure that can be easily upgraded to many different areas. For this reason, if you are going to use the user vehicle on flat terrains, it can be used quickly with small tires or you can easily make power and battery upgrades in more difficult areas. In this project, we implemented a mobile robot that has an autonomous function in dynamic and static environments. There are general problems in almost all the robots or gadgets that are autonomous. These problems go as path planning, localization, etc. In this project, implementations and algorithms that solve these problems can be found.
Low-voltage Power Meter
Advisor: Asst. Prof. Arif Mustafazade
Team: Ali Bahadır Bulut, Uğurcan Akdemir
Keywords: Power measurement, voltage measurement, current measurement, instantaneous power
Abstract: Low-voltage power meter is capable of measuring and displaying the instantaneous power consumption of a load from a low-voltage AC or DC source. It displays instantaneous power consumption of a circuit on an LCD screen. It is used determining electrical problems in circuits and verifying functional correctness of circuits.
Solar Powered Straw-man
Advisor: Asst. Prof. Tuba Ayhan
Team: Murat Cafer Demirci, Berkay Sığırcı, Bora Toros
Keywords: Straw-Man, Arduino Nano, Solar Energy, Servo Motor, Solar Panel, Battery
Abstract: Birds are causing a lot of trouble in various industries these days. The agricultural sector is one of the most important, while the aviation sector is another. Birds wreak havoc on farmers' crops, inflicting significant material damage and increased workload. Birds penetrate aircraft engines during takeoff and landing, causing technical parts to be disrupted, which is another important issue in the aviation business. This causes significant material damage and the death of birds. There is a solution to tackle this problem without wasting time or money by preventing the birds from dying. In addition, a specific amount of effort is spent on its maintenance. Many methods are currently being tested to modify the direction of birds near airports. However, issues like bird intrusion into aircraft engines continue to exist. To address these issues, we've built a considerably less expensive, solar powered straw-man device. As a result, we have found definitive answers to these issues. With this idea, farmers' crops will not go to waste. Unnecessary tasks will be eliminated, allowing them to produce more of them. Airport corporations, on the other hand, will be spared from significant financial losses at a far lower cost. Birds will not perish, and no one's life will be endangered. Using Arduino Nano, the design and prototype of a solar powered straw-man system were constructed in this thesis. Servo motors were used to simulate our straw man's arms. Buzzer was also used as sound output device. A remote control system was designed to turn the system on and off. In the system, a battery was used to feed the remote control system, microprocessor and servo motors. The necessary charging unit was added to the system and the charging of the battery with the solar panel was observed.The system has been tested to be stable.
Financial Robo-Advisor
Advisor: Asst. Prof. Serap Kırbız
Team: Eyüp Yasuntimur, Can Doğrul
Keywords: Exchange Rate, Commodity, Artificial Neural Network (ANN), Autoregressive Integrated Moving Average (ARIMA), Investment
Abstract: Different exchange rate prediction methods have received much attention in the past decades. Additionally, using machine learning models to predict investment tools is becoming an increasingly vital factor in finance. A major difficulty of exchange is that the financial instruments of the developing countries are highly volatile. Predicting the price movements of currency accurately is a massive advantage for investment. The aim of our project is to build an application for users with low financial literacy to make investments. In this project, a Financial Robo-Advisor application is being developed that uses machine learning techniques such as Artificial Neural Network (ANN) and Autoregressive Integrated Moving Average (ARIMA) to predict future financial time series data. The application offers users suitable investment portfolios based on these estimates. We believe that the design will be an innovative solution for users with low financial literacy.
IoT Authentication with Visual Cryptology
Advisor: Asst. Prof. Tuba Ayhan
Team: Feyza Türker, Osman Can Güdül
Keywords: IoT, Visual cryptography, RISC-V
Abstract: Data security is getting important day by day. While people using IoT devices, they want their data to be secured. Most of them do not know the protocol of the IoT devices and they do not know where their data went. For securing your data for in IoT devices engineers start to use authentication methods. In this project, Visual cryptography method will be explained for the authentication. Besides that, for cheaper and low powered project RISC-V will be used and explained. Visual cryptography will be explained with its data flow. RISC-V will be explained more like instruction way for repeatability. For the final section results from both hardware and software part will be explained. This report aims to explain both RISC-V and visual cryptography.
Automated Question Answering System for Personalized Online/Distance Learning Experience
Advisor: Asst. Prof. Ebru Arısoy
Team: Erdal Sidal Doğan, Mert Kömürcüoğlu
Keywords: Sentiment Analysis, Deep Learning, Question Answering, Keyword Extraction
Abstract: Online learning has become the inseperable part of many lives. However, there are certain aspects of the online learning that has negative eects on the knowledge acquisition process, lack of interaction between learner and the instructor is a prominent one. Aim of this project is to develop a system that would increase the student participation to the online lectures. Our systems keeps the learner active in various ways, depending on their feedbacks about the topic. This inclusion occurs either by asking a question in order to assess their learning, answering their question or redirecting them to topic they didn't understand well. Initial stage is to infer the students' feedback to determine whether they understand the topic or they have a question. In this step, Sentiment Analysis (SA) is utilized in order to extract the sentiment in the sentence. This project presents the implementation details of preliminary work on an automatic systems that has been developed for the online learning platforms.
--
This project is conducted with Computer Engineering Department.
--
Wireless Health Monitoring System
Advisor: Asst. Prof. Tuba Ayhan, Assoc. Prof Şuayb Arslan
Team: İrem Kuzgunkaya, İbrahim Ata Mercan, Ozan Sönmez
Keywords: Health system, wireless communication, web application, sensors, machine learning, sleep apnea.
Abstract: A portable electronic device using microcontrollers and sensors assisted by wireless transmitters will be used to monitor the patients' health conditions. The system will include sleep apnea detection using a heartbeat sensor. With the help of microcontrollers, the project aims to solve some of the emerging problems in terms of the clinical needs of the aging population via home-based, easy to use wireless devices that can be used for diagnosis and observations. By using these wireless sensors, the overwhelmed healthcare system in underdeveloped and developing countries may relieve their hospitals and health centers as those solutions will drastically decrease the number of visits needed for illness diagnosis and patient monitoring. The system will consist of two different components which are a web platform for monitoring the relevant data retrieved from the device and a portable electronic device. The web platform will be designed to detect and alert the observer in the times of detection of diagnosis such as tachyarrhythmia. This project is believed to bring value to the wearable device environment by using wearable devices for solely health purposes without overwhelming the current health systems in place within the country. The system will require little-to-nonprior training for using web applications as it will be planned to be made very similar to the current systems that are used in the hospitals. This project aims to detect sleep apnea at patients' own houses' comfort without forcing the patient to stay at the hospital for the night.