Mehmet Yakut

Neural network training based on FPGA with floating point number format and it’s performance

In this paper, two-layered feed forward artificial neural network’s (ANN) training by back propagation and its implementation on FPGA (field programmable gate array) using floating point number format with different bit lengths are remarked based on EX-OR problem. In the study, being suitable with the parallel data-processing specification on ANN’s nature, it is especially ensured to realize ANN training operations parallel over FPGA. On the training, Virtex2vp30 chip of Xilinx FPGA family is used. The network created on FPGA is coded by using VHDL. By comparing the results to available literature, the technique developed here proved to consume less space for the subjected ANN training which has the same structure and bit length, it is shown to have better performance.

Identification of IM Resistance Using Artificial Neural Network in Low Speed Region

This paper presents a new method of estimation for the stator and rotor resistances of the induction motor for speed sensorless motor control drives, using artificial neural networks. The error between the motor quantity based on a neural network model and a voltage model is back propagated to adjust the weights of the neural network model for the stator resistance estimation. For the rotor resistance estimation, the error between the measured stator current and the estimated stator resistance using neural network is back propagated to adjust the weights of the neural network. The rotor speed is extracted from the induction motor state equations. The performance of the stator and rotor resistance estimators are investigated with the help of measured the stator voltage and current. Both resistances are estimated experimentally, using the proposed neural network in an induction motor drive.

Design and implementation of a communication protocol for mobile device controlled smart home management system

Recently, with the increase in mobile and internet accessibility, it has become widespread that smart home systems can be managed remotely and monitored by mobile devices. In this study, a communication protocol between a mobil devices-embedded PC based control unit-sensors/actuators is designed and implemented. A handshaking mechanism between a mobile device and control unit, and packet structure for messaging between control unit and sensors/actuators are explained. With this system, the smart home system can be controlled over a wireless LAN by authorized users using an Android based mobile device on which the implemented GUI software is running.

Fast video vector quantization

In this paper fast video vector quantization algorithm was developed utilizing high correlation between successive frames which can be used for motion compensation as a next step. Best matching code vector search speed is increased using previously encoded frame code vector index list, which carry the available information of that frame. By using interframe correlation, search speeds of tree structured vector quantization (TSVQ) was achieved Additionally the coding structure developed here is directly applicable to motion estimation (ME) and motion compensation. (MC) The main objective of this work is real time video encoding using high calculation power of server computer and at clients side requiring very low computation power, meanwhile requiring low electric energy. The algorithm developed here is very suitable especially for mobile equipments which can cary only limited electric power on it.

Determination of Product Surface Quality

Physical changes in the product surface can be determined by using digital image processing and laser scanning techniques. In this study, how to determine surface quality of the product by using both two techniques is investigated. Car bonnet is choosen as the product. Images have been taken while its surface is scanning and, faulty regions on its surface is determined by using two dimensional digital image processing methods. And finally, area of this faulty regions are calculated by using Gausse method

Tracking control based on neural network for robot manipulator

In this paper, a control algorithm based on neural networks is presented. This control algorithm has been applied to a robot arm which has a highly nonlinear structure. The model based approaches for robot control require high computational time and can result in a poor control performance, if the specific model structure selected does not properly reflect all the dynamics. The control technique proposed here has provided satisfactory results. A decentralized model has been assumed here where a controller is associated with each joint and a separate neural network is used to adjust the parameters of each controller. Neural networks have been used to adjust the parameters of the controllers, being the outputs of the neural networks, the control parameters.