Theses supervised by Doç. Dr. Orhan Gazi
16 theses · Çankaya University
Detecting subthreshold signals in field effect transistors by using stochastic resonance phenomenon
Until recently it is accepted that noise is always a degrading factor for the performance of electronic communication and signal processing systems. For this reason many of the studies on signal processing and communication systems focused on eliminating the negative effects of noise signal on information bearing signal. However, recently it is discovered that noise can play a positive role for the detection of weak signals for electronic systems employing nonlinear electronic devices. It is seen that the detection of the weak signals in nonlinear systems is possible when an optimum amount of noise is added to the information bearing signal before passing the signal through the nonlinear system. And this concept is named as stochastic resonance, i.e., SR. In this thesis it is shown that it is possible to develop a training based model for the SR systems that decides optimum noise signal to be added to the weak information signal directly for the best system performance. In the second part of the thesis study a practical application of the SR concept on a nonlinear electronic device, field effect transistor (FET), is demonstrated. For this purpose a user interface program using C# is developed and some part of the signal processing is done using the field programmable gate arrays (FPGA).
Hiperspektral görüntüde boyut indirgeme yöntemleri
Hyperspectral Images has huge dimensions of data compared to single band or multispectral band images. This results from the fact that it contains hundreds of spectral bands with a high spectral resolution. Therefore, hyperspectral data processing, storing, and transmitting are critical issues to deal with. Additionally, it is a fact that required sample size for training a specific classification method increases exponentially with increasing number of spectral bands. In order to handle these problems, either the training data size has to be enlarged or dimensionality of hyperspectral images has to be reduced with some dimension reduction techniques. In this thesis, supervised and unsupervised dimension reduction methods are investigated, and some new methods are proposed. The proposed methods aim to reduce the dimensionality of the hyperspectral data before classification while preserving the classification accuracy as much as possible and to achieve reduced dimension with a low computational complexity.
Polar coded communication system design
In 2008, polar codes, which are the only channel codes whose performance are mathematically proven, are introduced by Arıkan. The introduction of polar codes was a milestone in coding society. The implementation of polar codes at high speeds is of critical importance, especially in communication systems such as 5G where high speed and parallel processing ICs such as FPGA are employed. In FPGA technology, it is of critical importance how much space the algorithm occupies in the integrated chip. The area occupied by the algorithm is directly proportional to the energy use of the FPGA and the delay of the process. In this thesis, we propose a polar encoding method that occupies less hardware implementation space. It is seen that with the innovative polar encoding scheme, considering the implementation area used, a noticeable advantage for FPGA implementations is achieved. We also analyse the performance of polar codes with pilot-based channel estimation and equalization methods. By using the frozen bits of the polar codes, the pulse and frequency responses of the communication channels can be estimated. Among the estimation and equalizer methods, least squares and minimum mean square (MMSE) methods are used. The employment of frozen bits provides an advantage in increasing both the processing speed and the data rate in the receiver at the same time. This is because there is no need for different packet structures and operations in the communication protocol for channel estimation and equalization. Keywords: Pole codes, error correction, channel estimation, channel matching, FPGA, code optimization, successive cancelation, VHDL implementation
The use of convolutional product codes in cooperative communication systems
In this thesis work, the performances of cooperative communication systems employing convolutional product codes (CPCs) and serially concatenated convolution codes (SCCCs) are inspected in detail and compared to each other. Different combining methods at destination are proposed. One of these methods is based on the combination of bit probabilities on the other hand other method focuses combining the signals considering their signal to noise rations (SNRs). The effect of relay number on the system performance is inspected. Reduced complexity cooperative communication systems utilizing convolutional product codes are considered. The proposed systems are all iteratively parallel decodable and have low latency.
Joint structures involving pdsccs, tcm and space time codes
In this thesis we propose new concatenated joint communication structures involving convolutional product codes, space time codes and trellis coded modulation. The first structure consists of convolutional product codes and space time codes. The second one includes convolutional product codes and trellis coded modulation. The proposed structures are all parallel decodable due to the structure of convolutional product codes and have low decoding latencies, which are the main advantages of these proposed structures considering their classical counterparts constructed using serially concatenated convolutional codes, space time codes and trellis coded modulation. The use of multi-antennas at the proposed structures increases the spectral efficiency.
PAPR reduction of OFDM system symbole via optimal rotation of information symbols
Orthogonal frequency division multiplexing (OFDM) is an effective multicarrier transmission for wireless communication systems. The demand for high data rate for multimedia applications made OFDM widely used in wireless communication. The main drawback of OFDM communication systems is their high peak-to-average power ratios (PAPRs) which limit their use in practical applications. There are two well-known PAPR reduction techniques known in the literature, partial transmit sequence (PTS), and selective mapping which are closely related to each other. In this thesis work, we propose a new PAPR reduction approach. Our method is based on rotation of the information symbols before inverse fast Fourier transform (IFFT) operation. The selection of the information symbols is performed inspecting their combinations in OFDM symbols, and this is achieved detailing decimation in frequency IFFT algorithm, i.e., inspecting the combination of information symbols while performing decimation in frequency IFFT algorithm. The proposed reduction method has better performance than that of the PTS method, and has much less complexity when compared to that of the PTS technique. Keywords : OFDM, Peak-to-average power ratio (PAPR), Complexity, Distortion, IFFT, Partial Transmit Sequences, Decimation in Frequency Fast Fourier Transform, BER.
Design of high performance low latency rateless codes
Luby Transform (LT) codes are one of the best rateless codes mainly designed for binary erasure channel. The characteristics of such codes perfectly performing when used with bulk data files, however a performance degradation has been observed when using them with short length messages. In this thesis, we present a new design for rateless codes, particularly an efficient LT codes using robust soliton distribution (RSD) as a degree generation method and tested in both binary erasure channel (BEC) and noisy channels like the additive white Gaussian noise (AWGN) channel. First, a new proposed decoding technique is defined as belief propagation-pattern recognition (BP-PR) is implemented to enhance the decoding ability of the conventional (BP) algorithm to overcome the problem of losing degree-one coded symbols which caused early decoding termination. The simulation results approve the improvement of the BP-PR when used with LT-RSD and outperforms the bit error rate (BER) records for the state of art techniques like memory-based robust Soliton distribution using conventional BP (LT-MBRSD-BP) or the Gaussian elimination assisted belief propagation (LT-RSD-BP-GE) and improve the records for the BER when used with MBRSD, ISD and optimal degree distribution (ODD), to form the new code called (LT-MBRSD-BP-PR),(LT-ISD-BP-PR) and (LT-ODD-BP-PR) respectively. Second, a new efficient deterministic encoding technique using deterministic degree generator with random data selection (LT-DE) is applied for extremely short data lengths. The degree generation method is based on creating the degrees in a repeated frame with a limited upper value called repetition period (𝑅𝑝) and the data symbols are chosen sequentially from a truncated data file. The data file is truncated to segments of length (𝑅𝑝) and each segment is chosen based on a random sequence. Testing this (LT-DE) against (LT-RSD-BP-PR), (LT-MBRSD-BP-PR) and (LT-ODD-BP-PR) in a BEC environment had approved the superiority of such code over all the other mentioned techniques. It has the lower error floor and higher successful decoding rate with minimum overhead and computational cost. The formation of this (LT-DE) associates a mutual relation between the successive coded symbols which motivate us to present a new sequential decoding technique mainly used over (AWGN) channel. With such new encoding-decoding technique LT codes can approach the decoding complexity cost of Raptor codes with smaller overhead and less encoding complexity as well.
Efficient decoding of polar codes
Polar Codes are the first mathematically provable capacity achieving error correcting codes which have low complexity encoding and decoding algorithms. For the decoding of polar codes, as a preliminary decoding algorithm, the successive cancellation (SC) decoding algorithm is used. SC algorithm is a sequential decoding algorithm which suffers from error propagation. For this reason, SC algorithm does not show good performance for moderate codeword lengths. Polar codes with SC decoding show worse performance than that of the modern channel codes, such as LDPC and turbo codes. To improve the performances of the polar codes improved versions of SC algorithm such as SC list (SCL) and SC stack are introduced in the literature, and these algorithms show much better performance than that of the classical SC decoding algorithm although they have larger complexity compared to SC. Besides, cyclic redundancy check codes are concatenated with polar codes which are decoded using the SCL algorithm, and such a concatenated system shows better performance than the other modern channel codes. In this thesis, we first propose a tree structure for the successive cancelation (SC) decoding of polar codes. The proposed structure is easy to implement in hardware and suitable for parallel processing operations. Next, using the proposed tree structure, we propose a technique for the fast decoding of polar codes. With the proposed method, it is possible to decode all the information bits simultaneously at the same time, i.e., in parallel. Lastly, we introduce and improved version of the proposed high-speed decoding algorithm. The proposed high-speed decoding approach and its improved version are simulated on computer environment, and their BER performances are compared to the performance of the classical successive cancelation method. Furthermore, we introduce a new approach to the successive cancelation of polar codes. The proposed approach uses the soft likelihood ratios of the predecessor information bits for the determination of successor information bits. The proposed method can be considered for the construction of joint iterative communication systems exchanging soft likelihoods. It is shown that the proposed soft decoding approach shows better performance than the classical successive cancelation algorithm introduced in Arikan's original work. As we know, polar codes are decoded in a sequential manner using successive cancelation algorithm introduced by Arikan. The sequential nature of the decoding process suffers from error propagation. We inspect the effects of error propagation on the performance of polar codes and propose some methods to alleviate the degrading effects of error propagation on the code performance for short and long frame lengths.
Pre-distortion design for non-linear power amplifiers
Power amplifiers are vital part of communication systems for long distance communication. They are used to amplify the power of incoming signals. The power gain characteristics of these amplifiers show a non-linear behavior for larger input power signals. The non-linear power gain response decreases the efficiency of power amplifier. To alleviate the negative effects of non-linear gain responses of the power amplifiers, researchers model the gain responses and calculate or approximate their inverse called pre-distortions, and the incoming signal is passed through the pre-distortion unit before it is fed to the power amplifier such that the gain responses of the power amplifiers show a linear characteristic. In this thesis, we first review the well-known pre-distortion methods available in the literature, and then propose a novel approach for pre-distortion design. The proposed approach uses the concept of analog signal reconstruction from its samples. We use sinc(⋅) function for the modeling of amplifier gain response and pre-distortion design. The proposed method can be seen as the optimum implementation of indirect learning algorithm used for pre-distortion design.
Improved successive cancellation decoding of polar codes
In this thesis, we propose improved successive cancellation polar code decoding algorithms. Polar codes are fragile to error propagation. Considering this issue, in our first proposal, we consider 𝐿𝑅=1, for which decision is made for the favor of bit 0 in classical successive cancellation algorithm, and propose multi-SC decoders, where we consider more than one decoders working in parallel and these decoders make opposite decisions for 𝐿𝑅=1. The proposed technique provides a flexible configuration and leads to the pruning of unnecessary path searching operations, which provide low complexity compared to successive cancelation list decoding algorithm. Multi-Parallel SC decoding shows a significant performance improvement compared with the original SC decoding and its performance is comparable to that of the successive cancellation list decoding algorithm. In our next proposal, we propose a method for the iterative decoding of polar codes replacing the unreliable received samples with randomly generated samples. In this method, first, a straight decoding operation is performed for the received frame and CRC check is performed, and if it is not satisfied, the received symbols are passed through virtual random channels before they are sent to the polar decoders employing successive cancellation decoding algorithm. When the received symbols are passed through virtual random channels, randomly generated noise is added to the inputs of the VRCs falling into a threshold interval, which contains unreliable information about the transmitted polar code-bit before they are sent to the polar decoder. For the decoded sequence, if the CRC check is not satisfied, a different randomly generated noise sequence is added to the unreliable inputs and the decoding operation is repeated. This procedure is repeated until a predefined maximum iteration number as long as CRC is not satisfied.
Spatial methods for direction of arrival estimation and hardware implementation
In this thesis, spatial methods for direction of arrival estimation and it hardware implementation are discussed. Estimation of the direction of arrival (DOA) of signals are widely used in different fields such as radar, sonar, acoustics, astronomy, and communications technologies. Spatial spectrum focuses on researching the spatial characteristics of the signal and the direction of the source. It displays signal propagation across all directions to the receiver. Therefore, if the spatial range of the signal is detected, then DOA can be found. This technology is very essential in signal processing, and it has expanded rapidly in recent years in particular finding the DOA of multiple signal sources. In this context, many algorithms have been used and have made great accomplishments over the last few years. In this thesis, firstly, the Capon beamforming and conventional beamforming based on ULA arrangement are simulated in MATLAB. The effects of the number of sensors, the distance between sensors, the number of samples, and the effect of SNR value were examined. The high number of sensors, high SNR values and the high number of samples increased the resolution and accuracy, while the distance between the sensors was chosen more than half of the wavelength, which led to the prediction of false angles. So it is observed that it would be best to keep the spacing at half of the wavelength. After MATLAB simulations, conventional beamforming is implemented in VHDL which is a hardware description language. For the VHDL implementation, signed fixed point numbers are used. The DOA estimation is implemented in VHDL for ULA. According to the simulation results, the VHDL algorithm achieved the angle values with a small margin of error.
Calculation of trigonometric functions using cordic algorithm
CORDIC which is the abbreviation of coordinate rotation digital computer is an algorithm proposed in 1959 by Jack. E. Volder. Since its introduction, numerous studies are performed for improved versions of the CORDIC algorithm. CORDIC algorithm is initially introduced for the computation of trigonometric functions, multiplication and division operations. Later on, this algorithm is further developed for the calculation of other elementary transcendental functions such as logarithms, exponentials, square roots. CORDIC algorithm is used in many diverse areas such as robotics, signal processing, graphics and animation, digital communication, image processing. CORDIC algorithm is developed for the hardware implementation of mathematical functions, and it is shown by the researchers that CORDIC algorithm is a good choice for scientific calculators. The cost and size of the hardware equipment needed for the implementation of a mathematical function depends on the computation complexity of the algorithm under concern. In time, CORDIC algorithms with higher precision and faster convergence rates are proposed in literature. In this thesis work we study radix-2, radix-4, angle recoding, and extended angle recoding CORDIC techniques and compare the algorithms considering the number of iterations required for a defined precision. Algorithms are simulated via computer programs. The results show that the radix-2 has requires more number of iterations compared to radix-4, angle recoding and extend angle recoding methods.
Performance of space time block coded bit interleaved coded modulation systems for mobile communication channels
In this thesis concatenated systems involving bit interleaved coded modulation (BICM) and space time codes (STC) are inspected in detail. Space time block codes (STBCs) are used while forming the joint structures. Performance of the joint structures involving BICM and STCs for additive white Gaussian and mobile fading channels are measured via computer simulations. Iterative decoding logic is applied for the decoding of these joint structures. An alternative feedback path for the iterative decoding of BICM-STBC joint structures is proposed and simulation results showed that the proposed path result in better bit error rate performance. Concatenated BICM-STBCs are also used in cooperative communication structures. A signal combination method based on the linear combination of the signal probabilities coming from the relays is proposed and it is seen that proper choosing of the coefficients for the linear combination of the probabilities is an important criteria for the performance of the cooperated systems.
Effects of channel estimation on turbo equalizer performance
Equalizer design is an important concept for the wireless communication systems suffering from multipath effects such as ISI. Turbo equalizers are one of the most powerful equalizers introduced several decades ago. For the equalizers to be efficient in removing ISI reliable channel estimation is essential. In this thesis the performance of turbo equalizers for blind and non-blind channel estimators is inspected. For blind channel estimation per-survivor processing with least mean square method is employed. For non-blind channel estimation training sequences are used for channel estimation. An improved channel estimation approach for non-blind technique was suggested and via simulation results, it was seen that the proposed method shows better performance in turbo equalizer than the classical blind and non-blind channel estimation methods do.
Effects of degree distribution in rateless coding
In this thesis rateless codes which are adopted by a variety of applications such as, wireless transmission, 3GPP, data storage, multicasting, video streaming are inspected in details. The performance of two important types of rateless codes which are Luby Transform and Raptor codes are measured via computer simulations. Both hard decision and soft decision methods are used while measuring the performance of these codes. For the soft decision decoding Belief Propagation algorithm was used in an iterative manner. Degree distribution is an important criteria for the performance of Luby Transform codes. A new degree distribution called random degree (or exponential random) distribution is proposed for Luby Transform codes. And simulation results support that the proposed distribution shows better performance than the classical degree distributions such as all-at-once, ideal soliton, robust soliton, and sparse.
Kanal tahmimi kullanarak kazanan patika
In this thesis per-survivor processing has been used for estimating the coefficients of a frequency selective channel. Two well-known methods, least mean square and recursive least mean square, along with per-survivor processing has been used during channel estimation. To increase the accuracy of the estimated channel coefficients an improved per-survivor processing channel estimation technique has been proposed and from the simulation results it is seen that the proposed technique results in better estimated channel coefficients than the channel coefficients obtained using classical per-survivor approach.