Yazar "Pakyurek, Muhammet" seçeneğine göre listele

Listeleniyor 1 - 3 / 3
  • Küçük Resim Yok
    Öğe
    A Dynamic Optimization based Algorithm for Pilot Assignment in Massive MIMO
    (Inst Information Science, 2023) Pakyurek, Muhammet; Kulac, Selman
    In this study, we propose an efficient algorithm for the optimization of pilot assignment to reduce pilot contamination in multi-cell multi-user massive multiple-input multiple-output (MIMO) systems. The object function of the proposed optimization approach is based on the cumulative average uplink Signal to Interference plus Noise Ratio (SINR) values. This method effectively optimizes pilot-user matchings using an efficient search which wisely reduces both search spaces of users and pilots. It dynamically calculates search space reductions after updating each cell iteratively. This reduction is performed by selecting only the users whose SINR value is below the average SINR of the each cell. On the other hand, the pilot's search space is reduced by choosing only the least-used pilots within close neighboring cells' pilot usages. This calculation is performed for each cell to increase the cumulative average SINR (global SINR) value. Pilot contamination is reduced to minimum levels with less complexity and memory requirements with the proposed method in this manner. The simulation results indicate that the proposed method outperforms the recent popular analytical and deep learning-based pilot assignment approaches. The achieved improvement is even more prominent when all the pilots are fully used.
  • Küçük Resim
    Öğe
    Extraction of Novel Features Based on Histograms of MFCCs Used in Emotion Classification from Generated Original Speech Dataset
    (Kaunas Univ Technology, 2020) Pakyurek, Muhammet; Atmis, Mahir; Kulac, Selman; Uludag, Umut
    This paper introduces two significant contributions: one is a new feature based on histograms of MFCC (Mel-Frequency Cepstral Coefficients) extracted from the audio files that can be used in emotion classification from speech signals, and the other - our new multi-lingual and multi-personal speech database, which has three emotions. In this study, Berlin Database (BD) (in German) and our custom PAU database (in English) created from YouTube videos and popular TV shows are employed to train and evaluate the test results. Experimental results show that our proposed features lead to better classification of results than the current state-of-the-art approaches with Support Vector Machine (SVM) from the literature. Thanks to our novel feature, this study can outperform a number of MFCC features and SVM classifier based studies, including recent researches. Due to the lack of our novel feature based approaches, one of the most common MFCC and SVM framework is implemented and one of the most common database Berlin DB is used to compare our novel approach with these kind of approaches.
  • Küçük Resim
    Öğe
    Web Page Information Extraction System by Using Deep Learning
    (Ieee, 2019) Pakyurek, Muhammet; Sezgin, Mehmet Selman; Kulac, Selman
    In many companies, business units that aim to online sell, need every type of referential data about the market. In order to collect this data which can be group of price, content, survey etc. with a predefined format, websites which sell similar products can be used. The methods used in the data collection process are generally categorized by 3 main groups: 1 Manual 2-Half Manual 3-Auto. Statically coded data collectors (type 1 and type 2) are unable to collect healthy data in the long term and require continuous development and maintenance effort, as Internet pages are dynamic and changes would happen frequently in their page designs. In this study, a data scraping application (type 3) which is not affected by structural changes in web pages was developed. This study aims to obtain data from images of web pages using Deep CNNs.