Researcher: Köprü, Berkay
Name Variants
Köprü, Berkay
Email Address
Birth Date
5 results
Search Results
Now showing 1 - 5 of 5
Publication Metadata only Affective burst detection from speech using Kernel-fusion dilated convolutional neural networks(IEEE, 2022) N/A; N/A; Department of Computer Engineering; Köprü, Berkay; Erzin, Engin; N/A; Faculty Member; Department of Computer Engineering; N/A; College of Engineering; N/A; 34503As speech interfaces are getting richer and widespread, speech emotion recognition promises more attractive applications. In the continuous emotion recognition (CER) problem, tracking changes across affective states is an essential and desired capability. Although CER studies widely use correlation metrics in evaluations, these metrics do not always capture all the high-intensity changes in the affective domain. In this paper, we define a novel affective burst detection problem to capture high-intensity changes of the affective attributes accurately. We formulate a two-class classification approach to isolate affective burst regions over the affective state contour for this problem. The proposed classifier is a kernel-fusion dilated convolutional neural network (KFDCNN) architecture driven by speech spectral features to segment the affective attribute contour into idle and burst sections. Experimental evaluations are performed on the RECOLA and CreativeIT datasets. The proposed KFDCNN outperforms baseline feedforward neural networks on both datasets.Publication Metadata only Multiplicity estimating random access protocol for resource efficiency in contention based NOMA(IEEE, 2018) Guersu, H. Murat; Kellerer, Wolfgang; N/A; Department of Electrical and Electronics Engineering; Köprü, Berkay; Ergen, Sinem Çöleri; PhD Student; Faculty Member; Department of Electrical and Electronics Engineering; Graduate School of Sciences and Engineering; College of Engineering; N/A; 7211Emerging technologies enforce strict requirements on future wireless networks such as massive connectivity that cannot be supported with scheduled access. Contention based Non-Orthogonal Multiple Access is a novel technique to overcome strict massive connectivity requirements by efficient use of wireless resources. However, most of the solutions proposed in this direction assumes different loads which would degrade the performance significantly if they would not hold. To stress these assumptions a resource efficiency metric is defined and state of the art solutions are evaluated for varying load regarding this metric. It is shown that the resource efficiency problem in the state of the art can be improved with multiplicity estimation, and hence, we propose Multiplicity estimating Random Access protocol, that adapts to the dynamic loads. This adaptation is evaluated through analytical calculation against the state of the art and it is shown that resource efficiency against with a slight decrease in the metric any load from 1 up to > 10(3) users is supported. In addition, we show how this protocol can be dimensioned and integrated to contention based NOMA.Publication Metadata only Neural network based sleep phases classification for resource constraint environments(Ieee, 2021) Aslan, Murat; Kholmatov, Alisher; N/A; Köprü, Berkay; PhD Student; Graduate School of Sciences and Engineering; N/ASleep is restoration process of the body. The efficiency of this restoration process is directly correlated to the amount of time spent at each sleep phase. Hence, automatic tracking of sleep via wearable devices has attracted both the researchers and industry. Current state-of-the-art sleep tracking solutions are memory and processing greedy and they require cloud or mobile phone connectivity. We propose a memory efficient sleep tracking architecture which can work in the embedded environment without needing any cloud or mobile phone connection. In this study, a novel architecture is proposed that consists of a feature extraction and Artificial Neural Networks based stacking classifier. Besides, we discussed how to tackle with sequential nature of the sleep staging for the memory constraint environments through the proposed framework. To verify the system, a dataset is collected from 24 different subjects for 31 nights with a wrist worn device having 3-axis accelerometer (ACC) and photoplethysmogram (PPG) sensors. Over the collected dataset, the proposed classification architecture achieves 20% and 14% better F1 scores than its competitors. Apart from the superior performance, proposed architecture is a promising solution for resource constraint embedded systems by allocating only 4.2 kilobytes of memory (RAM).Publication Metadata only Deep learning based minimum length scheduling for half duplex wireless powered communication networks(Institute of Electrical and Electronics Engineers (IEEE), 2022) N/A; N/A; N/A; Department of Electrical and Electronics Engineering; Önalan, Aysun Gurur; Köprü, Berkay; Ergen, Sinem Çöleri; PhD Student; PhD Student; Faculty Member; Department of Electrical and Electronics Engineering; Graduate School of Sciences and Engineering; Graduate School of Sciences and Engineering; College of Engineering; N/A; N/A; 7211Minimum length scheduling is used to ensure the strict delay requirements of time-critical applications in wireless powered communications networks (WPCNs). The previous optimal and sub-optimal solutions of the problem suffer from the run-time complexity of the iterative algorithms, which makes real-time applications unpractical. This paper proposes a deep learning based framework for a low-complexity solution to the minimum length scheduling problem in half-duplex WPCNs. The objective of the problem is to minimize the duration of the schedule for energy harvesting (EH) and information transmission (IT), subject to the data demand, energy causality, and maximum transmit power constraints. Multi-input multi-output feed-forward deep neural network (DNN) architecture is considered, where the inputs are channel state information and two parameters derived from the optimality conditions of the problem; and outputs are the transmit powers, EH and IT lengths. To ensure the feasibility of the DNN outputs, we design a final layer which maps the estimated transmit powers to the feasible EH and IT lengths. The DNN is trained offline with both supervised and unsupervised techniques. Simulation results indicate that the proposed DNN-based approaches are up to 8.5 times faster than the benchmark iterative algorithms. These approaches also outperform benchmark sub-optimal algorithms in terms of accuracy with only 0.12% optimality gap and robustness against varying network conditions.Publication Open Access Multiplicity estimating random access protocol for resource efficiency in contention based NOMA(Institute of Electrical and Electronics Engineers (IEEE), 2018) Gürsu, H. Murat; Kellerer, Wolfgang; Department of Electrical and Electronics Engineering; Ergen, Sinem Çöleri; Köprü, Berkay; Faculty Member; Department of Electrical and Electronics Engineering; Graduate School of Sciences and Engineering; 7211; N/AEmerging technologies enforce strict requirements on future wireless networks such as massive connectivity that cannot be supported with scheduled access. Contention based Non-Orthogonal Multiple Access is a novel technique to overcome strict massive connectivity requirements by efficient use of wireless resources. However, most of the solutions proposed in this direction assumes different loads which would degrade the performance significantly if they would not hold. To stress these assumptions a resource efficiency metric is defined and state of the art solutions are evaluated for varying load regarding this metric. It is shown that the resource efficiency problem in the state of the art can be improved with multiplicity estimation, and hence, we propose Multiplicity estimating Random Access protocol, that adapts to the dynamic loads. This adaptation is evaluated through analytical calculation against the state of the art and it is shown that resource efficiency against with a slight decrease in the metric any load from 1 up to > 10(3) users is supported. In addition, we show how this protocol can be dimensioned and integrated to contention based NOMA.