1,723,380 research outputs found
Eliminating blocking artifacts in halftoning-based block truncation coding
No full text24th European Signal Processing Conference, EUSIPCO 2016, Hungary, 28-2 August 2016Block Truncation Coding (BTC) is an effective lossy image coding technique that enjoys both high efficiency and low complexity especially when halftoning techniques are employed to shape the noise spectrum of its output. However, due to its block-based nature, blocking artifacts are commonly found in the coding outputs. In this work, a real-time halftoningbased BTC algorithm is proposed to solve this problem by eliminating the cause of blocking artifacts while maintaining a complexity comparable to the best stateofthe-art halftoning-based BTC algorithm. Both objective and subjective comparisons demonstrate the visual quality improvement in its encoding outputs.Department of Electronic and Information Engineerin
Efficient and Accurate Neural Fingerprints Obtained via Mean Curve Length of High Dimensional Model Representation of EEG Signals
Full text link31st European Signal Processing Conference, EUSIPCO 2023 -- 4 September 2023 through 8 September 2023 -- -- 194070In this study, we propose and evaluate a feature extraction methodology for the purpose of EEG-based person recognition. To this end, the mean curve length (MCL) was employed subsequent to the representation of EEG signals in an orthogonal geometry through High Dimensional Model Representation (HDMR). To analyze the effectiveness of the methodology, we executed it on a standard publicly available EEG dataset containing 109 subjects and acquired from 64 channels for eyes-open (EO) and eyes-closed (EC) resting-state conditions. The proposed feature was evaluated by comparing it to MCL, beta, and gamma band activities. According to the performance results, applying MCL to the output of the HDMR instead of raw data provides superior performances for identification and authentication. The attained results promise a novel simple, fast, and accurate biometric recognition scheme, named HDMRMCL. © 2023 European Signal Processing Conference, EUSIPCO. All rights reserved
Speakers counting by proposed nested microphone array in combination with limited space SRP
No full text© 2021 European Signal Processing Conference. All rights reserved.In this paper, a novel method is presented for estimating the number of speakers based on the microphone arrays. Firstly, a 3D snowflake nested microphone array (SNMA) is proposed for recording the speech signals. In the following, the steered response power (SRP) algorithm is implemented on subbands in limited spaces conditions for all microphone pairs related to the subarrays. Therefore, a weighted averaging method is implemented on subband limited spaces SRPs (LSRP), and the final energy map is compared with the histogram of the maximums of the SRP function on different subbands for various time frames. The passed candidate points are categorized by unsupervised K-means clustering and the number of speakers is estimated by the silhouette criteria. The accuracy of the proposed method is compared with PENS, i-vector PLDA, and wavelet-GEVD algorithms. The results show the superiority of the proposed method in comparison with other previous research.ANIDFONDECY
Detection of Attention Deficit Hyperactivity Disorder by Using Eeg Feature Maps and Deep Learning
No full text31st European Signal Processing Conference, EUSIPCO 2023 -- 4 September 2023 through 8 September 2023 -- 194070Attention deficit hyperactivity disorder (ADHD) is a mental disorder that affects the behavior of the persons, and usually onsets in childhood. ADHD generally causes impulsivity, hyperactivity, and inattention which impairs day-to-day life even in the adulthood if left undiagnosed and untreated. Although various guidelines for diagnosis of ADHD exist, a universally accepted objective diagnostic procedure is not established. Since current diagnosis of ADHD heavily relies on the expertise of healthcare providers, an EEG Topographic Feature Map (EEG-FM) based method is proposed in this study which aims to objectively diagnose ADHD. 6 different features extracted from EEG recordings acquired from 33 participants, 15 ADHD patients and 18 control subjects, converted into EEG-FM images and fed into a convolutional neural network (CNN) based classifier. Results indicate that the proposed method can accurately classify ADHD patients with up to 99% accuracy, precision, and recall. © 2023 European Signal Processing Conference, EUSIPCO. All rights reserved.2022-07*This study was partially supported by Izmir University of Economics, Scientific Research Projects Coordination Unit. Project number: 2022-07
A Dynamic Mode Decomposition Based Approach for Epileptic Eeg Classification
No full text28th European Signal Processing Conference, EUSIPCO 2020 -- 24 August 2020 through 28 August 2020 -- 165944Epilepsy is a neurological disorder that affects many people all around the world, and its early detection is a topic of research widely studied in signal processing community. In this paper, a new technique that was introduced to solve problems of fluid dynamics called Dynamic Mode Decomposition (DMD), is used to classify seizure and non-seizure epileptic EEG signals. The DMD decomposes a given signal into the intrinsic oscillations called modes which are used to define a DMD spectrum. In the proposed approach, the DMD spectrum is obtained by applying either multi-channel or single-channel based DMD technique. Then, subband and total power features extracted from the DMD spectrum and various classifiers are utilized to classify seizure and non-seizure epileptic EEG segments. Outstanding classification results are achieved by both the single-channel based (96.7%), and the multi-channel based (96%) DMD approaches. © 2021 European Signal Processing Conference, EUSIPCO. All rights reserved.2017-ÖNAP-MÜMF-0002, 2019-TDR-FEBE-0005This study was supported by Izmir Katip Celebi University Scientific Research Projects Coordination Unit. Project numbers: 2019-TDR-FEBE-0005 and 2017-ÖNAP-MÜMF-0002
Combinations of Eeg Topographic Feature Maps for the Classification of Adhd
No full text31st European Signal Processing Conference, EUSIPCO 2023 -- 4 September 2023 through 8 September 2023 -- 194070Attention-Deficit/Hyperactivity Disorder (ADHD) is a common mental disorder affecting both children and adults. It is characterized by issues with concentration, hyperactivity, and impulsivity, which can interfere with everyday duties and interpersonal relationships. Although behavioral studies are utilized to treat the disease, there is no proven method for detecting it. The Electroencephalogram (EEG) is a non-invasive method that monitors electrical activity in the brain and is commonly used to identify neurological and mental illnesses such as ADHD. In this study, the topographic EEG feature maps (EEG-FMs) were obtained from 6 traditional time-domain characteristics known as Hjorth activity, Hjorth mobility, Hjorth complexity, kurtosis, and skewness. The feature maps were concatenated and used as input to Convolutional Neural Network (CNN) model for ADHD classification. To show the efficacy of the recommended approach, EEG data from 15 ADHD individuals and 18 control subjects (CS) were analyzed. The results showed that concatenated EEG-FMs were successful to classify ADHD with up to 99.72% accuracy. © 2023 European Signal Processing Conference, EUSIPCO. All rights reserved.2022-07*This study was partially supported by Izmir University of Economics, Scientific Research Projects Coordination Unit. Project number: 2022-07
Three-dimensional sound source localization by distributed microphone arrays
No full text© 2021 European Signal Processing Conference. All rights reserved.Multiple sound source localization (SSL) is one of the applicable and important areas in the speech signal processing. In this paper, a two-step method is proposed for multiple 3D SSL based on the time delay estimation (TDE) in combination with distributed microphone arrays (DMA). In the first step, the direction of speakers are estimated by the use of a circular microphone array (CMA) in the center of the room and implementing the generalized cross-correlation (GCC) function. In the second step, the distributed T-shaped microphone arrays on the walls are considered for 3D SSL. The two most closed T-shaped array to each speaker are selected, where one of them is used for horizontal and the other one for vertical direction of arrival (DOA) estimation by the use of generalized eigenvalue decomposition (GEVD) algorithm. The experiments on the simulated data for 2 and 3 simultaneous speakers show the superiority of the proposed distributed microphone array-direction of arrival estimators (DMA-DOAE) method in comparison with other previous works in noisy and reverberant environments.ANIDFONDECY
Bispectral analysis of heart rate variability signal
No full text11th European Signal Processing Conference, EUSIPCO 2002 --3 September 2002 through 6 September 2002 -- --This article explores the potential of third-order statistics to analysis of heart rate variability (HRV) signal. Bispectral analysis of short-term HRV signals, obtained from a group of of healthy subjects under various experimental settings, showed the HRV activity to be located on specific bifrequency regions of the magnitude bispectrum. Nine strength measures were defined and were found to respond selectively to induced perturbation of sympathetic-vagal control of heart rate. In general the measures contained the spectral information and provided complementary information. © 2002 EUSIPCO
Curriculum learning for face recognition
No full textWe present a novel curriculum learning (CL) algorithm for face recognition using convolutional neural networks. Curriculum learning is inspired by the fact that humans learn better, when the presented information is organized in a way that covers the easy concepts first, followed by more complex ones. It has been shown in the literature that that CL is also beneficial for machine learning tasks by enabling convergence to a better local minimum. In the proposed CL algorithm for face recognition, we divide the training set of face images into subsets of increasing difficulty based on the head pose angle obtained from the absolute sum of yaw, pitch and roll angles. These subsets are introduced to the deep CNN in order of increasing difficulty. Experimental results on the large-scale CASIA-WebFace-Sub dataset show that the increase in face recognition accuracy is statistically significant when CL is used, as compared to organizing the training data in random batches. © 2021 European Signal Processing Conference, EUSIPCO. All rights reserved
Dyslexia detection in children using eye tracking data based on VGG16 network
No full textPublisher Copyright: © 2022 European Signal Processing Conference, EUSIPCO. All rights reserved.Considering the negative impact dyslexia has on school achievements, dyslexia diagnosis and treatment are found to be of great importance. In this paper, a deep convolutional neural network was developed to detect dyslexia in children ages 7-13, based on gathered eye tracking data. The children read a text written in Serbian on 13 different color configurations (including background and overlay color variations) and the raw gaze coordinates gathered during the trials were formatted into colored images and used to train a deep learning model based on the VGG16 architecture. Several configurations of the convolutional neural network were evaluated, as well as several trial segmentation configurations in order to provide the best overall result. The method was evaluated using subject-wise cross-validation and an accuracy of 87% was achieved. The obtained results show that a combination of convolutional neural network and visual encoding of the eye tracking data shows promising results in dyslexia detection with minimal preprocessing.Peer reviewe
- …
