Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302121
Süreyya Atasever, Ilker Özçelik, Ş. Sağiroğlu
Many detection approaches have been proposed to address growing threat of Distributed Denial of Service (DDoS) attacks on the Internet. The attack detection is the initial step in most of the mitigation systems. This study examined the methods used to detect DDoS attacks with the focus on learning based approaches. These approaches were compared based on their efficiency, operating load and scalability. Finally, it is discussed in details.
{"title":"An Overview of Machine Learning Based Approaches in DDoS Detection","authors":"Süreyya Atasever, Ilker Özçelik, Ş. Sağiroğlu","doi":"10.1109/SIU49456.2020.9302121","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302121","url":null,"abstract":"Many detection approaches have been proposed to address growing threat of Distributed Denial of Service (DDoS) attacks on the Internet. The attack detection is the initial step in most of the mitigation systems. This study examined the methods used to detect DDoS attacks with the focus on learning based approaches. These approaches were compared based on their efficiency, operating load and scalability. Finally, it is discussed in details.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127692055","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302325
Mahmut Yurt, Tolga Cukurmm
Özetçe —Yüksek çözünürlüklü manyetik rezonans görüntülerinin (MRG) farklı kontrastlar altında edinimi klinik tanıda gerekli olan teşhis bilgisini artırır. Ancak, artan gürültü oranı, uzun tarama süreleri ve donanım maliyetlerinden ötürü yüksek çözünürlüklü görüntülerin edinimi pratikte mümkün olmayabilir. Bu durumlarda, düşük çözünürlüklü görüntülerden yüksek çözünürlüklü görüntülerin üretilebilmesi alternatif bir çözüm olabilir. Yaygın yöntemler tek bir görüntünün süper çözünürlüğünü yapar. Ancak, çok kontrastlı MRG’de, tek bir kontrastın düşük çözünürlüklü görüntüsü başarılı bir netleştirme için gerekli ön bilgiyi içermez. Gerekli bilgiyi zenginleştirebilmek için, farklı kontrastlardaki tamamlayıcı ön bilgiler kullanılabilir. Bu sebeple, bu çalışmada birden çok kontrasta ait görüntüleri eşzamanlı olarak netleştiren bir çoklu kontrast MRG süper çözünürlük yöntemi önerilmiştir. Önerilen yöntem yüksek frekans detaylarını daha iyi kurtararak olabildiğince gerçekçi hedef görüntüler üretebilen koşullu çekişmeli üretici ağlara dayanmaktadır. Çoklu kontrast MR görüntüleri içeren veri setinde yapılan sayısal ve görsel değerlendirmeler, önerilen yöntemin alternatif tekli görüntü MRG süper çözünürlük yönteminden daha üstün performans gösterdiğini ortaya koymuştur.
{"title":"Multi-Image Super Resolution in Multi-Contrast MRI","authors":"Mahmut Yurt, Tolga Cukurmm","doi":"10.1109/SIU49456.2020.9302325","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302325","url":null,"abstract":"Özetçe —Yüksek çözünürlüklü manyetik rezonans görüntülerinin (MRG) farklı kontrastlar altında edinimi klinik tanıda gerekli olan teşhis bilgisini artırır. Ancak, artan gürültü oranı, uzun tarama süreleri ve donanım maliyetlerinden ötürü yüksek çözünürlüklü görüntülerin edinimi pratikte mümkün olmayabilir. Bu durumlarda, düşük çözünürlüklü görüntülerden yüksek çözünürlüklü görüntülerin üretilebilmesi alternatif bir çözüm olabilir. Yaygın yöntemler tek bir görüntünün süper çözünürlüğünü yapar. Ancak, çok kontrastlı MRG’de, tek bir kontrastın düşük çözünürlüklü görüntüsü başarılı bir netleştirme için gerekli ön bilgiyi içermez. Gerekli bilgiyi zenginleştirebilmek için, farklı kontrastlardaki tamamlayıcı ön bilgiler kullanılabilir. Bu sebeple, bu çalışmada birden çok kontrasta ait görüntüleri eşzamanlı olarak netleştiren bir çoklu kontrast MRG süper çözünürlük yöntemi önerilmiştir. Önerilen yöntem yüksek frekans detaylarını daha iyi kurtararak olabildiğince gerçekçi hedef görüntüler üretebilen koşullu çekişmeli üretici ağlara dayanmaktadır. Çoklu kontrast MR görüntüleri içeren veri setinde yapılan sayısal ve görsel değerlendirmeler, önerilen yöntemin alternatif tekli görüntü MRG süper çözünürlük yönteminden daha üstün performans gösterdiğini ortaya koymuştur.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128145044","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302305
E. Arslan, A. T. Dogukan, E. Başar
Ultra-reliable and low-latency communications (URLLC) partake a major role in future communication systems. A possible strong candidate for future URLLC networks is sparse vector coding (SVC), which enables a superior performance in terms of bit error rate (BER). In SVC, virtual digital domain (VDD) and compressed sensing (CS) algorithms are used to encode and decode information. In this paper, orthogonal frequency division multiplexing (OFDM)-based a novel system called orthogonal frequency division multiplexing with codebook index modulation (OFDM-CIM) and which can meet the needs of URLLC systems has been proposed. In OFDM-CIM, information bits are transmitted via both active subcarrier indices and codebook indices. As a result of computer simulations, OFDMCIM is presented as a strong candidate for next generation communication systems.
{"title":"Orthogonal Frequency Division Multiplexing with Codebook Index Modulation","authors":"E. Arslan, A. T. Dogukan, E. Başar","doi":"10.1109/SIU49456.2020.9302305","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302305","url":null,"abstract":"Ultra-reliable and low-latency communications (URLLC) partake a major role in future communication systems. A possible strong candidate for future URLLC networks is sparse vector coding (SVC), which enables a superior performance in terms of bit error rate (BER). In SVC, virtual digital domain (VDD) and compressed sensing (CS) algorithms are used to encode and decode information. In this paper, orthogonal frequency division multiplexing (OFDM)-based a novel system called orthogonal frequency division multiplexing with codebook index modulation (OFDM-CIM) and which can meet the needs of URLLC systems has been proposed. In OFDM-CIM, information bits are transmitted via both active subcarrier indices and codebook indices. As a result of computer simulations, OFDMCIM is presented as a strong candidate for next generation communication systems.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132537658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302468
Alperen Erdoğan, S. Guney
Nowadays, one of the most important illness is heart disease which cause of mostly patients dead. Medical diagnosis of heart diseases is very difficult. While heart diseases are diagnosed medically, they can be confused with other diseases that show same symptoms such as chest pain, shortness of breath, palpitations and nausea. This makes it difficult to diagnose heart diseases medically. In this study, the presence of heart diseases was determined by using machine learning algorithms. In this study, the data obtained from the patients were weighted according to their effects on the success rate. In this study, a method is proposed for determine weight coefficient. According to proposed method's results, 86,90% success was achieved with 13 different features obtained from the patients.
{"title":"Heart Disease Prediction by Using Machine Learning Algorithms","authors":"Alperen Erdoğan, S. Guney","doi":"10.1109/SIU49456.2020.9302468","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302468","url":null,"abstract":"Nowadays, one of the most important illness is heart disease which cause of mostly patients dead. Medical diagnosis of heart diseases is very difficult. While heart diseases are diagnosed medically, they can be confused with other diseases that show same symptoms such as chest pain, shortness of breath, palpitations and nausea. This makes it difficult to diagnose heart diseases medically. In this study, the presence of heart diseases was determined by using machine learning algorithms. In this study, the data obtained from the patients were weighted according to their effects on the success rate. In this study, a method is proposed for determine weight coefficient. According to proposed method's results, 86,90% success was achieved with 13 different features obtained from the patients.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"E-29 3","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132693793","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302370
Fatih Yazıcı, Ayhan Sefa Yıldız, Alper Yazar, E. G. Schmidt
In this paper, we propose a scalable on-chip packet switch architecture for hardware accelerated cloud computing systems. Our proposed switch architecture is implemented on the FPGA and interconnects reconfigurable regions, 40 Gbps Ethernet interfaces and a PCIe interface. The switch fabric operates at line speed to achieve scalability. We propose a new algorithm that grants access to the fabric according to the allocated prioritization to input-output port pairs. The switch is implemented on Xilinx Zynq 7000-SoC and can work at 40 Gbps rate. Our simulation results show that our proposed algorithm achieves desired prioritization without degrading the throughput. Keywords—cloud computing, on-chip switch, switch fabric arbitration.
{"title":"An On-chip Switch Architecture for Hardware Accelerated Cloud Computing Systems","authors":"Fatih Yazıcı, Ayhan Sefa Yıldız, Alper Yazar, E. G. Schmidt","doi":"10.1109/SIU49456.2020.9302370","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302370","url":null,"abstract":"In this paper, we propose a scalable on-chip packet switch architecture for hardware accelerated cloud computing systems. Our proposed switch architecture is implemented on the FPGA and interconnects reconfigurable regions, 40 Gbps Ethernet interfaces and a PCIe interface. The switch fabric operates at line speed to achieve scalability. We propose a new algorithm that grants access to the fabric according to the allocated prioritization to input-output port pairs. The switch is implemented on Xilinx Zynq 7000-SoC and can work at 40 Gbps rate. Our simulation results show that our proposed algorithm achieves desired prioritization without degrading the throughput. Keywords—cloud computing, on-chip switch, switch fabric arbitration.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133082430","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302090
Cahfer Güngen, Özlem Polat, R. Karakis
The classification of brain tumors has great importance in medical applications that benefit from computer-assisted diagnosis. Misdiagnosis of brain tumor types, both prevents the patient's response to treatment effectively and reduce the chance of survival. This study proposes a solution for the classification of brain tumors using MR images. The most common brain tumors, glioma, meningioma and pituitary, are detected using convolutional neural networks. The convolutional network is trained and tested on an accessible Figshare dataset containing 3064 MR images using four different optimizers. AUC, sensitivity, specificity and accuracy are used as performance measure. The proposed method is comparable to the literature and classifies brain tumors with an average accuracy of 96.84% and a maximum accuracy of 97.75%.
{"title":"Classification of Brain Tumors using Convolutional Neural Network from MR Images","authors":"Cahfer Güngen, Özlem Polat, R. Karakis","doi":"10.1109/SIU49456.2020.9302090","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302090","url":null,"abstract":"The classification of brain tumors has great importance in medical applications that benefit from computer-assisted diagnosis. Misdiagnosis of brain tumor types, both prevents the patient's response to treatment effectively and reduce the chance of survival. This study proposes a solution for the classification of brain tumors using MR images. The most common brain tumors, glioma, meningioma and pituitary, are detected using convolutional neural networks. The convolutional network is trained and tested on an accessible Figshare dataset containing 3064 MR images using four different optimizers. AUC, sensitivity, specificity and accuracy are used as performance measure. The proposed method is comparable to the literature and classifies brain tumors with an average accuracy of 96.84% and a maximum accuracy of 97.75%.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132120474","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302119
Anil Çelik, B. Yildirim
In this work, a hybrid artificial intelligence approach to classify a text as clean or profane is proposed. Focus of the proposed approach is to filter profane text shared by people who abuse the anonymous nature of the internet. Independent of any platform, proposed approach only requires a piece of text to work. In the first step, input text gets pre-processed by an iterative cleaning algorithm. Then, processed text is sent to a heuristic or artificial intelligence based approach depending on the word count. Artificial intelligence based approach uses three unique models to process text. Each model independently calculates profanity probability. For the last step, calculations are sent to a judge model to oversee the classification. Finally, either heuristic or artificial intelligence based approach generates a binary response for classification. It is presumed that the proposed approach will be beneficial for Turkish profanity detection on internet platforms.
{"title":"Turkish Profanity Detection Enhanced by Artificial Intelligence","authors":"Anil Çelik, B. Yildirim","doi":"10.1109/SIU49456.2020.9302119","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302119","url":null,"abstract":"In this work, a hybrid artificial intelligence approach to classify a text as clean or profane is proposed. Focus of the proposed approach is to filter profane text shared by people who abuse the anonymous nature of the internet. Independent of any platform, proposed approach only requires a piece of text to work. In the first step, input text gets pre-processed by an iterative cleaning algorithm. Then, processed text is sent to a heuristic or artificial intelligence based approach depending on the word count. Artificial intelligence based approach uses three unique models to process text. Each model independently calculates profanity probability. For the last step, calculations are sent to a judge model to oversee the classification. Finally, either heuristic or artificial intelligence based approach generates a binary response for classification. It is presumed that the proposed approach will be beneficial for Turkish profanity detection on internet platforms.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115750407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302109
Muzaffer Özbey, T. Çukur
In recent years, deep learning techniques have been used in computer science as well as in many other disciplines. Successful detection of complex relationships and connections within big data enables effective use of deep learning in many areas. Deep learning with the detection of patterns and abnormalities in images is also a promising method for the field of Radiology. Detection of abnormalities in MR images enables detection of brain tumor and can automate this process. However, the deep learning models developed for brain tumor detection are sensitive to missing MR images in the input data and therefore the model is not robust enough. One of the models used for brain tumor detection requires a combined MR image in 4 different contrast and sequence; T1, T2, Flair and T1c images. In this study, it is proposed to synthesize the missing contrast image in the input data with another deep learning technique. Incomplete T1c MR image was synthesized by the generative adversarial networks (GAN) method and brain tumor detection performance was examined.
{"title":"T1-Weighted Contrast-Enhanced Synthesis for Multi-Contrast MRI Segmentation","authors":"Muzaffer Özbey, T. Çukur","doi":"10.1109/SIU49456.2020.9302109","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302109","url":null,"abstract":"In recent years, deep learning techniques have been used in computer science as well as in many other disciplines. Successful detection of complex relationships and connections within big data enables effective use of deep learning in many areas. Deep learning with the detection of patterns and abnormalities in images is also a promising method for the field of Radiology. Detection of abnormalities in MR images enables detection of brain tumor and can automate this process. However, the deep learning models developed for brain tumor detection are sensitive to missing MR images in the input data and therefore the model is not robust enough. One of the models used for brain tumor detection requires a combined MR image in 4 different contrast and sequence; T1, T2, Flair and T1c images. In this study, it is proposed to synthesize the missing contrast image in the input data with another deep learning technique. Incomplete T1c MR image was synthesized by the generative adversarial networks (GAN) method and brain tumor detection performance was examined.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"179 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124472368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302278
M. Koroglu, F. Arıkan, O. Arikan
Ionosphere has several effects on radio signals. Traveling wave-like ionospheric disturbances can occur due to solar radiations, seismic activities. Receiver loss of locks and some positioning errors can occur because of these disturbances. The time frequency structure of ionosphere has to be determined in order to detect this kind of disturbances. Slant Total Electron Content (STEC) is a metric that characterize the ionosphere and defined as the line integral of electron density on a ray path from satellite to receiver. In this study, variational mode decomposition method is applied on STEC values to understand the time frequency behavior for disturbances for the first time. Daily STEC values are decomposed into its components in time and frequency domain by using VMD method. By the help of this method, periodic structures of ionospheric disturbances are determined. The behavior of the disturbances varies with the satellite direction and both in time and frequency domain. Probability density functions of frequencies in West and North directions are similar whereas probability density function computed in east direction differs from other direction over Turkey.
{"title":"Time-Frequency Analysis of Ionosphere Disturbances by Using Variational Mode Decomposition","authors":"M. Koroglu, F. Arıkan, O. Arikan","doi":"10.1109/SIU49456.2020.9302278","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302278","url":null,"abstract":"Ionosphere has several effects on radio signals. Traveling wave-like ionospheric disturbances can occur due to solar radiations, seismic activities. Receiver loss of locks and some positioning errors can occur because of these disturbances. The time frequency structure of ionosphere has to be determined in order to detect this kind of disturbances. Slant Total Electron Content (STEC) is a metric that characterize the ionosphere and defined as the line integral of electron density on a ray path from satellite to receiver. In this study, variational mode decomposition method is applied on STEC values to understand the time frequency behavior for disturbances for the first time. Daily STEC values are decomposed into its components in time and frequency domain by using VMD method. By the help of this method, periodic structures of ionospheric disturbances are determined. The behavior of the disturbances varies with the satellite direction and both in time and frequency domain. Probability density functions of frequencies in West and North directions are similar whereas probability density function computed in east direction differs from other direction over Turkey.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125077480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-10-05DOI: 10.1109/SIU49456.2020.9302410
Sedat Demirbag, Mustafa Erden, L. Arslan
In this work, we have developed an end-to-end approach for text dependent speaker verification task. With this method, phonetic labels are fused with spectral features, and used to train a neural network for same/different speaker decision. The data used for tests is obtained from a real call center integrated voice response system. It consists of audio taken from calls made by people at different times in which they utter a specific, short sentence in Turkish. Contribution of in-domain data with target sentence and free format human-human call data for model training is investigated. For the inclusion of phonetic information in modelling three different methods are applied which are phoneme boundary, utterance boundary and phoneme boundary group. Test results show that, we attain an equal error rate of 10.7% for speaker verification on given dataset. Keywords—speaker verification, text dependent, voice biometrics, artificial neural network, end-to-end.
{"title":"End-To-End Phonetic Neural Network Approach for Speaker Verification","authors":"Sedat Demirbag, Mustafa Erden, L. Arslan","doi":"10.1109/SIU49456.2020.9302410","DOIUrl":"https://doi.org/10.1109/SIU49456.2020.9302410","url":null,"abstract":"In this work, we have developed an end-to-end approach for text dependent speaker verification task. With this method, phonetic labels are fused with spectral features, and used to train a neural network for same/different speaker decision. The data used for tests is obtained from a real call center integrated voice response system. It consists of audio taken from calls made by people at different times in which they utter a specific, short sentence in Turkish. Contribution of in-domain data with target sentence and free format human-human call data for model training is investigated. For the inclusion of phonetic information in modelling three different methods are applied which are phoneme boundary, utterance boundary and phoneme boundary group. Test results show that, we attain an equal error rate of 10.7% for speaker verification on given dataset. Keywords—speaker verification, text dependent, voice biometrics, artificial neural network, end-to-end.","PeriodicalId":312627,"journal":{"name":"2020 28th Signal Processing and Communications Applications Conference (SIU)","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125444010","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：