Pub Date : 2017-11-02DOI: 10.23919/EUSIPCO.2017.8081637
Abdulrahman M. Alanazi, Tarig Ballal, M. Masood, T. Al-Naffouri
The image restoration problem deals with images in which information has been degraded by blur or noise. In this work, we present a new method for image deblurring by solving a regularized linear least-squares problem. In the proposed method, a synthetic perturbation matrix with a bounded norm is forced into the discrete ill-conditioned model matrix. This perturbation is added to enhance the singular-value structure of the matrix and hence to provide an improved solution. A method is proposed to find a near-optimal value of the regularization parameter for the proposed approach. To reduce the computational complexity, we present a technique based on the bootstrapping method to estimate the regularization parameter for both low and high-resolution images. Experimental results on the image deblurring problem are presented. Comparisons are made with three benchmark methods and the results demonstrate that the proposed method clearly outperforms the other methods in terms of both the output PSNR and SSIM values.
{"title":"Image deblurring using a perturbation-basec regularization approach","authors":"Abdulrahman M. Alanazi, Tarig Ballal, M. Masood, T. Al-Naffouri","doi":"10.23919/EUSIPCO.2017.8081637","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081637","url":null,"abstract":"The image restoration problem deals with images in which information has been degraded by blur or noise. In this work, we present a new method for image deblurring by solving a regularized linear least-squares problem. In the proposed method, a synthetic perturbation matrix with a bounded norm is forced into the discrete ill-conditioned model matrix. This perturbation is added to enhance the singular-value structure of the matrix and hence to provide an improved solution. A method is proposed to find a near-optimal value of the regularization parameter for the proposed approach. To reduce the computational complexity, we present a technique based on the bootstrapping method to estimate the regularization parameter for both low and high-resolution images. Experimental results on the image deblurring problem are presented. Comparisons are made with three benchmark methods and the results demonstrate that the proposed method clearly outperforms the other methods in terms of both the output PSNR and SSIM values.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122299259","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 : 2017-10-31DOI: 10.23919/EUSIPCO.2017.8081436
S. Sthapit, J. Hopgood, J. Thompson
Mobile Cloud Computing or Fog computing refer to offloading computationally intensive algorithms from a mobile device to a cloud or a intermediate cloud in order to save resources (time and energy) in the mobile device. In this paper, we look at alternative solution when the cloud or fog is not available. We modelled sensors using network of queues and use linear programming to make scheduling decisions. We then propose novel algorithms which can improve efficiency of the overall system. Results show significant performance improvement at the cost of using some extra energy. Particularly, when incoming job rate is higher, we found our Proactive Centralised gives the best compromise between performance and energy whereas Reactive Distributed is more effective when job rate is lower.
{"title":"Distributed computational load balancing for real-time applications","authors":"S. Sthapit, J. Hopgood, J. Thompson","doi":"10.23919/EUSIPCO.2017.8081436","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081436","url":null,"abstract":"Mobile Cloud Computing or Fog computing refer to offloading computationally intensive algorithms from a mobile device to a cloud or a intermediate cloud in order to save resources (time and energy) in the mobile device. In this paper, we look at alternative solution when the cloud or fog is not available. We modelled sensors using network of queues and use linear programming to make scheduling decisions. We then propose novel algorithms which can improve efficiency of the overall system. Results show significant performance improvement at the cost of using some extra energy. Particularly, when incoming job rate is higher, we found our Proactive Centralised gives the best compromise between performance and energy whereas Reactive Distributed is more effective when job rate is lower.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125448815","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 : 2017-10-26DOI: 10.23919/EUSIPCO.2017.8081629
Yissel Rodríguez Aldana, B. Hunyadi, E. J. M. Reyes, V. Rodriguez, S. Huffel
Nonconvulsive status epilepticus (NCSE) is observed when the patient undergoes a persistent electroencephalographic epileptic episode without physical symptoms. This condition is commonly found in critically ill patients from intensive care units and constitutes a medical emergency. This paper proposes a method to detect nonconvulsive epileptic seizures (NCES). To perform the NCES detection the electroencephalogram (EEG) is represented as a third order tensor with axes frequency χ time χ channels using Wavelet or Hilbert-Huang transform. The signatures obtained from the tensor decomposition are used to train five classifiers to separate between the normal and seizure EEG. Classification is performed in two ways: (1) with each signature of the different modes separately, (2) with all signatures assembled. The algorithm is tested on a database containing 139 nonconvulsive seizures. From all performed analysis, Hilbert-Huang Tensors Space and assembled signatures demonstrate to be the best features to classify between seizure and non-seizure EEG.
{"title":"Nonconvulsive epileptic seizures detection using multiway data analysis","authors":"Yissel Rodríguez Aldana, B. Hunyadi, E. J. M. Reyes, V. Rodriguez, S. Huffel","doi":"10.23919/EUSIPCO.2017.8081629","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081629","url":null,"abstract":"Nonconvulsive status epilepticus (NCSE) is observed when the patient undergoes a persistent electroencephalographic epileptic episode without physical symptoms. This condition is commonly found in critically ill patients from intensive care units and constitutes a medical emergency. This paper proposes a method to detect nonconvulsive epileptic seizures (NCES). To perform the NCES detection the electroencephalogram (EEG) is represented as a third order tensor with axes frequency χ time χ channels using Wavelet or Hilbert-Huang transform. The signatures obtained from the tensor decomposition are used to train five classifiers to separate between the normal and seizure EEG. Classification is performed in two ways: (1) with each signature of the different modes separately, (2) with all signatures assembled. The algorithm is tested on a database containing 139 nonconvulsive seizures. From all performed analysis, Hilbert-Huang Tensors Space and assembled signatures demonstrate to be the best features to classify between seizure and non-seizure EEG.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115876316","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 : 2017-10-26DOI: 10.23919/EUSIPCO.2017.8081434
I. Mitiche, G. Morison, A. Nesbitt, P. Boreham, B. Stewart
In this paper we investigate the application of feature extraction and machine learning techniques to fault identification in power systems. Specifically we implement the novel application of Permutation Entropy-based measures known as Weighted Permutation and Dispersion Entropy to field Electro-Magnetic Interference (EMI) signals for classification of discharge sources, also called conditions, such as partial discharge, arcing and corona which arise from various assets of different power sites. This work introduces two main contributions: the application of entropy measures in condition monitoring and the classification of real field EMI captured signals. The two simple and low dimension features are fed to a Multi-Class Support Vector Machine for the classification of different discharge sources contained in the EMI signals. Classification was performed to distinguish between the conditions observed within each site and between all sites. Results demonstrate that the proposed approach separated and identified the discharge sources successfully.
{"title":"Classification of partial discharge EMI conditions using permutation entropy-based features","authors":"I. Mitiche, G. Morison, A. Nesbitt, P. Boreham, B. Stewart","doi":"10.23919/EUSIPCO.2017.8081434","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081434","url":null,"abstract":"In this paper we investigate the application of feature extraction and machine learning techniques to fault identification in power systems. Specifically we implement the novel application of Permutation Entropy-based measures known as Weighted Permutation and Dispersion Entropy to field Electro-Magnetic Interference (EMI) signals for classification of discharge sources, also called conditions, such as partial discharge, arcing and corona which arise from various assets of different power sites. This work introduces two main contributions: the application of entropy measures in condition monitoring and the classification of real field EMI captured signals. The two simple and low dimension features are fed to a Multi-Class Support Vector Machine for the classification of different discharge sources contained in the EMI signals. Classification was performed to distinguish between the conditions observed within each site and between all sites. Results demonstrate that the proposed approach separated and identified the discharge sources successfully.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130587207","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 : 2017-10-26DOI: 10.23919/EUSIPCO.2017.8081529
Cian O'Brien, Mark D. Plumbley
Automatic Music Transcription (AMT) is concerned with the problem of producing the pitch content of a piece of music given a recorded signal. Many methods rely on sparse or low rank models, where the observed magnitude spectra are represented as a linear combination of dictionary atoms corresponding to individual pitches. Some of the most successful approaches use Non-negative Matrix Decomposition (NMD) or Factorization (NMF), which can be used to learn a dictionary and pitch activation matrix from a given signal. Here we introduce a further refinement of NMD in which we assume the transcription itself is approximately low rank. The intuition behind this approach is that the total number of distinct activation patterns should be relatively small since the pitch content between adjacent frames should be similar. A rank penalty is introduced into the NMD objective function and solved using an iterative algorithm based on Singular Value thresholding. We find that the low rank assumption leads to a significant increase in performance compared to NMD using β-divergence on a standard AMT dataset.
{"title":"Automatic music transcription using low rank non-negative matrix decomposition","authors":"Cian O'Brien, Mark D. Plumbley","doi":"10.23919/EUSIPCO.2017.8081529","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081529","url":null,"abstract":"Automatic Music Transcription (AMT) is concerned with the problem of producing the pitch content of a piece of music given a recorded signal. Many methods rely on sparse or low rank models, where the observed magnitude spectra are represented as a linear combination of dictionary atoms corresponding to individual pitches. Some of the most successful approaches use Non-negative Matrix Decomposition (NMD) or Factorization (NMF), which can be used to learn a dictionary and pitch activation matrix from a given signal. Here we introduce a further refinement of NMD in which we assume the transcription itself is approximately low rank. The intuition behind this approach is that the total number of distinct activation patterns should be relatively small since the pitch content between adjacent frames should be similar. A rank penalty is introduced into the NMD objective function and solved using an iterative algorithm based on Singular Value thresholding. We find that the low rank assumption leads to a significant increase in performance compared to NMD using β-divergence on a standard AMT dataset.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130303516","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 : 2017-10-26DOI: 10.23919/EUSIPCO.2017.8081686
S. Sanei, C. C. Took, Shirin Enshaeifar, T. Lee
The recovery of periodic signals from their noisy single channel mixtures has made wide use of the adaptive line enhancer (ALE). The ALE, however, is not designed for detection of two-(2-D) or three-dimensional (3-D) periodic signals such as tremor in an unconstrained hand motion. An ALE which can perform restoration of 3-D periodic signals is therefore required for such purposes. These signals may not exhibit periodicity in a single dimension. To address and solve this problem a quaternion adaptive line enhancer (QALE) is introduced in this paper for the first time which exploits the quaternion least mean square (QLMS) algorithm for the detection of 3-D (extendable to 4-D) periodic signals.
{"title":"Quaternion adaptive line enhancer","authors":"S. Sanei, C. C. Took, Shirin Enshaeifar, T. Lee","doi":"10.23919/EUSIPCO.2017.8081686","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081686","url":null,"abstract":"The recovery of periodic signals from their noisy single channel mixtures has made wide use of the adaptive line enhancer (ALE). The ALE, however, is not designed for detection of two-(2-D) or three-dimensional (3-D) periodic signals such as tremor in an unconstrained hand motion. An ALE which can perform restoration of 3-D periodic signals is therefore required for such purposes. These signals may not exhibit periodicity in a single dimension. To address and solve this problem a quaternion adaptive line enhancer (QALE) is introduced in this paper for the first time which exploits the quaternion least mean square (QLMS) algorithm for the detection of 3-D (extendable to 4-D) periodic signals.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"108 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127948052","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 : 2017-10-26DOI: 10.23919/EUSIPCO.2017.8081647
M. R. Anbiyaei, W. Liu, D. McLernon
A method is proposed for reducing the effect of white noise in wideband sparse arrays via a combination of a judiciously designed transformation followed by highpass filters. The reduced noise level leads to a higher signal to noise ratio for the system, which can have a significant effect on the performance of various beamforming methods. As a representative example, the reference signal based (RSB) and the Linearly Constrained Minimum Variance (LCMV) beamformers are employed here to demonstrate the improved beamforming performance, as confirmed by simulation results.
{"title":"Performance improvement for wideband beamforming with white noise reduction based on sparse arrays","authors":"M. R. Anbiyaei, W. Liu, D. McLernon","doi":"10.23919/EUSIPCO.2017.8081647","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081647","url":null,"abstract":"A method is proposed for reducing the effect of white noise in wideband sparse arrays via a combination of a judiciously designed transformation followed by highpass filters. The reduced noise level leads to a higher signal to noise ratio for the system, which can have a significant effect on the performance of various beamforming methods. As a representative example, the reference signal based (RSB) and the Linearly Constrained Minimum Variance (LCMV) beamformers are employed here to demonstrate the improved beamforming performance, as confirmed by simulation results.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116208912","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 : 2017-10-26DOI: 10.23919/EUSIPCO.2017.8081581
Mingyang Chen, Wenwu Wang, M. Barnard, J. Chambers
We study the problem of wideband direction of arrival (DoA) estimation by joint optimisation of array and spatial sparsity. Two-step iterative process is proposed. In the first step, the wideband signal is reshaped and used as the input to derive the weight coefficients using a sparse array optimisation method. The weights are then used to scale the observed signal model for which a compressive sensing based spatial sparsity optimisation method is used for DoA estimation. Simulations are provided to demonstrate the performance of the proposed method for both stationary and moving sources.
{"title":"Wideband DoA estimation based on joint optimisation of array and spatial sparsity","authors":"Mingyang Chen, Wenwu Wang, M. Barnard, J. Chambers","doi":"10.23919/EUSIPCO.2017.8081581","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081581","url":null,"abstract":"We study the problem of wideband direction of arrival (DoA) estimation by joint optimisation of array and spatial sparsity. Two-step iterative process is proposed. In the first step, the wideband signal is reshaped and used as the input to derive the weight coefficients using a sparse array optimisation method. The weights are then used to scale the observed signal model for which a compressive sensing based spatial sparsity optimisation method is used for DoA estimation. Simulations are provided to demonstrate the performance of the proposed method for both stationary and moving sources.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127279591","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 : 2017-10-23DOI: 10.23919/EUSIPCO.2017.8081321
T. V. D. Laar, M.G.H. Cox, A. V. Diepen, B. Vries
State-space modeling of non-stationary natural signals is a notoriously difficult task. As a result of context switches, the memory depth of the model should ideally be adapted online. Stabilized linear forgetting (SLF) has been proposed as an elegant method for state-space tracking in context-switching environments. In practice, SLF leads to state and parameter estimation tasks for which no analytical solutions exist. In the literature, a few approximate solutions have been derived, making use of specific model simplifications. This paper proposes an alternative approach, in which SLF is described as an inference task on a generative probabilistic model. SLF is then executed by a variational message passing algorithm on a factor graph representation of the generative model. This approach enjoys a number of advantages relative to previous work. First, variational message passing (VMP) is an automatable procedure that adapts appropriately under changing model assumptions. This eases the search process for the best model. Secondly, VMP easily extends to estimate model parameters. Thirdly, the modular make-up of the factor graph framework allows SLF to be used as a click-on feature in a large variety of complex models. The functionality of the proposed method is verified by simulating an SLF state-space model in a context-switching data environment.
{"title":"Variational stabilized linear forgetting in state-space models","authors":"T. V. D. Laar, M.G.H. Cox, A. V. Diepen, B. Vries","doi":"10.23919/EUSIPCO.2017.8081321","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081321","url":null,"abstract":"State-space modeling of non-stationary natural signals is a notoriously difficult task. As a result of context switches, the memory depth of the model should ideally be adapted online. Stabilized linear forgetting (SLF) has been proposed as an elegant method for state-space tracking in context-switching environments. In practice, SLF leads to state and parameter estimation tasks for which no analytical solutions exist. In the literature, a few approximate solutions have been derived, making use of specific model simplifications. This paper proposes an alternative approach, in which SLF is described as an inference task on a generative probabilistic model. SLF is then executed by a variational message passing algorithm on a factor graph representation of the generative model. This approach enjoys a number of advantages relative to previous work. First, variational message passing (VMP) is an automatable procedure that adapts appropriately under changing model assumptions. This eases the search process for the best model. Secondly, VMP easily extends to estimate model parameters. Thirdly, the modular make-up of the factor graph framework allows SLF to be used as a click-on feature in a large variety of complex models. The functionality of the proposed method is verified by simulating an SLF state-space model in a context-switching data environment.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134436891","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 : 2017-10-23DOI: 10.23919/EUSIPCO.2017.8081478
Motoya Ohnishi, M. Yukawa
We present a novel online learning paradigm for nonlinear function estimation based on iterative orthogonal projections in an L2 space reflecting the stochastic property of input signals. An online algorithm is built upon the fact that any finite dimensional subspace has a reproducing kernel, which is given in terms of the Gram matrix of its basis. The basis used in the present study involves multiple Gaussian kernels. The sequence generated by the algorithm is expected to approach towards the best approximation, in the L2-norm sense, of the nonlinear function to be estimated. This is in sharp contrast to the conventional kernel adaptive filtering paradigm because the best approximation in the reproducing kernel Hilbert space generally differs from the minimum mean squared error estimator over the subspace (Yukawa and Müller 2016). Numerical examples show the efficacy of the proposed approach.
我们提出了一种基于L2空间中反映输入信号随机特性的迭代正交投影的非线性函数估计的新的在线学习范式。一个在线算法是建立在任何有限维子空间都有一个复制核的基础上的，这个复制核是用它的基的格拉姆矩阵给出的。在本研究中使用的基础涉及多个高斯核。在l2范数意义上，期望算法生成的序列接近要估计的非线性函数的最佳近似值。这与传统的核自适应滤波范例形成鲜明对比，因为再现核希尔伯特空间中的最佳近似值通常不同于子空间上的最小均方误差估计量(Yukawa and m ller 2016)。数值算例表明了该方法的有效性。
{"title":"Online learning in L2 space with multiple Gaussian kernels","authors":"Motoya Ohnishi, M. Yukawa","doi":"10.23919/EUSIPCO.2017.8081478","DOIUrl":"https://doi.org/10.23919/EUSIPCO.2017.8081478","url":null,"abstract":"We present a novel online learning paradigm for nonlinear function estimation based on iterative orthogonal projections in an L2 space reflecting the stochastic property of input signals. An online algorithm is built upon the fact that any finite dimensional subspace has a reproducing kernel, which is given in terms of the Gram matrix of its basis. The basis used in the present study involves multiple Gaussian kernels. The sequence generated by the algorithm is expected to approach towards the best approximation, in the L2-norm sense, of the nonlinear function to be estimated. This is in sharp contrast to the conventional kernel adaptive filtering paradigm because the best approximation in the reproducing kernel Hilbert space generally differs from the minimum mean squared error estimator over the subspace (Yukawa and Müller 2016). Numerical examples show the efficacy of the proposed approach.","PeriodicalId":346811,"journal":{"name":"2017 25th European Signal Processing Conference (EUSIPCO)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121873168","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：