Pub Date : 2014-05-26DOI: 10.1109/CIP.2014.6844505
Jair Montoya-Martínez, Antonio Artés-Rodríguez, M. Pontil
We consider the estimation of the Brain Electrical Sources (BES) matrix from noisy EEG measurements, commonly named as the EEG inverse problem. We propose a new method based on the factorization of the BES as a product of a sparse coding matrix and a dense latent source matrix. This structure is enforced by minimizing a regularized functional that includes the ℓ21-norm of the coding matrix and the squared Frobenius norm of the latent source matrix. We develop an alternating optimization algorithm to solve the resulting nonsmooth-nonconvex minimization problem. We have evaluated our approach under a simulated scenario consisting on estimating a synthetic BES matrix with 5124 sources. We compare the performance of our method respect to the Lasso, Group Lasso, Sparse Group Lasso and Trace norm regularizers.
我们考虑从噪声脑电测量中估计脑电源(BES)矩阵,通常称为脑电逆问题。我们提出了一种基于将BES分解为稀疏编码矩阵和密集潜在源矩阵乘积的新方法。这种结构是通过最小化一个正则泛函来实现的,该泛函包含编码矩阵的l21范数和潜在源矩阵的Frobenius范数的平方。我们开发了一种交替优化算法来解决由此产生的非光滑-非凸最小化问题。我们在一个模拟场景下评估了我们的方法,该场景包括估算具有5124个源的合成BES矩阵。我们比较了我们的方法在Lasso、Group Lasso、Sparse Group Lasso和Trace范数正则化方面的性能。
{"title":"Structured sparse-low rank matrix factorization for the EEG inverse problem","authors":"Jair Montoya-Martínez, Antonio Artés-Rodríguez, M. Pontil","doi":"10.1109/CIP.2014.6844505","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844505","url":null,"abstract":"We consider the estimation of the Brain Electrical Sources (BES) matrix from noisy EEG measurements, commonly named as the EEG inverse problem. We propose a new method based on the factorization of the BES as a product of a sparse coding matrix and a dense latent source matrix. This structure is enforced by minimizing a regularized functional that includes the ℓ21-norm of the coding matrix and the squared Frobenius norm of the latent source matrix. We develop an alternating optimization algorithm to solve the resulting nonsmooth-nonconvex minimization problem. We have evaluated our approach under a simulated scenario consisting on estimating a synthetic BES matrix with 5124 sources. We compare the performance of our method respect to the Lasso, Group Lasso, Sparse Group Lasso and Trace norm regularizers.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126239987","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844509
Camilla Birgitte Falk Jensen, Michael Kai Petersen, J. E. Larsen
With neuroimaging studies showing promising results for discrimination of affective responses, the perspectives of applying these to create more personalised interfaces that adapt to our preferences in real-time seems within reach. Additionally the emergence of wireless electroencephalograph (EEG) neuroheadsets and smartphone brainscanners widens the possibilities for this to be used in mobile settings on a consumer level. However the neural signatures of emotional responses are characterized by small voltage changes that would be highly susceptible to noise if captured in a mobile context. Hypothesizing that retrieval of emotional responses in mobile usage scenarios could be enhanced through spatial filtering, we compare a standard EEG electrode-based analysis against an approach based on independent component analysis (ICA). By clustering scalp maps and time series responses we identify neural signatures that are differentially modulated when passively viewing neutral, pleasant and unpleasant images. While early responses can be detected from the raw EEG signal, we identify multiple early and late ICA components that are modulated by emotional content. We propose that similar approaches to spatial filtering might allow us to retrieve more robust signals in real-life mobile usage scenarios, and potentially facilitate design of cognitive interfaces that adapt the selection of media to our emotional responses.
{"title":"Emotional responses as independent components in EEG","authors":"Camilla Birgitte Falk Jensen, Michael Kai Petersen, J. E. Larsen","doi":"10.1109/CIP.2014.6844509","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844509","url":null,"abstract":"With neuroimaging studies showing promising results for discrimination of affective responses, the perspectives of applying these to create more personalised interfaces that adapt to our preferences in real-time seems within reach. Additionally the emergence of wireless electroencephalograph (EEG) neuroheadsets and smartphone brainscanners widens the possibilities for this to be used in mobile settings on a consumer level. However the neural signatures of emotional responses are characterized by small voltage changes that would be highly susceptible to noise if captured in a mobile context. Hypothesizing that retrieval of emotional responses in mobile usage scenarios could be enhanced through spatial filtering, we compare a standard EEG electrode-based analysis against an approach based on independent component analysis (ICA). By clustering scalp maps and time series responses we identify neural signatures that are differentially modulated when passively viewing neutral, pleasant and unpleasant images. While early responses can be detected from the raw EEG signal, we identify multiple early and late ICA components that are modulated by emotional content. We propose that similar approaches to spatial filtering might allow us to retrieve more robust signals in real-life mobile usage scenarios, and potentially facilitate design of cognitive interfaces that adapt the selection of media to our emotional responses.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122255469","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844501
P. Kendrick, T. Cox, Francis F. Li, B. Fazenda, Iain Jackson
Microphone handling noise is a common problem with user-generated content. It can occur when the operator inadvertently knocks or brushes a recording device. Handling noise may be impulsive, where a microphone is knocked, or a more sustained rubbing noise, when the microphone is brushed against something. A detector able to accurately detect handling noises caused by rubbing while recording speech, music or quotidian sounds has been developed. Ensembles of decision trees were trained to classify handling noise level over 23 ms frames; a second ensemble flags frames when the noise may be masked by foreground audio. Aggregation of the detection over 1 s yielded a Matthews correlation coefficient of 0.91.
{"title":"Automatic detection of microphone handling noise","authors":"P. Kendrick, T. Cox, Francis F. Li, B. Fazenda, Iain Jackson","doi":"10.1109/CIP.2014.6844501","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844501","url":null,"abstract":"Microphone handling noise is a common problem with user-generated content. It can occur when the operator inadvertently knocks or brushes a recording device. Handling noise may be impulsive, where a microphone is knocked, or a more sustained rubbing noise, when the microphone is brushed against something. A detector able to accurately detect handling noises caused by rubbing while recording speech, music or quotidian sounds has been developed. Ensembles of decision trees were trained to classify handling noise level over 23 ms frames; a second ensemble flags frames when the noise may be masked by foreground audio. Aggregation of the detection over 1 s yielded a Matthews correlation coefficient of 0.91.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134536452","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844498
Mikkel N. Schmidt, Tue Herlau, Morten Mørup
Hierarchical clustering is a widely used tool for structuring and visualizing complex data using similarity. Traditionally, hierarchical clustering is based on local heuristics that do not explicitly provide assessment of the statistical saliency of the extracted hierarchy. We propose a non-parametric generative model for hierarchical clustering of similarity based on multifurcating Gibbs fragmentation trees. This allows us to infer and display the posterior distribution of hierarchical structures that comply with the data. We demonstrate the utility of our method on synthetic data and data of functional brain connectivity.
{"title":"Discovering hierarchical structure in normal relational data","authors":"Mikkel N. Schmidt, Tue Herlau, Morten Mørup","doi":"10.1109/CIP.2014.6844498","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844498","url":null,"abstract":"Hierarchical clustering is a widely used tool for structuring and visualizing complex data using similarity. Traditionally, hierarchical clustering is based on local heuristics that do not explicitly provide assessment of the statistical saliency of the extracted hierarchy. We propose a non-parametric generative model for hierarchical clustering of similarity based on multifurcating Gibbs fragmentation trees. This allows us to infer and display the posterior distribution of hierarchical structures that comply with the data. We demonstrate the utility of our method on synthetic data and data of functional brain connectivity.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123958615","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844510
J. Andersen
This paper sums up the preliminary observations and challenges encountered during my first engaging with the music intelligence company Echo Nest's automatically derived data of more than 35 million songs. The overall purpose is to investigate whether musicologists can draw benefit from Echo Nest's API, and to explore what practical and analytical consideration one should take into account when engaging with the numbers derived from the Echo Nest API. This paper suggests that the Echo Nest API hold a large potential of doing new types of analyses and visualizing the results. But it concurrently argues that a careful and critical approach is requisite, when interpreting the results.
{"title":"Using the Echo Nest's automatically extracted music features for a musicological purpose","authors":"J. Andersen","doi":"10.1109/CIP.2014.6844510","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844510","url":null,"abstract":"This paper sums up the preliminary observations and challenges encountered during my first engaging with the music intelligence company Echo Nest's automatically derived data of more than 35 million songs. The overall purpose is to investigate whether musicologists can draw benefit from Echo Nest's API, and to explore what practical and analytical consideration one should take into account when engaging with the numbers derived from the Echo Nest API. This paper suggests that the Echo Nest API hold a large potential of doing new types of analyses and visualizing the results. But it concurrently argues that a careful and critical approach is requisite, when interpreting the results.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127807466","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844507
B. K. Das, L. A. Azpicueta-Ruiz, M. Chakraborty, J. Arenas-García
In this paper, we review two families for sparsity-aware adaptive filtering. Proportionate-type NLMS filters try to accelerate filter convergence by assigning each filter weight a different gain that depends on its actual value. Sparsity-norm regularized filters penalize the cost function minimized by the filter using sparsity-promoting norms (such as ℓ0 or ℓ1) and derive new stochastic gradient descent rules from the regularized cost function. We compare both families of algorithms in terms of computational complexity and studying how well they deal with the convergence vs steady-state error tradeoff. We conclude that sparsity-norm regularized filters are computationally less expensive and can achieve a better tradeoff, making them more attractive in principle. However, selection of the strength of the regularization term seems to be a critical element for the good performance of these filters.
{"title":"A comparative study of two popular families of sparsity-aware adaptive filters","authors":"B. K. Das, L. A. Azpicueta-Ruiz, M. Chakraborty, J. Arenas-García","doi":"10.1109/CIP.2014.6844507","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844507","url":null,"abstract":"In this paper, we review two families for sparsity-aware adaptive filtering. Proportionate-type NLMS filters try to accelerate filter convergence by assigning each filter weight a different gain that depends on its actual value. Sparsity-norm regularized filters penalize the cost function minimized by the filter using sparsity-promoting norms (such as ℓ0 or ℓ1) and derive new stochastic gradient descent rules from the regularized cost function. We compare both families of algorithms in terms of computational complexity and studying how well they deal with the convergence vs steady-state error tradeoff. We conclude that sparsity-norm regularized filters are computationally less expensive and can achieve a better tradeoff, making them more attractive in principle. However, selection of the strength of the regularization term seems to be a critical element for the good performance of these filters.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115457908","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844515
F. Castaldo, F. Palmieri
Propagation of Gaussian belief messages in factor graphs in normal form is applied to data fusion for tracking moving objects in maritime scenarios, as crowded harbors. The data are yielded by multiple cameras, deployed in the region under surveillance, and AIS system, wherever is available. The track model and the estimates coming from the sensors are integrated bi-directionally, providing a flexible framework for comprehensive inference. The framework is applied to tracking a large cargo ship in a harbor from frames recorded with three commercial cameras.
{"title":"Application of factor graphs to multi-camera fusion for maritime tracking","authors":"F. Castaldo, F. Palmieri","doi":"10.1109/CIP.2014.6844515","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844515","url":null,"abstract":"Propagation of Gaussian belief messages in factor graphs in normal form is applied to data fusion for tracking moving objects in maritime scenarios, as crowded harbors. The data are yielded by multiple cameras, deployed in the region under surveillance, and AIS system, wherever is available. The track model and the estimates coming from the sensors are integrated bi-directionally, providing a flexible framework for comprehensive inference. The framework is applied to tracking a large cargo ship in a harbor from frames recorded with three commercial cameras.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122093907","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844506
I. Valera, Francisco J. R. Ruiz, F. Pérez-Cruz
Bayesian nonparametric models allow solving estimation and detection problems with an unbounded number of degrees of freedom. In multiuser multiple-input multiple-output (MIMO) communication systems we might not know the number of active users and the channel they face, and assuming maximal scenarios (maximum number of transmitters and maximum channel length) might degrade the receiver performance. In this paper, we propose a Bayesian nonparametric prior and its associated inference algorithm, which is able to detect an unbounded number of users with an unbounded channel length. This generative model provides the dispersive channel model for each user and a probabilistic estimate for each transmitted symbol in a fully blind manner, i.e., without the need of pilot (training) symbols.
{"title":"Infinite factorial unbounded hidden Markov model for blind multiuser channel estimation","authors":"I. Valera, Francisco J. R. Ruiz, F. Pérez-Cruz","doi":"10.1109/CIP.2014.6844506","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844506","url":null,"abstract":"Bayesian nonparametric models allow solving estimation and detection problems with an unbounded number of degrees of freedom. In multiuser multiple-input multiple-output (MIMO) communication systems we might not know the number of active users and the channel they face, and assuming maximal scenarios (maximum number of transmitters and maximum channel length) might degrade the receiver performance. In this paper, we propose a Bayesian nonparametric prior and its associated inference algorithm, which is able to detect an unbounded number of users with an unbounded channel length. This generative model provides the dispersive channel model for each user and a probabilistic estimate for each transmitted symbol in a fully blind manner, i.e., without the need of pilot (training) symbols.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132129319","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844496
P. Bouboulis, G. Papageorgiou, S. Theodoridis
We present a robust method for the image denoising task based on kernel ridge regression and sparse modeling. Added noise is assumed to consist of two parts. One part is impulse noise assumed to be sparse (outliers), while the other part is bounded noise. The noisy image is divided into small regions of interest, whose pixels are regarded as points of a two-dimensional surface. A kernel based ridge regression method, whose parameters are selected adaptively, is employed to fit the data, whereas the outliers are detected via the use of the increasingly popular orthogonal matching pursuit (OMP) algorithm. To this end, a new variant of the OMP rationale is employed that has the additional advantage to automatically terminate, when all outliers have been selected.
{"title":"Robust image denoising in RKHS via orthogonal matching pursuit","authors":"P. Bouboulis, G. Papageorgiou, S. Theodoridis","doi":"10.1109/CIP.2014.6844496","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844496","url":null,"abstract":"We present a robust method for the image denoising task based on kernel ridge regression and sparse modeling. Added noise is assumed to consist of two parts. One part is impulse noise assumed to be sparse (outliers), while the other part is bounded noise. The noisy image is divided into small regions of interest, whose pixels are regarded as points of a two-dimensional surface. A kernel based ridge regression method, whose parameters are selected adaptively, is employed to fit the data, whereas the outliers are detected via the use of the increasingly popular orthogonal matching pursuit (OMP) algorithm. To this end, a new variant of the OMP rationale is employed that has the additional advantage to automatically terminate, when all outliers have been selected.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121755564","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 : 2014-05-26DOI: 10.1109/CIP.2014.6844500
F. Palmieri, A. Buonanno
In modeling time series, convolution multi-layer graphs are able to capture long-term dependence at a gradually increasing scale. We present an approach to learn a layered factor graph architecture starting from a stationary latent models for each layer. Simulations of belief propagation are reported for a three-layer graph on a small data set of characters.
{"title":"Belief propagation and learning in convolution multi-layer factor graphs","authors":"F. Palmieri, A. Buonanno","doi":"10.1109/CIP.2014.6844500","DOIUrl":"https://doi.org/10.1109/CIP.2014.6844500","url":null,"abstract":"In modeling time series, convolution multi-layer graphs are able to capture long-term dependence at a gradually increasing scale. We present an approach to learn a layered factor graph architecture starting from a stationary latent models for each layer. Simulations of belief propagation are reported for a three-layer graph on a small data set of characters.","PeriodicalId":117669,"journal":{"name":"2014 4th International Workshop on Cognitive Information Processing (CIP)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2014-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126190120","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}