This paper proposes methods for acoustic pattern recognition in dynamically changing noise. Parallel model combination and vector Taylor series model compensation techniques are used to adapt acoustic models to noisy conditions are applied together with a time-varying noise estimation algorithm. The noise estimation produces biased noise estimates and therefore we propose methods to accommodate the compensation to the bias. We apply the methods in robust voice activity detection, where frame-wise speech/non-speech classifier is first trained in clean conditions and then tested in and adapted to non-stationary noise conditions. The simulations show that a model compensation with the time-varying noise estimator improves clearly the accuracy of voice activity detection.
{"title":"Non-stationary noise model compensation in voice activity detection","authors":"M. Myllymäki, T. Virtanen","doi":"10.5281/ZENODO.41553","DOIUrl":"https://doi.org/10.5281/ZENODO.41553","url":null,"abstract":"This paper proposes methods for acoustic pattern recognition in dynamically changing noise. Parallel model combination and vector Taylor series model compensation techniques are used to adapt acoustic models to noisy conditions are applied together with a time-varying noise estimation algorithm. The noise estimation produces biased noise estimates and therefore we propose methods to accommodate the compensation to the bias. We apply the methods in robust voice activity detection, where frame-wise speech/non-speech classifier is first trained in clean conditions and then tested in and adapted to non-stationary noise conditions. The simulations show that a model compensation with the time-varying noise estimator improves clearly the accuracy of voice activity detection.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115365919","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}
K. Chahine, V. Baltazart, Yide Wang, Xavier Dérobert, C. L. Bastard
In this paper, we investigate the effects of frequency-dependent attenuation on the performance of some time delay estimation techniques applied to ground penetrating radar (GPR) data. The signal model is based on a complex power law of frequency for dielectric permittivity which describes wave propagation using two parameters, the quality factor Q and the phase velocity at an arbitrary reference frequency Vr. Hence, the adopted model deviates from the damped exponential model and it is this mismatch that is likely to deteriorate the performance of the employed techniques. At first, we carry out a sensitivity study by determining the variations of the relative root mean square error of the time delay estimates as a function of the SNR, Q and the product Bτ for three algorithms, namely, root-MUSIC, ESPRIT and the matrix pencil method (MPM). These variations reveal a systematic error which is quantified by means of a first-order approximation and is found to be the ratio of the phase delay to the group delay. The bandwidth over which this approximation is reasonably accurate depends on the reference frequency and the quality factor of the medium. Then, we use this error to compensate for the bias introduced by the model mismatch with the aim of improving the estimates.
{"title":"Effects of frequency-dependent attenuation on the performance of time delay estimation techniques using ground penetrating radar","authors":"K. Chahine, V. Baltazart, Yide Wang, Xavier Dérobert, C. L. Bastard","doi":"10.5281/ZENODO.41540","DOIUrl":"https://doi.org/10.5281/ZENODO.41540","url":null,"abstract":"In this paper, we investigate the effects of frequency-dependent attenuation on the performance of some time delay estimation techniques applied to ground penetrating radar (GPR) data. The signal model is based on a complex power law of frequency for dielectric permittivity which describes wave propagation using two parameters, the quality factor Q and the phase velocity at an arbitrary reference frequency Vr. Hence, the adopted model deviates from the damped exponential model and it is this mismatch that is likely to deteriorate the performance of the employed techniques. At first, we carry out a sensitivity study by determining the variations of the relative root mean square error of the time delay estimates as a function of the SNR, Q and the product Bτ for three algorithms, namely, root-MUSIC, ESPRIT and the matrix pencil method (MPM). These variations reveal a systematic error which is quantified by means of a first-order approximation and is found to be the ratio of the phase delay to the group delay. The bandwidth over which this approximation is reasonably accurate depends on the reference frequency and the quality factor of the medium. Then, we use this error to compensate for the bias introduced by the model mismatch with the aim of improving the estimates.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"2016 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127353492","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}
The extension of lattice based operators to multivariate images is still a challenging theme in mathematical morphology. In this paper, we propose to explicitly construct complete lattices and replace each element of a multivariate image by its rank, creating a rank image suitable for classical morphological processing. Manifold learning is considered as the basis for the construction of a complete lattice after reducing a multivariate image to its main data by Vector Quantization. A quantitative comparison between usual ordering criteria is performed and experimental results illustrate the abilities of our proposal.
{"title":"Rank transformation and manifold learning for multivariate mathematical morphology","authors":"O. Lézoray, C. Charrier, A. Elmoataz","doi":"10.5281/ZENODO.41282","DOIUrl":"https://doi.org/10.5281/ZENODO.41282","url":null,"abstract":"The extension of lattice based operators to multivariate images is still a challenging theme in mathematical morphology. In this paper, we propose to explicitly construct complete lattices and replace each element of a multivariate image by its rank, creating a rank image suitable for classical morphological processing. Manifold learning is considered as the basis for the construction of a complete lattice after reducing a multivariate image to its main data by Vector Quantization. A quantitative comparison between usual ordering criteria is performed and experimental results illustrate the abilities of our proposal.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127477839","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}
Joint source-channel (JSC) coding is an important alternative to classic separate coding in wireless applications that require robustness without feedback or under stringent delay constraints. JSC schemes based on arithmetic coding can be implemented with finite-state encoders (FSE) generating finite-state codes (FSC). The performance of an FSC is primarily characterized by its free distance, which can be computed with efficient algorithms. This work shows how all FSEs corresponding to a set of initial parameters (source probabilities, arithmetic precision, design rate) can be ordered in a tree data structure. Since an exhaustive search of the code with the largest free distance is very difficult in most cases, a criterion for optimized exploration of the tree of FSEs is provided. Three methods for exploring the tree are proposed and compared with respect to the speed of finding a code with the largest free distance.
{"title":"Optimizing the search of finite-state joint source-channel codes based on arithmetic coding","authors":"A. Diallo, C. Weidmann, M. Kieffer","doi":"10.5281/ZENODO.41593","DOIUrl":"https://doi.org/10.5281/ZENODO.41593","url":null,"abstract":"Joint source-channel (JSC) coding is an important alternative to classic separate coding in wireless applications that require robustness without feedback or under stringent delay constraints. JSC schemes based on arithmetic coding can be implemented with finite-state encoders (FSE) generating finite-state codes (FSC). The performance of an FSC is primarily characterized by its free distance, which can be computed with efficient algorithms. This work shows how all FSEs corresponding to a set of initial parameters (source probabilities, arithmetic precision, design rate) can be ordered in a tree data structure. Since an exhaustive search of the code with the largest free distance is very difficult in most cases, a criterion for optimized exploration of the tree of FSEs is provided. Three methods for exploring the tree are proposed and compared with respect to the speed of finding a code with the largest free distance.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116013630","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}
Dae-Young Hyun, Junhee Heu, Chang-Su Kim, Sang Uk Lee
In this work, we propose an efficient image and video colorization algorithm. To achieve natural colorization, we first define initial color values and compute their reliabilities. For image colorization, initial colors are assigned by the user interaction. For video colorization, initial colors are transferred from the previous frame and their reliabilities are computed based on the motion information. Then, we formulate an energy function and minimize the function to refine the initial colors. Simulation results show that the proposed algorithm provides more natural colorization results on various images and videos than the conventional algorithms.
{"title":"Reliable colorization algorithm for images and videos","authors":"Dae-Young Hyun, Junhee Heu, Chang-Su Kim, Sang Uk Lee","doi":"10.5281/ZENODO.41599","DOIUrl":"https://doi.org/10.5281/ZENODO.41599","url":null,"abstract":"In this work, we propose an efficient image and video colorization algorithm. To achieve natural colorization, we first define initial color values and compute their reliabilities. For image colorization, initial colors are assigned by the user interaction. For video colorization, initial colors are transferred from the previous frame and their reliabilities are computed based on the motion information. Then, we formulate an energy function and minimize the function to refine the initial colors. Simulation results show that the proposed algorithm provides more natural colorization results on various images and videos than the conventional algorithms.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129493184","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}
We show that by considering the estimation of the amplitudes of sinusoids in colored, Gaussian noise as a joint amplitude and noise covariance matrix estimation problem, we obtain an iterative estimator that has the Capon spectral estimator as a special case. The estimator is also closely related to the amplitude and phase estimator (APES). In experiments, the proposed joint estimator in most cases outperforms Capon and APES.
{"title":"An amplitude and covariance matrix estimator for signals in colored Gaussian noise","authors":"J. Jensen, M. G. Christensen, S. H. Jensen","doi":"10.5281/ZENODO.41432","DOIUrl":"https://doi.org/10.5281/ZENODO.41432","url":null,"abstract":"We show that by considering the estimation of the amplitudes of sinusoids in colored, Gaussian noise as a joint amplitude and noise covariance matrix estimation problem, we obtain an iterative estimator that has the Capon spectral estimator as a special case. The estimator is also closely related to the amplitude and phase estimator (APES). In experiments, the proposed joint estimator in most cases outperforms Capon and APES.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"103 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129363623","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}
The analysis of data is typically accompanied by concern as to the correctness of recorded data points; some of the points might be contaminated, thereby distorting the result of the analysis. This paper proposes a novel cluster-based and distribution-independent method for outlier detection. Based on Monte Carlo simulations, the new method is tested with different data distributions and compared with the method of standardised residuals (also known as the z-score). It is shown that the cluster-based approach identifies outliers more reliably, even for a normal data distribution, and the advantages are discussed in detail.
{"title":"Adaptive cluster-based outlier detection","authors":"T. Strutz","doi":"10.5281/ZENODO.41323","DOIUrl":"https://doi.org/10.5281/ZENODO.41323","url":null,"abstract":"The analysis of data is typically accompanied by concern as to the correctness of recorded data points; some of the points might be contaminated, thereby distorting the result of the analysis. This paper proposes a novel cluster-based and distribution-independent method for outlier detection. Based on Monte Carlo simulations, the new method is tested with different data distributions and compared with the method of standardised residuals (also known as the z-score). It is shown that the cluster-based approach identifies outliers more reliably, even for a normal data distribution, and the advantages are discussed in detail.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126482615","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}
We consider the problem of image alignment using random measurements. More specifically, this paper is concerned with estimating a transformation that aligns a given reference image with a query image, assuming that not the images themselves but only random measurements are available. According to the theory behind compressed sensing, random projections of signal manifolds nearly preserve pairwise Euclidean distances when the reduced space is sufficiently large. This suggests that image alignment can be performed effectively based on a sufficient number of random measurements. We build on our previous work in order to show that the corresponding objective function can be decomposed as the difference of two convex functions (DC). Thus, the optimization problem becomes equivalent to a DC program that can be solved by an outer-approximation cutting plane method, which always converges to the globally optimal solution.
{"title":"Optimal image alignment with random measurements","authors":"E. Kokiopoulou, D. Kressner, P. Frossard","doi":"10.5281/ZENODO.41484","DOIUrl":"https://doi.org/10.5281/ZENODO.41484","url":null,"abstract":"We consider the problem of image alignment using random measurements. More specifically, this paper is concerned with estimating a transformation that aligns a given reference image with a query image, assuming that not the images themselves but only random measurements are available. According to the theory behind compressed sensing, random projections of signal manifolds nearly preserve pairwise Euclidean distances when the reduced space is sufficiently large. This suggests that image alignment can be performed effectively based on a sufficient number of random measurements. We build on our previous work in order to show that the corresponding objective function can be decomposed as the difference of two convex functions (DC). Thus, the optimization problem becomes equivalent to a DC program that can be solved by an outer-approximation cutting plane method, which always converges to the globally optimal solution.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128023124","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}
A. Almhdie, P. Lopes-Pereira, S. Même, C. Colombier, V. Brault, F. Szeremeta, B. Doan, R. Lédée, R. Harba, Y. Hérault, J. Beloeil, C. Léger
A semi automatic active contour method based on Chan-Vese model is proposed for the segmentation of mouse brain MR images. First, a 2 ½ D strategy is applied on the axial images to segment the 3D volume of interest. The method takes into account the special shape of the object to segment. Moreover, the user defines the limits where to search these contours and also provides an initial contour. This semi automatic method makes that human intervention is limited and the tedious manual handling is greatly reduced. Results have shown that the brain volumes estimated by the method are identical to expert manually estimated volumes. Last but not least, the new method was used in the analysis of the cerebral malformations linked to Trisomy 21: no significant difference of the brain volumes between Trisomy 21 mice and the control ones were found.
{"title":"Chan-Vese based method to segment mouse brain MRI images: Application to cerebral malformation analysis in Trisomy 21","authors":"A. Almhdie, P. Lopes-Pereira, S. Même, C. Colombier, V. Brault, F. Szeremeta, B. Doan, R. Lédée, R. Harba, Y. Hérault, J. Beloeil, C. Léger","doi":"10.5281/ZENODO.41440","DOIUrl":"https://doi.org/10.5281/ZENODO.41440","url":null,"abstract":"A semi automatic active contour method based on Chan-Vese model is proposed for the segmentation of mouse brain MR images. First, a 2 ½ D strategy is applied on the axial images to segment the 3D volume of interest. The method takes into account the special shape of the object to segment. Moreover, the user defines the limits where to search these contours and also provides an initial contour. This semi automatic method makes that human intervention is limited and the tedious manual handling is greatly reduced. Results have shown that the brain volumes estimated by the method are identical to expert manually estimated volumes. Last but not least, the new method was used in the analysis of the cerebral malformations linked to Trisomy 21: no significant difference of the brain volumes between Trisomy 21 mice and the control ones were found.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127937351","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}
M. Mouri, A. Funase, A. Cichocki, I. Takumi, H. Yasukawa, M. Hata
Anomalous environmental electromagnetic (EM) radiation waves have been reported as the portents of earthquakes. We have been measuring the Extremely Low Frequency (ELF) range all over Japan. Our goal is to predict earthquakes using EM radiation waves. The recorded data often contain signals unrelated to earthquakes. These signals, as noise, confound earthquake prediction efforts. It is necessary to eliminate noises from observed signals in a preprocessing step. In previous researches, we used ISRA, an algorithm of the Non-negative Matrix Factorization (NMF), to estimate source signal. However, ISRA is not robust for outliers because ISRA's cost function is based on square distance. In order to improve robustness, we should use lower order cost function. In this paper, we propose matrix factorization method based on quasi-absolute distance for global signal elimination.
{"title":"Implementation of matrix factorization based on minimizing quasi-absolute distance for electromagnetic global signal elimination","authors":"M. Mouri, A. Funase, A. Cichocki, I. Takumi, H. Yasukawa, M. Hata","doi":"10.5281/ZENODO.41628","DOIUrl":"https://doi.org/10.5281/ZENODO.41628","url":null,"abstract":"Anomalous environmental electromagnetic (EM) radiation waves have been reported as the portents of earthquakes. We have been measuring the Extremely Low Frequency (ELF) range all over Japan. Our goal is to predict earthquakes using EM radiation waves. The recorded data often contain signals unrelated to earthquakes. These signals, as noise, confound earthquake prediction efforts. It is necessary to eliminate noises from observed signals in a preprocessing step. In previous researches, we used ISRA, an algorithm of the Non-negative Matrix Factorization (NMF), to estimate source signal. However, ISRA is not robust for outliers because ISRA's cost function is based on square distance. In order to improve robustness, we should use lower order cost function. In this paper, we propose matrix factorization method based on quasi-absolute distance for global signal elimination.","PeriodicalId":106989,"journal":{"name":"2009 17th European Signal Processing Conference","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2009-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121766654","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：