Pub Date : 2002-11-07DOI: 10.1109/NNSP.2002.1030032
Deniz Erdoğmuş, J. Príncipe, Sung-Phil Kim, Justin C. Sanchez
Estimating the entropy of a sample set is required, in solving numerous learning scenarios involving information theoretic optimization criteria. A number of entropy estimators are available in the literature; however, these require a batch of samples to operate on in order to yield an estimate. We derive a recursive formula to estimate Renyi's (1970) quadratic entropy on-line, using each new sample to update the entropy estimate to obtain more accurate results in stationary situations or to track the changing entropy of a signal in nonstationary situations.
{"title":"A recursive Renyi's entropy estimator","authors":"Deniz Erdoğmuş, J. Príncipe, Sung-Phil Kim, Justin C. Sanchez","doi":"10.1109/NNSP.2002.1030032","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030032","url":null,"abstract":"Estimating the entropy of a sample set is required, in solving numerous learning scenarios involving information theoretic optimization criteria. A number of entropy estimators are available in the literature; however, these require a batch of samples to operate on in order to yield an estimate. We derive a recursive formula to estimate Renyi's (1970) quadratic entropy on-line, using each new sample to update the entropy estimate to obtain more accurate results in stationary situations or to track the changing entropy of a signal in nonstationary situations.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"114 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117213993","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030082
Akihito Sudou, P. Hartono, R. Saegusa, S. Hashimoto
For reconstructing a signal from sampling data, the method based on Shannon's sampling theorem is usually employed. The reconstruction error appears when the signal does not satisfy the Nyquist condition. This paper proposes a new reconstruction method by using a linear perceptron and multilayer perceptron as FIR filter. The perceptron, which has weights obtained by learning when adapting the original signal, suppresses the difference between the reconstructed signal and the original signal even when the Nyquist condition does not stand. Although the proposed method needs weight data, the total data size is much smaller than the ordinary sampling method, as the most suitable reconstruction filter is exclusively adapted to the given sampling data.
{"title":"Signal reconstruction from sampled data using neural network","authors":"Akihito Sudou, P. Hartono, R. Saegusa, S. Hashimoto","doi":"10.1109/NNSP.2002.1030082","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030082","url":null,"abstract":"For reconstructing a signal from sampling data, the method based on Shannon's sampling theorem is usually employed. The reconstruction error appears when the signal does not satisfy the Nyquist condition. This paper proposes a new reconstruction method by using a linear perceptron and multilayer perceptron as FIR filter. The perceptron, which has weights obtained by learning when adapting the original signal, suppresses the difference between the reconstructed signal and the original signal even when the Nyquist condition does not stand. Although the proposed method needs weight data, the total data size is much smaller than the ordinary sampling method, as the most suitable reconstruction filter is exclusively adapted to the given sampling data.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126627013","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030080
A. Flexer, G. Gruber, G. Dorffner
We report improvements on automatic continuous sleep staging using hidden Markov models (HMM). Contrary to our previous efforts, we trained the HMMs on data from single sleep labs instead of generalizing to data from diverse sleep labs. Our totally unsupervised approach detects the cornerstones of human sleep (wakefulness, deep and rem sleep) with around 80% accuracy based on data from a single EEG channel recorded at the sleep lab for which we already achieved the best results so far. Experiments with data from the worst sleep lab so far cannot be improved by training a separate model. This means that our previous problem of detecting rem sleep is not a general problem of our method but rather due to insufficient information in the data for some of the sleep labs.
{"title":"Improvements on continuous unsupervised sleep staging","authors":"A. Flexer, G. Gruber, G. Dorffner","doi":"10.1109/NNSP.2002.1030080","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030080","url":null,"abstract":"We report improvements on automatic continuous sleep staging using hidden Markov models (HMM). Contrary to our previous efforts, we trained the HMMs on data from single sleep labs instead of generalizing to data from diverse sleep labs. Our totally unsupervised approach detects the cornerstones of human sleep (wakefulness, deep and rem sleep) with around 80% accuracy based on data from a single EEG channel recorded at the sleep lab for which we already achieved the best results so far. Experiments with data from the worst sleep lab so far cannot be improved by training a separate model. This means that our previous problem of detecting rem sleep is not a general problem of our method but rather due to insufficient information in the data for some of the sleep labs.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121813795","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030050
G. Nathalie, L.B. Herve, H. Jeanny, G. Anne
We develop a method to optimize a machine-based semantic categorization of natural images according to human perception. First, the categories are determined through a psychophysical experiment. The similarity matrices obtained from the human responses are analyzed by a multidimensional scaling technique called curvilinear component analysis (CCA). The same is done with an automatic image indexing system based on similarities between the outputs of Gabor filters applied to the images. Then we show that, by using the human categorization to balance the filter outputs, the system's performance may be significantly improved.
{"title":"Towards the introduction of human perception in a natural scene classification system","authors":"G. Nathalie, L.B. Herve, H. Jeanny, G. Anne","doi":"10.1109/NNSP.2002.1030050","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030050","url":null,"abstract":"We develop a method to optimize a machine-based semantic categorization of natural images according to human perception. First, the categories are determined through a psychophysical experiment. The similarity matrices obtained from the human responses are analyzed by a multidimensional scaling technique called curvilinear component analysis (CCA). The same is done with an automatic image indexing system based on similarities between the outputs of Gabor filters applied to the images. Then we show that, by using the human categorization to balance the filter outputs, the system's performance may be significantly improved.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122043097","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030051
R. V. Andreão, B. Dorizzi, P. C. Cortez, J. Mota
In this article, we explore the use of a unique type of wavelets for ECG beat detection and classification. Once the different beats are segmented, classification is performed using at the input of a neural network different wavelet scales. This improves the noise resistance and allows a better representation of the different morphologies. The results, evaluated on the MIT/BIH database, are excellent (97.69% on the normal and PVC classes) thanks to the use of a regularization technique.
{"title":"Efficient ECG multi-level wavelet classification through neural network dimensionality reduction","authors":"R. V. Andreão, B. Dorizzi, P. C. Cortez, J. Mota","doi":"10.1109/NNSP.2002.1030051","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030051","url":null,"abstract":"In this article, we explore the use of a unique type of wavelets for ECG beat detection and classification. Once the different beats are segmented, classification is performed using at the input of a neural network different wavelet scales. This improves the noise resistance and allows a better representation of the different morphologies. The results, evaluated on the MIT/BIH database, are excellent (97.69% on the normal and PVC classes) thanks to the use of a regularization technique.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124270223","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030068
S. Chitroub, A. Houacine, B. Sansal
It has been shown (see Chitroub, S. et al., Signal Processing, vol.82, no.1, p.69-92, 2002) that the model for POLSAR (polarimetric synthetic aperture radar) images is a mixture model that results from the product of two distributions, one characterizes the target response and the other characterizes the speckle phenomenon. For scene interpretation purpose, it is desirable to separate between the target response and the speckle information. We propose here to use some unsupervised learning rules for POLSAR images analysis via a PCA-ICA neural network model. Based on its rigorous statistical formulation (see Chitroub et al., Intelligent Data Analysis International Journal, vol.6, no.2, 2002), a neuronal PCA approach for the simultaneous diagonalization of the signal and noise covariance matrices is proposed. The goal is to provide PC images that are uncorrelated and have an improved SNR. Speckle is a non-Gaussian multiplicative noise, and the higher order statistics contain additional information about it. ICA is used to separate the speckle from the PC images and providing new IC images that have an improved contrast. The method has been applied on real POLSAR images. The extracted features are quite effective for scene interpretation.
它已经被证明(见Chitroub, S. et al.， Signal Processing, vol.82, no. 6)。(1, p.69-92, 2002)， POLSAR(偏振合成孔径雷达)图像模型是两种分布的乘积的混合模型，其中一种分布表征目标响应，另一种表征散斑现象。为了场景解释的目的，最好将目标响应和散斑信息分开。我们在此建议使用一些无监督学习规则，通过PCA-ICA神经网络模型对POLSAR图像进行分析。基于其严格的统计公式(参见Chitroub et al.， Intelligent Data Analysis International Journal, vol.6, no. 6)。(2, 2002)，提出了一种同时对角化信号和噪声协方差矩阵的神经主成分分析方法。目标是提供不相关的PC图像，并具有改进的信噪比。散斑是一种非高斯乘性噪声，高阶统计量包含了关于它的附加信息。ICA用于从PC图像中分离散斑，并提供具有改进对比度的新IC图像。该方法已在真实的POLSAR图像上得到应用。提取的特征对场景解释非常有效。
{"title":"Unsupervised learning rules for POLSAR images analysis","authors":"S. Chitroub, A. Houacine, B. Sansal","doi":"10.1109/NNSP.2002.1030068","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030068","url":null,"abstract":"It has been shown (see Chitroub, S. et al., Signal Processing, vol.82, no.1, p.69-92, 2002) that the model for POLSAR (polarimetric synthetic aperture radar) images is a mixture model that results from the product of two distributions, one characterizes the target response and the other characterizes the speckle phenomenon. For scene interpretation purpose, it is desirable to separate between the target response and the speckle information. We propose here to use some unsupervised learning rules for POLSAR images analysis via a PCA-ICA neural network model. Based on its rigorous statistical formulation (see Chitroub et al., Intelligent Data Analysis International Journal, vol.6, no.2, 2002), a neuronal PCA approach for the simultaneous diagonalization of the signal and noise covariance matrices is proposed. The goal is to provide PC images that are uncorrelated and have an improved SNR. Speckle is a non-Gaussian multiplicative noise, and the higher order statistics contain additional information about it. ICA is used to separate the speckle from the PC images and providing new IC images that have an improved contrast. The method has been applied on real POLSAR images. The extracted features are quite effective for scene interpretation.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128650248","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030063
A. Cichocki, Tomasz M. Rutkowski, K. Siwek
Blind sources separation, independent component analysis (ICA) and related methods are promising approaches for analysis of biomedical signals, especially for EEG/MEG and fMRI data. However, most of the methods extract all sources simultaneously, so it is time consuming and not reliable especially, when the number of sensors is large (more than 100 sensors) and signals are contaminated by huge noise. The main objective of this paper is to present a new method for extraction of specific source signals using bandpass filters approach. Such a method allows us to extract source signals with specific stochastic properties, e.g., extraction of narrow band sources with specific frequency bandwidth.
{"title":"Blind signal extraction of signals with specified frequency band","authors":"A. Cichocki, Tomasz M. Rutkowski, K. Siwek","doi":"10.1109/NNSP.2002.1030063","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030063","url":null,"abstract":"Blind sources separation, independent component analysis (ICA) and related methods are promising approaches for analysis of biomedical signals, especially for EEG/MEG and fMRI data. However, most of the methods extract all sources simultaneously, so it is time consuming and not reliable especially, when the number of sensors is large (more than 100 sensors) and signals are contaminated by huge noise. The main objective of this paper is to present a new method for extraction of specific source signals using bandpass filters approach. Such a method allows us to extract source signals with specific stochastic properties, e.g., extraction of narrow band sources with specific frequency bandwidth.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123789368","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030034
Vlad Popovici, J. Thiran
The autocorrelations have been previously used as features for 1D or 2D signal classification in a wide range of applications, like texture classification, face detection and recognition, EEG signal classification, and so on. However, in almost all the cases, the high computational costs have hampered the extension to higher orders (more than the second order). We present a method which avoids the computation of the autocorrelation coefficients and which can be applied to a large set of tools commonly used in statistical pattern recognition. We discuss different scenarios of using the autocorrelations and we show that the order of autocorrelations is no longer an obstacle.
{"title":"Pattern recognition using higher-order local autocorrelation coefficients","authors":"Vlad Popovici, J. Thiran","doi":"10.1109/NNSP.2002.1030034","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030034","url":null,"abstract":"The autocorrelations have been previously used as features for 1D or 2D signal classification in a wide range of applications, like texture classification, face detection and recognition, EEG signal classification, and so on. However, in almost all the cases, the high computational costs have hampered the extension to higher orders (more than the second order). We present a method which avoids the computation of the autocorrelation coefficients and which can be applied to a large set of tools commonly used in statistical pattern recognition. We discuss different scenarios of using the autocorrelations and we show that the order of autocorrelations is no longer an obstacle.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130709269","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030057
Hossein Roufarshbaf, H. Amindavar
We describe two real-time classifiers of unknown finite point QAM constellations over an nonideal channel. In the proposed schemes, first the transmitted symbols are recovered over a band-limited channel using the inherent cyclostationary characteristics of QAM signals. After equalization, the constellation is determined in the face of an unknown rotation due to an equalizer using a clustering approach, or Zernike moments. These methods are found to be effective in nonminimum. phase channels since they use the cyclostationary characteristics of the input signals to mitigate the destructive nature of the channel. The performance of the new classifiers are shown for high bit rate high density QAM constellations in presence of AWGN.
{"title":"High-speed voiceband QAM constellation classification in multipath environment","authors":"Hossein Roufarshbaf, H. Amindavar","doi":"10.1109/NNSP.2002.1030057","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030057","url":null,"abstract":"We describe two real-time classifiers of unknown finite point QAM constellations over an nonideal channel. In the proposed schemes, first the transmitted symbols are recovered over a band-limited channel using the inherent cyclostationary characteristics of QAM signals. After equalization, the constellation is determined in the face of an unknown rotation due to an equalizer using a clustering approach, or Zernike moments. These methods are found to be effective in nonminimum. phase channels since they use the cyclostationary characteristics of the input signals to mitigate the destructive nature of the channel. The performance of the new classifiers are shown for high bit rate high density QAM constellations in presence of AWGN.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"295 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133071808","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 : 2002-11-07DOI: 10.1109/NNSP.2002.1030045
Yankun Zhang, Chong-qing Liu
Kernel principal component analysis (KPCA) as a powerful nonlinear feature extraction method has proven as a preprocessing step for classification algorithm. A face recognition approach based on KPCA and genetic algorithms (GAs) is proposed. By the use of the polynomial functions as a kernel function in KPCA, the high order relationships can be utilized and the nonlinear principal components can be obtained. After we obtain the nonlinear principal components, we use GAs to select the optimal feature set for classification. At the recognition stage, we employed linear support vector machines (SVM) as classifier for the recognition tasks. Two face databases were used to test our algorithm and higher recognition rates were obtained which show that our algorithm is effective.
{"title":"Face recognition using kernel principal component analysis and genetic algorithms","authors":"Yankun Zhang, Chong-qing Liu","doi":"10.1109/NNSP.2002.1030045","DOIUrl":"https://doi.org/10.1109/NNSP.2002.1030045","url":null,"abstract":"Kernel principal component analysis (KPCA) as a powerful nonlinear feature extraction method has proven as a preprocessing step for classification algorithm. A face recognition approach based on KPCA and genetic algorithms (GAs) is proposed. By the use of the polynomial functions as a kernel function in KPCA, the high order relationships can be utilized and the nonlinear principal components can be obtained. After we obtain the nonlinear principal components, we use GAs to select the optimal feature set for classification. At the recognition stage, we employed linear support vector machines (SVM) as classifier for the recognition tasks. Two face databases were used to test our algorithm and higher recognition rates were obtained which show that our algorithm is effective.","PeriodicalId":117945,"journal":{"name":"Proceedings of the 12th IEEE Workshop on Neural Networks for Signal Processing","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116123777","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：