Pub Date : 2004-09-29DOI: 10.1109/MLSP.2004.1422993
R. Anitha, D. S. Satish, C. Sekhar
In this paper, we address the issues in classification of varying duration segments of speech using support vector machines. Commonly used methods for mapping the varying duration segments into fixed dimension patterns may lead to loss of crucial information necessary for classification. We propose a method in which the representation of a segment of speech is considered as a trajectory in a multidimensional space. A fixed dimension pattern vector derived from the outerproduct operation on the matrix representation of a multidimensional trajectory is given as input to the support vector machines. For acoustic modeling of speech segments consisting of multiple phonemes, the outerproduct operation is carried out for the trajectory matrix of each phoneme. The effectiveness of the proposed methods is demonstrated in recognition of isolated utterances of the E-set of English alphabet
{"title":"Outerproduct of trajectory matrix for acoustic modeling using support vector machines","authors":"R. Anitha, D. S. Satish, C. Sekhar","doi":"10.1109/MLSP.2004.1422993","DOIUrl":"https://doi.org/10.1109/MLSP.2004.1422993","url":null,"abstract":"In this paper, we address the issues in classification of varying duration segments of speech using support vector machines. Commonly used methods for mapping the varying duration segments into fixed dimension patterns may lead to loss of crucial information necessary for classification. We propose a method in which the representation of a segment of speech is considered as a trajectory in a multidimensional space. A fixed dimension pattern vector derived from the outerproduct operation on the matrix representation of a multidimensional trajectory is given as input to the support vector machines. For acoustic modeling of speech segments consisting of multiple phonemes, the outerproduct operation is carried out for the trajectory matrix of each phoneme. The effectiveness of the proposed methods is demonstrated in recognition of isolated utterances of the E-set of English alphabet","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"29 1","pages":"355-363"},"PeriodicalIF":0.0,"publicationDate":"2004-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86963919","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 : 2004-09-29DOI: 10.1109/MLSP.2004.1422961
Deniz Erdoğmuş, R. Jenssen, Y. Rao, J. Príncipe
Multivariate density estimation is an important problem that is frequently encountered in statistical learning and signal processing. One of the most popular techniques is Parzen windowing, also referred to as kernel density estimation. Gaussianization is a procedure that allows one to estimate multivariate densities efficiently from the marginal densities of the individual random variables. In this paper, we present an optimal density estimation scheme that combines the desirable properties of Parzen windowing and Gaussianization, using minimum Kullback-Leibler divergence as the optimality criterion for selecting the kernel size in the Parzen windowing step. The performance of the estimate is illustrated in a classifier design example
{"title":"Multivariate density estimation with optimal marginal parzen density estimation and gaussianization","authors":"Deniz Erdoğmuş, R. Jenssen, Y. Rao, J. Príncipe","doi":"10.1109/MLSP.2004.1422961","DOIUrl":"https://doi.org/10.1109/MLSP.2004.1422961","url":null,"abstract":"Multivariate density estimation is an important problem that is frequently encountered in statistical learning and signal processing. One of the most popular techniques is Parzen windowing, also referred to as kernel density estimation. Gaussianization is a procedure that allows one to estimate multivariate densities efficiently from the marginal densities of the individual random variables. In this paper, we present an optimal density estimation scheme that combines the desirable properties of Parzen windowing and Gaussianization, using minimum Kullback-Leibler divergence as the optimality criterion for selecting the kernel size in the Parzen windowing step. The performance of the estimate is illustrated in a classifier design example","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"47 1","pages":"73-82"},"PeriodicalIF":0.0,"publicationDate":"2004-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82883034","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 : 2004-09-29DOI: 10.1109/MLSP.2004.1423029
Serhiy Koisnov, S. Marchand-Maillet
This paper presents a hierarchical ensemble learning method applied in the context of multimedia autoannotation. In contrast to the standard multiple-category classification setting that assumes independent, non-overlapping and exhaustive set of categories, the proposed approach models explicitly the hierarchical relationships among target classes and estimates their relevance to a query as a trade-off between the goodness of fit to a given category description and its inherent uncertainty. The promising results of the empirical evaluation confirm the viability of the proposed approach, validated in comparison to several techniques of ensemble learning, as well as with different type of baseline classifiers
{"title":"Hierarchical ensemble learning for multimedia categorization and autoannotation","authors":"Serhiy Koisnov, S. Marchand-Maillet","doi":"10.1109/MLSP.2004.1423029","DOIUrl":"https://doi.org/10.1109/MLSP.2004.1423029","url":null,"abstract":"This paper presents a hierarchical ensemble learning method applied in the context of multimedia autoannotation. In contrast to the standard multiple-category classification setting that assumes independent, non-overlapping and exhaustive set of categories, the proposed approach models explicitly the hierarchical relationships among target classes and estimates their relevance to a query as a trade-off between the goodness of fit to a given category description and its inherent uncertainty. The promising results of the empirical evaluation confirm the viability of the proposed approach, validated in comparison to several techniques of ensemble learning, as well as with different type of baseline classifiers","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"95 1","pages":"645-654"},"PeriodicalIF":0.0,"publicationDate":"2004-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79835946","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 : 2004-09-29DOI: 10.1109/MLSP.2004.1422965
C. Bottura, G. L. de Oliveira Serra
In this study an approach to fuzzy neural identification of MIMO discrete-time nonlinear dynamical systems is proposed. Based on the Takagi-Sugeno (TS) fuzzy neural network, off-line and on-line schemes are formulated as a NARX (nonlinear autoregressive with exogenous input) fuzzy neural model from samples of a nonlinear dynamical system where the consequent parameters are modified by an adaptive WIV (weighted instrumental variable) algorithm based on the numerically robust orthogonal householder transformation
{"title":"Computational intelligence applied to signal processing: a proposal for fuzzy neural identification","authors":"C. Bottura, G. L. de Oliveira Serra","doi":"10.1109/MLSP.2004.1422965","DOIUrl":"https://doi.org/10.1109/MLSP.2004.1422965","url":null,"abstract":"In this study an approach to fuzzy neural identification of MIMO discrete-time nonlinear dynamical systems is proposed. Based on the Takagi-Sugeno (TS) fuzzy neural network, off-line and on-line schemes are formulated as a NARX (nonlinear autoregressive with exogenous input) fuzzy neural model from samples of a nonlinear dynamical system where the consequent parameters are modified by an adaptive WIV (weighted instrumental variable) algorithm based on the numerically robust orthogonal householder transformation","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"29 1","pages":"113-122"},"PeriodicalIF":0.0,"publicationDate":"2004-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78117766","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}
Probability of bit error expressions are derived for direct sequence CDMA (DS-CDMA) and multicarrier CDMA (MC-CDMA) with imperfect diversity combining. Pilot and data channels are transmitted through a Rayleigh fading channel with an exponential multipath intensity profile. Channel statistics are estimated using simple integrators. Then the multipath in the DS system and the multiple subcarriers in the MC system are weighted by the imperfect channel estimates and combined. Keeping the data rate, the transmit power, and the fading power constant, as the bandwidth increases, the number of multipaths increases in the DS system, and the number of subcarriers increases in the MC system. At the same time, the signal strength in each path/subcarrier decreases, and results in larger errors in the channel estimates. We show that there is a tradeoff between diversity order and SNR available for channel estimation in both DS-CDMA and MC-CDMA. Moreover, we also show that MC-CDMA performs better than DS-CDMA.
{"title":"Comparing DS-CDMA and multicarrier CDMA with imperfect channel estimation","authors":"Lucy L. Chong, Laurence B. Milstein","doi":"10.1109/SSP.2001.955303","DOIUrl":"https://doi.org/10.1109/SSP.2001.955303","url":null,"abstract":"Probability of bit error expressions are derived for direct sequence CDMA (DS-CDMA) and multicarrier CDMA (MC-CDMA) with imperfect diversity combining. Pilot and data channels are transmitted through a Rayleigh fading channel with an exponential multipath intensity profile. Channel statistics are estimated using simple integrators. Then the multipath in the DS system and the multiple subcarriers in the MC system are weighted by the imperfect channel estimates and combined. Keeping the data rate, the transmit power, and the fading power constant, as the bandwidth increases, the number of multipaths increases in the DS system, and the number of subcarriers increases in the MC system. At the same time, the signal strength in each path/subcarrier decreases, and results in larger errors in the channel estimates. We show that there is a tradeoff between diversity order and SNR available for channel estimation in both DS-CDMA and MC-CDMA. Moreover, we also show that MC-CDMA performs better than DS-CDMA.","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"8 2‐3","pages":"385-388"},"PeriodicalIF":0.0,"publicationDate":"2001-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/SSP.2001.955303","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72401526","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}
Sparse arrays are attractive for direction-of-arrival (DOA) estimation since they can provide accurate estimates at a low cost. A problem of great interest in this matter is to determine the element positions that yield the best DOA estimation performance. A major difficulty with this problem is to define a suitable performance measure to optimize. A novel criterion is proposed for optimizing element positions. The ambiguity threshold of the Weiss-Weinstein bound (1985) is used to optimize the element positions of a sparse linear array. The array obtained from the optimization is compared with some other sparse array structures that have been proposed in the literature.
{"title":"Optimization of element positions for direction finding with sparse arrays","authors":"F. Athley","doi":"10.1109/SSP.2001.955336","DOIUrl":"https://doi.org/10.1109/SSP.2001.955336","url":null,"abstract":"Sparse arrays are attractive for direction-of-arrival (DOA) estimation since they can provide accurate estimates at a low cost. A problem of great interest in this matter is to determine the element positions that yield the best DOA estimation performance. A major difficulty with this problem is to define a suitable performance measure to optimize. A novel criterion is proposed for optimizing element positions. The ambiguity threshold of the Weiss-Weinstein bound (1985) is used to optimize the element positions of a sparse linear array. The array obtained from the optimization is compared with some other sparse array structures that have been proposed in the literature.","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"14 1","pages":"516-519"},"PeriodicalIF":0.0,"publicationDate":"2001-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81942654","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 presented a method for efficiently combining channel estimation and downlink beamforming for CDMA systems, in cases where the RAKE receiver cannot be used for channel estimation. This method relies on an iterative scheme which iterates between a channel estimation scheme which is only stable when the multi-user interference is low and a beamforming operation which maximises the received signal to noise ratio. The simulation results presented show that this iterative scheme seems to converge to solutions which maximise the signal to noise plus interference ratio (SINR) which is an attractive feature since it is achieved without taking into account other user's statistics. In this new method the channel estimation and beamforming operations are iterated several times until convergence to a fixed solution.
{"title":"Combined downlink beamforming and channel estimation for high data rates CDMA systems","authors":"S. Perreau","doi":"10.1109/SSP.2001.955237","DOIUrl":"https://doi.org/10.1109/SSP.2001.955237","url":null,"abstract":"We presented a method for efficiently combining channel estimation and downlink beamforming for CDMA systems, in cases where the RAKE receiver cannot be used for channel estimation. This method relies on an iterative scheme which iterates between a channel estimation scheme which is only stable when the multi-user interference is low and a beamforming operation which maximises the received signal to noise ratio. The simulation results presented show that this iterative scheme seems to converge to solutions which maximise the signal to noise plus interference ratio (SINR) which is an attractive feature since it is achieved without taking into account other user's statistics. In this new method the channel estimation and beamforming operations are iterated several times until convergence to a fixed solution.","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"55 1","pages":"122-125"},"PeriodicalIF":0.0,"publicationDate":"2001-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81231321","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 introduce importance sampling techniques for the assessment of a class of open-loop digital phase modulation receivers with random carrier phase tracking in additive white Gaussian noise channels. We consider a symbol-by-symbol phase detector consisting of a bank of nonlinear stochastic filters tracking the random phase carrier and a decision algorithm driven by the filters' innovations. For the irreducible error floor assessment we use an importance sampling technique relying on large deviations principles that results in a multiple mode simulation density. The noisy operation of the receiver is addressed with an adaptive importance sampling technique. Simulations yield practically the same results obtained with conventional Monte Carlo with remarkable time gains.
{"title":"Importance sampling analysis of digital phase detectors with carrier phase tracking","authors":"F. Silva, J. Leitão","doi":"10.1109/SSP.2001.955219","DOIUrl":"https://doi.org/10.1109/SSP.2001.955219","url":null,"abstract":"We introduce importance sampling techniques for the assessment of a class of open-loop digital phase modulation receivers with random carrier phase tracking in additive white Gaussian noise channels. We consider a symbol-by-symbol phase detector consisting of a bank of nonlinear stochastic filters tracking the random phase carrier and a decision algorithm driven by the filters' innovations. For the irreducible error floor assessment we use an importance sampling technique relying on large deviations principles that results in a multiple mode simulation density. The noisy operation of the receiver is addressed with an adaptive importance sampling technique. Simulations yield practically the same results obtained with conventional Monte Carlo with remarkable time gains.","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"58 1","pages":"50-53"},"PeriodicalIF":0.0,"publicationDate":"2001-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78546966","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 modified constant modulus algorithm (MCMA) for adaptive equalization of the wireless indoor channel for QAM signals is presented. The algorithm minimizes an error cost function that includes both the amplitude and phase of the equalizer output. In addition to the amplitude-dependent term that is provided by the conventional constant modulus algorithm (CMA), the cost function includes a signal constellation matched error (CME) term. This term speeds up convergence and allows the equalizer to switch to decision directed (DD), or any soft-decision mode, faster than the CMA applied alone. The constellation-matched error term is constructed using polynomials with desirable properties. The MCMA is applied to a decision feedback equalizer and shown to provide improved performance over dual mode techniques.
{"title":"A modified constant modulus algorithm for adaptive channel equalization for QAM signals","authors":"M. Amin, Lin He, C. Reed, R. Malkemes","doi":"10.1109/SSP.2001.955349","DOIUrl":"https://doi.org/10.1109/SSP.2001.955349","url":null,"abstract":"A modified constant modulus algorithm (MCMA) for adaptive equalization of the wireless indoor channel for QAM signals is presented. The algorithm minimizes an error cost function that includes both the amplitude and phase of the equalizer output. In addition to the amplitude-dependent term that is provided by the conventional constant modulus algorithm (CMA), the cost function includes a signal constellation matched error (CME) term. This term speeds up convergence and allows the equalizer to switch to decision directed (DD), or any soft-decision mode, faster than the CMA applied alone. The constellation-matched error term is constructed using polynomials with desirable properties. The MCMA is applied to a decision feedback equalizer and shown to provide improved performance over dual mode techniques.","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"1 1","pages":"563-566"},"PeriodicalIF":0.0,"publicationDate":"2001-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87510807","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 novel support vector machine (SVM)-based method for appearance-based face recognition is presented. The proposed method does not use any external feature extraction process. Accordingly the intensities of the raw pixels that make up the face pattern are fed directly to the SVM. However, it takes account of prior knowledge about facial structures in the form of a kernel embedded in the SVM architecture. The new kernel efficiently explores spatial relationships among potential eye, nose, and mouth objects and is compared with existing kernels. Experiments with the ORL database show a recognition rate of 98% and speed of 0.22 seconds per face with 40 classes.
{"title":"Recognition of facial images using support vector machines","authors":"Kwang In Kim, J. Kim, K. Jung","doi":"10.1109/SSP.2001.955324","DOIUrl":"https://doi.org/10.1109/SSP.2001.955324","url":null,"abstract":"A novel support vector machine (SVM)-based method for appearance-based face recognition is presented. The proposed method does not use any external feature extraction process. Accordingly the intensities of the raw pixels that make up the face pattern are fed directly to the SVM. However, it takes account of prior knowledge about facial structures in the form of a kernel embedded in the SVM architecture. The new kernel efficiently explores spatial relationships among potential eye, nose, and mouth objects and is compared with existing kernels. Experiments with the ORL database show a recognition rate of 98% and speed of 0.22 seconds per face with 40 classes.","PeriodicalId":70952,"journal":{"name":"信号处理","volume":"64 1","pages":"468-471"},"PeriodicalIF":0.0,"publicationDate":"2001-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87607842","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：