{"title":"Robust estimation for radial basis functions","authors":"A. Bors, I. Pitas","doi":"10.1109/NNSP.1994.366058","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366058","url":null,"abstract":"This paper presents a new learning algorithm for radial basis functions (RBF) neural network, based on robust statistics. The extention of the learning vector quantizer for second order statistics is one of the classical approaches in estimating the parameters of a RBF model. The paper provides a comparative study for these two algorithms regarding their application in probability density function estimation. The theoretical bias in estimating one-dimensional Gaussian functions are derived. The efficiency of the algorithm is shown in modelling two-dimensional functions.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"93 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117254673","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}
{"title":"A monolithic speech recognizer based on fully recurrent neural networks","authors":"K. Kasper, H. Reininger, D. Wolf, H. Wust","doi":"10.1109/NNSP.1994.366033","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366033","url":null,"abstract":"Reports on investigations concerning the application of fully recurrent neural networks (FRNN) for speaker independent speech recognition. In a phoneme based recognition system separate FRNN are used for feature scoring as well as for compensating variations in time durations of speech segments. A recognizer with a FRNN for feature scoring achieves the same recognition rate as a recognition system where the context information is provided. The performance of the FRNN used for time alignment is comparable to that of a viterbi based alignment with durational constraints. Additionally, a monolithic speech recognizer is realized by FRNN which directly classifies feature sequences. The performance of this FRNN is comparable to that of speech recognition systems which are based on discrete hidden Markov models and use a sophisticated durational modeling. Furthermore, simulation experiments revealed that FRNN are able to extract relevant information for speech recognition from noise contaminated speech and thus achieve a robust recognition performance.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120938241","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}
{"title":"Neural networks and higher order spectra for breast cancer detection","authors":"T. Stathaki, A. Constantinides","doi":"10.1109/NNSP.1994.366019","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366019","url":null,"abstract":"The research work contained in this paper is concerned with the use of higher order spectral estimation techniques for the derivation of the parameters of two dimensional autoregressive (AR) models. The specific application of the developed method is in mammography, an area in which it is very difficult to discern the appropriate features. The required segmentation of such 2-D random fields is effected through the additional stage of a neural network having as inputs the extracted autoregressive parameters. The results show significant discriminating gains through such techniques. The directionality of the cumulant space has been observed to influence the AR parameter estimation and this forms another area for examination.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117084164","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}
{"title":"OSA-a topological algorithm for constructing two-layer neural networks","authors":"F. Mascioli, G. Martinelli","doi":"10.1109/NNSP.1994.366066","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366066","url":null,"abstract":"Presents a constructive training algorithm for supervised neural networks: OSA (Oil-Spot Algorithm). It builds a two-layer neural network by involving successively binary examples. Its main learning rule, based on topological theorems on the cuts of a binary hypercube, is discussed. A convenient treatment of real-valued data is possible by means of a suitable real-to-binary codification. For binary target functions that have efficient halfspace union representations, the constructed networks are optimized in terms of number of neurons with respect to other constructive algorithms, as shown.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132078244","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}
{"title":"Word recognition using a neural network and a phonetically based DTW","authors":"Y. Matsuura, H. Miyazawa, T. Skinner","doi":"10.1109/NNSP.1994.366034","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366034","url":null,"abstract":"The authors have developed a speaker-independent, isolated-word recognition system using a neural network to recognize the underlying sequence of phonemes and a dynamic time warping (DTW) technique to time-align the recognized sequence of phonemes with corresponding lexical sequences of phonemes. A significant feature of this system is the ability to easily change the vocabulary, since the lexical entries are simply derived from their phoneme sequences.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134075082","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}
{"title":"A neural network scheme for Earthquake prediction based on the seismic electric signals","authors":"S. Lakkos, A. Hadjiprocopis, R. Comley, P. Smith","doi":"10.1109/NNSP.1994.365997","DOIUrl":"https://doi.org/10.1109/NNSP.1994.365997","url":null,"abstract":"Earthquake prediction based on the seismic electric signals usually employs statistical linear models. In this paper, an alternative scheme for earthquake prediction and modelling of the geophysical characteristics based on artificial neural networks, is presented. Several network configurations are investigated and the results are discussed and interpreted in various ways.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"191 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133113301","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}
{"title":"Improving the resolution of a sensor array pattern by neural networks","authors":"C. Bracco, S. Marcos, M. Benidir","doi":"10.1109/NNSP.1994.366008","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366008","url":null,"abstract":"The sensor array pattern is the spatial response of an array of sensors to an incident monochromatic plane wave. It is known to have a resolution which is a function of the number of sensors. In some applications the number of sensors may be small for technical or economical reasons. The authors present and analyse a feedforward neural network structure which is able, by learning, to improve the resolution of the array pattern for a fixed and small number of sensors.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127784753","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}
{"title":"Mixture density estimation via EM algorithm with deterministic annealing","authors":"N. Ueda, R. Nakano","doi":"10.1109/NNSP.1994.366062","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366062","url":null,"abstract":"Presents a new approach for the problem of estimating the parameters which determine a mixture density. The approach utilizes the principle of maximum entropy and statistical mechanics analogy. The EM process which is well known as a maximum likelihood estimation method is reformulated as the minimization problem of thermodynamic free energy. Unlike stochastic relaxation or simulated annealing, the minimization is deterministically performed. Moreover, the derived algorithm, unlike the conventional EM algorithm, can estimate the parameters free of initial parameter values.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114785472","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}
{"title":"A hybrid neural network architecture for automatic object recognition","authors":"T. Fechner, R. Tanger","doi":"10.1109/NNSP.1994.366049","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366049","url":null,"abstract":"This paper describes the application of a hybrid neural network architecture for automatic object recognition in inverse synthetic aperture radar (ISAR) imagery. The architecture employs a cascaded combination of an unsupervised and a supervised trained neural network. The unsupervised trained self-organizing feature map is used for object segmentation and the supervised trained multilayer perceptron classifies the segmented objects. The classification result is fed back to the feature map segmentor in order to improve segmentation and classification. The functionality of this approach is demonstrated by the use of simulated noisy ISAR images from different objects.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"98 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114666911","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}
{"title":"Connectionist model combination for large vocabulary speech recognition","authors":"M. Hochberg, G. Cook, S. Renals, A. J. Robinson","doi":"10.1109/NNSP.1994.366040","DOIUrl":"https://doi.org/10.1109/NNSP.1994.366040","url":null,"abstract":"Reports in the statistics and neural networks literature have expounded the benefits of merging multiple models to improve classification and prediction performance. The Cambridge University connectionist speech group has developed a hybrid connectionist-hidden Markov model system for large vocabulary talker independent speech recognition. The performance of this system has been greatly enhanced through the merging of connectionist acoustic models. This paper presents and compares a number of different approaches to connectionist model merging and evaluates them on the TIMIT phone recognition and ARPA Wall Street Journal word recognition tasks.<<ETX>>","PeriodicalId":250405,"journal":{"name":"Proceedings of IEEE Workshop on Neural Networks for Signal Processing","volume":"224 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1994-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121958108","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}