Pub Date : 1995-08-13DOI: 10.1109/MWSCAS.1995.504497
M.C.M. Teixeira, D. Lamas
The Backpropagation Algorithm (BA) is the standard method for training multilayer Artificial Neural Networks (ANN), although it converges very slowly and can stop in a local minimum. We present a new method for neural network training using the BA inspired on constuctivism, an alphabetization method proposed by Emilia Ferreiro (1985) based on Piaget philosophy. Simulation results show that the proposed configuration usually obtained a lower final mean square error, when compared with the standard BA and with the BA with momentum factor.
{"title":"Neural network training using the constructivism paradigms","authors":"M.C.M. Teixeira, D. Lamas","doi":"10.1109/MWSCAS.1995.504497","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.504497","url":null,"abstract":"The Backpropagation Algorithm (BA) is the standard method for training multilayer Artificial Neural Networks (ANN), although it converges very slowly and can stop in a local minimum. We present a new method for neural network training using the BA inspired on constuctivism, an alphabetization method proposed by Emilia Ferreiro (1985) based on Piaget philosophy. Simulation results show that the proposed configuration usually obtained a lower final mean square error, when compared with the standard BA and with the BA with momentum factor.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115400625","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.504495
M. Raus, L. Kreft
For evaluating parking rates without the usage of tickets the reading of car license plates from video image sequences taking from the fronts of incoming cars was implemented under use of artificial neural networks (ANNs). The detection of the number plate area and the classification of the classically separated character regions were implemented by different ANN approaches with convincing results.
{"title":"Reading car license plates by the use of artificial neural networks","authors":"M. Raus, L. Kreft","doi":"10.1109/MWSCAS.1995.504495","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.504495","url":null,"abstract":"For evaluating parking rates without the usage of tickets the reading of car license plates from video image sequences taking from the fronts of incoming cars was implemented under use of artificial neural networks (ANNs). The detection of the number plate area and the classification of the classically separated character regions were implemented by different ANN approaches with convincing results.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"70 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115663175","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.504391
Zhongde Wang, G. Jullien, W. Miller
There are two schemes of Chebyshev interpolation. Neagoe (1990) recently developed an approach, using the existing DCT algorithms, for computing the Chebyshev coefficients for one of the two schemes, but no algorithms have been developed for computing the interpolated samples. In this paper we first demonstrate that both schemes of Chebyshev interpolation relate to the type I and II discrete cosine transforms (DCT-I and DCT-II), respectively. Then we show that both schemes of Chebyshev interpolation can be computed using the existing fast algorithms for the DCT.
{"title":"Fast computation of Chebyshev optimal nonuniform interpolation","authors":"Zhongde Wang, G. Jullien, W. Miller","doi":"10.1109/MWSCAS.1995.504391","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.504391","url":null,"abstract":"There are two schemes of Chebyshev interpolation. Neagoe (1990) recently developed an approach, using the existing DCT algorithms, for computing the Chebyshev coefficients for one of the two schemes, but no algorithms have been developed for computing the interpolated samples. In this paper we first demonstrate that both schemes of Chebyshev interpolation relate to the type I and II discrete cosine transforms (DCT-I and DCT-II), respectively. Then we show that both schemes of Chebyshev interpolation can be computed using the existing fast algorithms for the DCT.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115733702","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.510278
L. Tang, Thomas S. Huang
This paper describes an approach to face recognition using synthesized intermediate views which are not in the training database but come in the test data set. A 3D face modeling procedure is used to create these synthesized views. This method is evaluated by correlating all synthesized views with real intermediate views.
{"title":"Face recognition using synthesized intermediate views","authors":"L. Tang, Thomas S. Huang","doi":"10.1109/MWSCAS.1995.510278","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.510278","url":null,"abstract":"This paper describes an approach to face recognition using synthesized intermediate views which are not in the training database but come in the test data set. A 3D face modeling procedure is used to create these synthesized views. This method is evaluated by correlating all synthesized views with real intermediate views.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121100164","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.510327
B. Wess, W. Kreuzer
The problem of translating graphs into optimal instruction code is NP-complete even for very simple target machine models. On the other hand, O(n) algorithms are available for translating expression trees into optimal instruction code. The purpose of this paper is to show that by transforming graphs into constrained expression trees, global optimized instruction code can be produced based on tree algorithms.
{"title":"Optimized DSP assembly code generation starting from homogeneous atomic data flow graphs","authors":"B. Wess, W. Kreuzer","doi":"10.1109/MWSCAS.1995.510327","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.510327","url":null,"abstract":"The problem of translating graphs into optimal instruction code is NP-complete even for very simple target machine models. On the other hand, O(n) algorithms are available for translating expression trees into optimal instruction code. The purpose of this paper is to show that by transforming graphs into constrained expression trees, global optimized instruction code can be produced based on tree algorithms.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127499142","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.510214
Yih-Fang Huang, S. Gollamudi
This paper proposes employing set-membership identification for adaptive equalization. A novel feature of the set-membership identification (SMI) is selective update of the estimates for the channel parameters. This is in sharp contrast with conventional recursive schemes such as recursive least-squares (RLS) which updates continually regardless of the benefit of updates. Simulation results show that the SMI algorithm uses less than 20% of the data for parameter updates in the training mode and less than 10% of the data in the decision-directed mode, without much performance degradation in terms of bit error rate.
{"title":"Set-membership identification for adaptive equalization","authors":"Yih-Fang Huang, S. Gollamudi","doi":"10.1109/MWSCAS.1995.510214","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.510214","url":null,"abstract":"This paper proposes employing set-membership identification for adaptive equalization. A novel feature of the set-membership identification (SMI) is selective update of the estimates for the channel parameters. This is in sharp contrast with conventional recursive schemes such as recursive least-squares (RLS) which updates continually regardless of the benefit of updates. Simulation results show that the SMI algorithm uses less than 20% of the data for parameter updates in the training mode and less than 10% of the data in the decision-directed mode, without much performance degradation in terms of bit error rate.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125153447","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.504510
D. Bhattacharya, A. Antoniou
A Hopfield-type neural network is proposed for the design of equiripple FIR digital filters. A weighted least-squares error function is minimized in an iterative fashion and weights are updated at the end of each iteration until the desired accuracy is achieved. The network is simulated and an example is included to show that this is an efficient way of solving the approximation problem and has a high potential for implementation in analog VLSI.
{"title":"Design of equiripple linear phase FIR filters by feedback neural networks","authors":"D. Bhattacharya, A. Antoniou","doi":"10.1109/MWSCAS.1995.504510","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.504510","url":null,"abstract":"A Hopfield-type neural network is proposed for the design of equiripple FIR digital filters. A weighted least-squares error function is minimized in an iterative fashion and weights are updated at the end of each iteration until the desired accuracy is achieved. The network is simulated and an example is included to show that this is an efficient way of solving the approximation problem and has a high potential for implementation in analog VLSI.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125191850","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.510266
J. Silva-Martínez, J.S. Suner
In this paper two preamplifiers for capacitive sources are proposed. The topologies present very low distortion components (THD<1%), low offset voltages (less than 40 mvolts at the output) and low noise (less than 30 /spl mu/v/sub RMS/) The dynamic range of the topologies, is 57 dB and 63 dB, respectively. The power consumption is 21 /spl mu/watts for the first topology and 10 /spl mu/watts for the second one. The supply voltages employed are /spl plusmn/1.5 volts.
{"title":"Low frequency amplifiers for electret microphones","authors":"J. Silva-Martínez, J.S. Suner","doi":"10.1109/MWSCAS.1995.510266","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.510266","url":null,"abstract":"In this paper two preamplifiers for capacitive sources are proposed. The topologies present very low distortion components (THD<1%), low offset voltages (less than 40 mvolts at the output) and low noise (less than 30 /spl mu/v/sub RMS/) The dynamic range of the topologies, is 57 dB and 63 dB, respectively. The power consumption is 21 /spl mu/watts for the first topology and 10 /spl mu/watts for the second one. The supply voltages employed are /spl plusmn/1.5 volts.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126192096","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.510282
L. D. de Oliveira, J. Rossi, F.A. de C. Pires
The performance of the three-phase core type transformers, under AC/DC double excitation is discussed in this work. It is presented a mathematical model that considers the mutual coupling between phases and the magnetic nonlinearity. The validity of the proposed model is verified by means of the experimental and simulated results.
{"title":"Asymmetrical magnetization in three-phase core type transformers","authors":"L. D. de Oliveira, J. Rossi, F.A. de C. Pires","doi":"10.1109/MWSCAS.1995.510282","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.510282","url":null,"abstract":"The performance of the three-phase core type transformers, under AC/DC double excitation is discussed in this work. It is presented a mathematical model that considers the mutual coupling between phases and the magnetic nonlinearity. The validity of the proposed model is verified by means of the experimental and simulated results.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123555635","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 : 1995-08-13DOI: 10.1109/MWSCAS.1995.504401
O. J. Tobias, R. Seara, F.A.P. Soares, J.C.M. Bermudez
This paper describes an algorithm which uses machine-vision techniques to perform automatic visual inspection tasks in the industry of ceramic tiles. This algorithm uses difference statistics to analyze images of ceramic tiles from the point of view of their texture contents.
{"title":"Automatic visual inspection using the co-occurrence approach","authors":"O. J. Tobias, R. Seara, F.A.P. Soares, J.C.M. Bermudez","doi":"10.1109/MWSCAS.1995.504401","DOIUrl":"https://doi.org/10.1109/MWSCAS.1995.504401","url":null,"abstract":"This paper describes an algorithm which uses machine-vision techniques to perform automatic visual inspection tasks in the industry of ceramic tiles. This algorithm uses difference statistics to analyze images of ceramic tiles from the point of view of their texture contents.","PeriodicalId":165081,"journal":{"name":"38th Midwest Symposium on Circuits and Systems. Proceedings","volume":"70 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1995-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123708250","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：