Pub Date : 2002-08-26DOI: 10.1109/ICOSP.2002.1180097
T. Shan, R. Tao, Yue Wang, Siyong Zhou
Cell average clutter map CFAR (CACM-CFAR) is analyzed at first. CACM-CFAR has better performance in the inhomogeneous clutter environment, but is affected by interference and self-masking which will cause the low probability of detection. To solve these problems, a novel algorithm named order statistic clutter map CFAR (OSCM-CFAR) is proposed, in which the clutter map is updated by the k-th ranked cell in the clutter map cell. The performance analysis shows that the probability of detection for slow moving targets is increased by using the OSCM-CFAR algorithm. And the same case occurs in the multi-target environment when the new method is used.
{"title":"Performance of order statistic clutter map CFAR","authors":"T. Shan, R. Tao, Yue Wang, Siyong Zhou","doi":"10.1109/ICOSP.2002.1180097","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1180097","url":null,"abstract":"Cell average clutter map CFAR (CACM-CFAR) is analyzed at first. CACM-CFAR has better performance in the inhomogeneous clutter environment, but is affected by interference and self-masking which will cause the low probability of detection. To solve these problems, a novel algorithm named order statistic clutter map CFAR (OSCM-CFAR) is proposed, in which the clutter map is updated by the k-th ranked cell in the clutter map cell. The performance analysis shows that the probability of detection for slow moving targets is increased by using the OSCM-CFAR algorithm. And the same case occurs in the multi-target environment when the new method is used.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132413621","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1181096
Wei Zhang, S. Gazor
The Gaussian and Laplacian models of speech signals are investigated in this paper. We use different hypothesis tests to compare these two models. The Gaussian model has been widely used while our experimental results show that the probability density functions (PDFs) of speech signals are more like the Laplacian distributions. Based on the fact that the KLT and DCT have been excessively used in speech signal processing, the distribution of speech components in both decorrelated domains are also investigated. All the results illustrate that the speech signals follow Laplacian distributions both in the time domain speech samples and in the KLT or DCT (excluding DC) domains. The distribution of speech signals in uncorrelated domains can be assumed as a multivariate Laplacian.
{"title":"Statistical modelling of speech signals","authors":"Wei Zhang, S. Gazor","doi":"10.1109/ICOSP.2002.1181096","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1181096","url":null,"abstract":"The Gaussian and Laplacian models of speech signals are investigated in this paper. We use different hypothesis tests to compare these two models. The Gaussian model has been widely used while our experimental results show that the probability density functions (PDFs) of speech signals are more like the Laplacian distributions. Based on the fact that the KLT and DCT have been excessively used in speech signal processing, the distribution of speech components in both decorrelated domains are also investigated. All the results illustrate that the speech signals follow Laplacian distributions both in the time domain speech samples and in the KLT or DCT (excluding DC) domains. The distribution of speech signals in uncorrelated domains can be assumed as a multivariate Laplacian.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133930848","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1181093
W. Lo, Pui-Fung Wong, M. Siu
A Chinese sentence is typically written as a sequence of characters. However, a word is a logical semantic and syntactic unit. Thus, a segmentation algorithm is necessary. to map the sequence of characters into a sequence of words. Forward maximum matching, which tries to find the longest words to match the characters in the sentence, is one of the most popular methods because of its simplicity and efficiency. However, because it makes decisions by finding the longest next word without regard to the whole sentence, it is not optimal. In this paper, we proposed two new segmentation algorithms: the dynamic matching algorithm and maximum likelihood segmentation algorithm. In the dynamic matching algorithm, dynamic programming is used to look for the best segmentation (longest average word length) for the whole sentence. In the maximum likelihood algorithm, we aim at obtaining the likely word segmentation given a particular language model. Because of ML, this algorithm also guarantees to give the best perplexity across different segmentations. While both algorithms yield limited gains in terms of perplexity reduction, both give significant reduction in recognition error on the 863 corpus.
{"title":"Maximum likelihood algorithm on Chinese word segmentation","authors":"W. Lo, Pui-Fung Wong, M. Siu","doi":"10.1109/ICOSP.2002.1181093","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1181093","url":null,"abstract":"A Chinese sentence is typically written as a sequence of characters. However, a word is a logical semantic and syntactic unit. Thus, a segmentation algorithm is necessary. to map the sequence of characters into a sequence of words. Forward maximum matching, which tries to find the longest words to match the characters in the sentence, is one of the most popular methods because of its simplicity and efficiency. However, because it makes decisions by finding the longest next word without regard to the whole sentence, it is not optimal. In this paper, we proposed two new segmentation algorithms: the dynamic matching algorithm and maximum likelihood segmentation algorithm. In the dynamic matching algorithm, dynamic programming is used to look for the best segmentation (longest average word length) for the whole sentence. In the maximum likelihood algorithm, we aim at obtaining the likely word segmentation given a particular language model. Because of ML, this algorithm also guarantees to give the best perplexity across different segmentations. While both algorithms yield limited gains in terms of perplexity reduction, both give significant reduction in recognition error on the 863 corpus.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133223707","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1181109
Yuanyuan Shi, Jia Liu, Runsheng Liu
The conventional hidden Markov model (HMM) only based on the spectral features does not have a high recognition performance for connected Mandarin digits, because highly confusable syllables exist. The main problems of Mandarin digit recognition are analyzed. It is revealed that to establish the precise classification models for Mandarin digits not only features extracted from the spectrum, energy and pitch contour are necessary but also they should be used with different emphases for different digits. So each-type of feature is used to train a single-stream HMM by maximum likelihood. Then a multi-stream HMM is obtained by combining the single-stream HMMs with exponents that weigh the log-likelihood of each stream. The exponents are estimated by means of the generalized probabilistic descent algorithm according to the digit minimum classification error rate criteria. The superiority of the multi-stream HMM is demonstrated: the relative string error rate is reduced by 54.5%. And the unknown length digit string error rate and its digit error rate decrease to 4.66% and 1.31% respectively.
{"title":"Discriminative HMM stream model for Mandarin digit string speech recognition","authors":"Yuanyuan Shi, Jia Liu, Runsheng Liu","doi":"10.1109/ICOSP.2002.1181109","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1181109","url":null,"abstract":"The conventional hidden Markov model (HMM) only based on the spectral features does not have a high recognition performance for connected Mandarin digits, because highly confusable syllables exist. The main problems of Mandarin digit recognition are analyzed. It is revealed that to establish the precise classification models for Mandarin digits not only features extracted from the spectrum, energy and pitch contour are necessary but also they should be used with different emphases for different digits. So each-type of feature is used to train a single-stream HMM by maximum likelihood. Then a multi-stream HMM is obtained by combining the single-stream HMMs with exponents that weigh the log-likelihood of each stream. The exponents are estimated by means of the generalized probabilistic descent algorithm according to the digit minimum classification error rate criteria. The superiority of the multi-stream HMM is demonstrated: the relative string error rate is reduced by 54.5%. And the unknown length digit string error rate and its digit error rate decrease to 4.66% and 1.31% respectively.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133342958","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1180122
Zhong Yan-ping, Song-hua Shen
By analyzing a new model for short-time complex cepstrum of voiced speech, this paper proposes a new method for homomorphic deconvolution to recover the vocal tract impulse response; The experimental results are given and the effects of pitch period on the recovery of vocal tract impulse response are discussed in the paper.
{"title":"A method of recovering vocal tract impulse response based on a new model for complex cepstrum of voiced speech","authors":"Zhong Yan-ping, Song-hua Shen","doi":"10.1109/ICOSP.2002.1180122","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1180122","url":null,"abstract":"By analyzing a new model for short-time complex cepstrum of voiced speech, this paper proposes a new method for homomorphic deconvolution to recover the vocal tract impulse response; The experimental results are given and the effects of pitch period on the recovery of vocal tract impulse response are discussed in the paper.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133926899","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1181069
A.R. Ghaderipoor, L. Beygi, S. H. Jamali
Multiple antenna systems that operate at high rates require simple yet effective space-time transmission schemes to handle the large traffic volume in real time. In the last investigations, LD (linear dispersion) codes were proposed that can handle any configuration of transmit and receive antennas for frequency non-selective fading channels. This scheme transmits substreams of data in linear combinations over space and time and the codes are designed to optimize the mutual information between the transmitted and received signals with low complexity decoding. In this paper we try to design LD codes for frequency selective fading channels. We suppose that in each signaling interval, L independent copy of transmitted signals arrive at each receiver antenna. We first focus on the structure of such a system in the presence of LD codes and extract the capacity relation of a multiple antenna system in a frequency selective fading channel. Then we exploit the capacity formula to find the best LD codes that can reach the capacity.
{"title":"Linear dispersion space-time codes for frequency selective fading channels","authors":"A.R. Ghaderipoor, L. Beygi, S. H. Jamali","doi":"10.1109/ICOSP.2002.1181069","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1181069","url":null,"abstract":"Multiple antenna systems that operate at high rates require simple yet effective space-time transmission schemes to handle the large traffic volume in real time. In the last investigations, LD (linear dispersion) codes were proposed that can handle any configuration of transmit and receive antennas for frequency non-selective fading channels. This scheme transmits substreams of data in linear combinations over space and time and the codes are designed to optimize the mutual information between the transmitted and received signals with low complexity decoding. In this paper we try to design LD codes for frequency selective fading channels. We suppose that in each signaling interval, L independent copy of transmitted signals arrive at each receiver antenna. We first focus on the structure of such a system in the presence of LD codes and extract the capacity relation of a multiple antenna system in a frequency selective fading channel. Then we exploit the capacity formula to find the best LD codes that can reach the capacity.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132893835","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1180074
Duan Pengfei, Wan Yonglian, Sun Wenfeng
The Hough transform (HT) is analyzed and applied to radar weak target detection and location in this paper. The literature, in which only search radar is used, is extended to more general radar system. The theoretical analysis and experimental results used real-life radar data show that the presented method performs well in the strong clutter background, and robust to the clutter distribution.
{"title":"Radar weak target detection and location based on Hough transform","authors":"Duan Pengfei, Wan Yonglian, Sun Wenfeng","doi":"10.1109/ICOSP.2002.1180074","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1180074","url":null,"abstract":"The Hough transform (HT) is analyzed and applied to radar weak target detection and location in this paper. The literature, in which only search radar is used, is extended to more general radar system. The theoretical analysis and experimental results used real-life radar data show that the presented method performs well in the strong clutter background, and robust to the clutter distribution.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"63 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133018845","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1180025
Pingyi Fan, Z. Cao, X. Xia
We investigate the principle of maximum ratio transmission (MRT) for flat Rayleigh fading channels. Theoretical analysis indicates that the overall signal to noise ratio gain of MRT by using the multiple antennas for both transmission and reception can be maximized by selecting the weighting vector properly. It can be considered as an expansion of that presented by Lo (IEEE Trans. Comm. vol.47, no.10, p.1458-61, 1999).
{"title":"Improved weighting vector selection method in maximum ratio transmission over flat Rayleigh fading channels","authors":"Pingyi Fan, Z. Cao, X. Xia","doi":"10.1109/ICOSP.2002.1180025","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1180025","url":null,"abstract":"We investigate the principle of maximum ratio transmission (MRT) for flat Rayleigh fading channels. Theoretical analysis indicates that the overall signal to noise ratio gain of MRT by using the multiple antennas for both transmission and reception can be maximized by selecting the weighting vector properly. It can be considered as an expansion of that presented by Lo (IEEE Trans. Comm. vol.47, no.10, p.1458-61, 1999).","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121640287","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1180013
Sheng Chen
The paper proposes to combine orthogonal least squares (OLS) model selection with local regularisation for efficient sparse kernel data modelling. By assigning each orthogonal weight in the regression model with an individual regularisation parameter, the ability for the OLS model selection to produce a very parsimonious model with excellent generalisation performance is greatly enhanced.
{"title":"Locally regularised orthogonal least squares algorithm for the construction of sparse kernel regression models","authors":"Sheng Chen","doi":"10.1109/ICOSP.2002.1180013","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1180013","url":null,"abstract":"The paper proposes to combine orthogonal least squares (OLS) model selection with local regularisation for efficient sparse kernel data modelling. By assigning each orthogonal weight in the regression model with an individual regularisation parameter, the ability for the OLS model selection to produce a very parsimonious model with excellent generalisation performance is greatly enhanced.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"144 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116411803","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 : 2002-08-26DOI: 10.1109/ICOSP.2002.1181207
R. Tjahyadi, Wanquan Liu
In this paper, we propose a method to extract features from images that may be related to their compression capability with wavelet filters - their fidelity. Based on these features, images are classified into three different classes corresponding to their fidelity: low, medium and high. We have found this classification schema is effective and can be used as a guideline for selecting wavelet filter for the images in the low fidelity class.
{"title":"Image classification for quality compression with wavelet filters based on image feature analysis","authors":"R. Tjahyadi, Wanquan Liu","doi":"10.1109/ICOSP.2002.1181207","DOIUrl":"https://doi.org/10.1109/ICOSP.2002.1181207","url":null,"abstract":"In this paper, we propose a method to extract features from images that may be related to their compression capability with wavelet filters - their fidelity. Based on these features, images are classified into three different classes corresponding to their fidelity: low, medium and high. We have found this classification schema is effective and can be used as a guideline for selecting wavelet filter for the images in the low fidelity class.","PeriodicalId":159807,"journal":{"name":"6th International Conference on Signal Processing, 2002.","volume":"331 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116528853","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：