W. M. Fisher, G. R. Doddington, and K. M. Goudie-Marshall. The DARPA speech recognition research database: specifications and status. In Proceedings of the DARPA speech recognition workshop, 1986.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....PARAMETERS The parameters of the FHMM are estimated using the EstimationMaximization algorithm [3] For further details refer to [5] and [9] 4. EXPERIMENTAL RESULTS Our experiments tested a factorial HMM system on a phoneme classification task. We used the phonetically balanced TIMIT database [4]. Training was performed on the sx and si training sentences. These create a training set with 3696 utterances from 168 different speakers. 250 sentences from the test set were used for testing. The factorial HMM had two layers and three states in each layer. The standard Lee phonetic ....

W. Fisher, G. Doddington and K. Goudie-Marshall, "The DARPA speech recognition research database: Specifications and status", Proceedings of the DARPA Speech Recognition Workshop, pp. 93-99, 1986.

....measure is evaluated as regards its classification ability using a 1 nearest neighbour decision rule. The possibility of rejection is not taken into account : the test speaker is always part of the set of references. 5.2 Databases For our experiments, we used TIMIT and NTIMIT databases. TIMIT [8] contains 630 speakers (438 male and 192 female) each of them having uttered 10 sentences. Two sentences have the prefix sa (sa1 and sa2) Sentences sa1 and sa2 are different, but they are the same across speakers. Three sentences have the prefix si and fivehave the prefix sx . These 8 ....

W. M. Fisher, G. R. Doddington, and K. M. Goudie-Marshall. The DARPAspeech recognition research database : specifications and status. In Proceedings of the DARPA workshop on speech recognition, pages 93--99, Feb. 1986.

....error rates M is the average number of frames for the training sentences across all speakers, and N is the average number of frames for the test sentences. The same symmetrisations are applied to f Y=X and f X=Y . 5. DATABASE AND SIGNAL ANALYSIS Weusethe first 63 speakers of TIMIT [14] and NTIMIT [15] for our experiments (19 females and 44 males) Each of them has read 10 sentences. The signal is sampled at 16 kHz, on 16 bits, on a linear amplitude scale. NTIMIT is a telephone channel version of TIMIT. Eachsentence is analysed as follows : for eachspeech token, the speech ....

William M. Fisher, George R. Doddington, and Kathleen M. Goudie-Marshall. The DARPAspeech recognition research database : specifications and status. In Proceedings of the DARPA workshop on speech recognition, pages 93--99, February 1986.

....to the speaker dependent PBMEC b evaluation in Table 5.1 are plotted in Figure 5.7. The PBMEC string accuracies, for different numbers of mixture components, are shown in Tables 5.3 and 5.4, for the speaker dependent and multi speaker modes, respectively. 5.6. 3 TIMIT The TIMIT database [Fisher et al. 1986, Lamel et al. 1986] was used to evaluate the effectiveness of MCE in the context of the same continuous phoneme recognition paradigm examined above, but for a more difficult task. The TIMIT corpus contains a total of 6300 sentences, 10 sentences spoken by each of 630 speakers from 8 major ....

Fisher, W.M., Doddington, G. and Goudie-Marshall, K.M. (1986). The DARPA speech recognition research database: Specifications and status. Proceedings of the DARPA Speech Recognition Workshop, Report No. SAIC-86/1546.

....respect to further five different test sets. alice 20 sentences from Alice , which was part of the training set (none of these test sentences from Alice were actually in the set of recorded utterances) timit 20 sentences from the TIMIT databases, a phonetically balanced set of 452 sentences [7]. comm 20 sentences from the CMU Communicator testing suite, used in a speech dialog system. festvox 20 sentences from the abstract of [1] story 20 sentences from a novel that was not part of the original database. A five point score was used, 5 being indistinguishable or nearly ....

Fisher, W., Doddington, G., and Goudie-Marshall, K. The DARPA speech recognition research database : specifications and status. In Proceedings of the DARPA workshop on speech recognition (1986), pp. 93--99.

....Fig. 1. Example demonstrating the process of converting conventional orthography to hiragana. A romaji transcription (transliteration of Japanese words into Roman letters) and an English translation are also shown. have found applications in the development of phonetically balanced sentence sets [2], 3] and in the evaluation of speech recognizers [4] The algorithm described in this paper, in fact, was used to develop a database of 10 000 phonetically balanced sentences for a Japanese speech database project [5] A. Overview of Japanese Orthography The Japanese writing system is ....

W. M. Fisher, G. R. Doddington, and K. M. Goudie-Marshall, "The DARPA speech recognition research database: Specifications and status, " in Proc. DARPA Speech Recognition Workshop, Feb. 1986, pp. 93--99.

....acoustic characteristics of the cellular environment for effective system design, the training database should accurately reflect the linguistic domain of interest. In general, training phoneme based recognizers requires a large, phonetically labeled database, such as the popular TIMIT database [1], to adequately capture the variation of continuous speech. Attractive features of the TIMIT database include multiple speakers, continuous speech, good coverage of North American standard dialects, and carefully designed breadth and depth of phonetic coverage. The collection of a similarly ....

....ffl 5 sx sentences, which were read from a list of 450 phonetically balanced sentences selected by MIT. ffl 3 si sentences, which were randomly selected by TI. 70 of the speakers are male. Most speakers are adult Caucasians. A complete description of the TIMIT database can be found in [1]. 3. CTIMIT DATABASE GENERATION A block diagram of the experimental setup used to generate the CTIMIT database is found on page 4 of the accompanying PostScript file poster.ps, and in [2] Clean speech from the training and testing portions of the TIMIT database was randomly ordered, then ....

W. M. Fisher et al. The DARPA speech recognition research database: specifications and status. In Proc. DARPA Workshop on Speech Recognition, pages 93--99, February 1986.

....across nativeproduced, Turkish, Mandarin, and German accented English. Two of the language accents studied 2 Detailed information about the TIMIT database is available through the Linguistics Data Consortium (LDC) URL address: http: www.cis.upenn.edu ldc home.html) Details can be found in Fisher et al. 1986). 3 The American English group Army Brat refers to a person who has moved frequently across the U.S. and therefore may possess less of a regional dialect. This term is derived from U.S. military personnel who are normally moved frequently across the U.S. 3 (Turkish, Mandarin) do not possess ....

Fisher, W., Doddington, G., and Goudie-Marshall, K. (1986). "The DARPA speech recognition research database: Specifications and status". In Proceedings of the DARPA Speech Recognition Workshop, pp. 93--99.

....quantization (VQPCA MS E) The multistage architecture reduces the number of distance calculations and hence the training time for VQPCA (Gray 1984) 4. 2 Dimension Reduction of Speech We used examples of the twelve monothongal vowels extracted from continuous speech drawn from the TIMIT database (Fisher and Doddington 1986). Each input vector consists of 32 DFT coefficients (spanning the frequency range 0 4kHz) timeaveraged over the central third of the utterance. We divided the data set into a training set containing 1200 vectors, a validation set containing 408 vectors and a test set containing 408 vectors. The ....

W. M. Fisher and G. R. Doddington. (1986) The DARPA speech recognition research database: specification and status. In Proceedings of the DARPA Speech Recognition Workshop, pages 93-99, Palo Alto, CA.

.... discussion can be found in Quackenbush, Barnwell, and Clements (1988) or Deller, Proakis, and Hansen (1993) For this study, sample phonetically balanced sentences where selected from a copy of the National Institute of Standards and Technology (NIST) production of the TIMIT DARPA speech database (Fisher et al. 1986; NIST, 1988) Sentences were degraded with two types of additive background noise (white Gaussian noise or noise from the cooling fan of an IBM PS 2 workstation) Noise levels were adjusted to ensure global signal to noise (SNR) ratios of 10, 20 and 30 decibels. Figure 4 summarizes objective ....

....statistical analysis of two parameter classes to understand how speech parameters vary under the Lombard effect. Though linear predictive coefficients (LPC, both predictor and PARCOR coefficients) were considered, only results from mel 2 A discussion of phoneme label definitions can be found in (Fisher, 1986; NIST, 1988) 5 SOURCE GENERATOR FRAMEWORK 8 cepstral parameters are discussed here since ICARUS is based on this parameterization method. This was done to determine the sensitivity of the parameters which characterize speech, as well as those algorithms which are used to estimate these ....

W. Fisher, G. Doddington, K. Goudie-Marshall, (1986), "The DARPA Speech Recognition Research Database: Specifications and Status," Proceedings of the 1986 DARPA Speech Recognition Workshop, pp. 93-99.

....this thesis. It describes briefly the domain, size and importance of each database. All bar the WSJCAM0 database (Section A.4) use native American English speakers, and all are continuous read speech corpora. A. 1 TIMIT The DARPA (US Defence Advanced Research Project Agency) TIMIT database [22, 54] was collected jointly by Texas Instruments, MIT and SRI. It is an acoustic phonetic corpus (i.e. the speech is phonetically transcribed and time aligned with the acoustic data) It consists of a set of calibration sentences (SA) a set of phonetically compact sentences (SX) plus a large number ....

W.M. Fisher, G.R. Doddington, and K.M. Goudie-Marshall. The DARPA Speech Recognition Research Database: Specifications and Status. In Proceedings Speech Recognition Workshop, 1986.

....measure is evaluated as regards its classification ability using a 1 nearest neighbour decision rule. The possibility of rejection is not taken into account : the test speaker is always part of the set of references. 5.2 Databases For our experiments, we used TIMIT and NTIMIT databases. TIMIT [8] contains 630 speakers (438 male and 192 female) each of them having uttered 10 sentences. Two sentences have the prefix sa (sa1 and sa2) Sentences sa1 and sa2 are different, but they are the same across speakers. Three sentences have the prefix si and five have the prefix sx . These 8 ....

W. M. Fisher, G. R. Doddington, and K. M. Goudie-Marshall. The DARPA speech recognition research database : specifications and status. In Proceedings of the DARPA workshop on speech recognition, pages 93--99, Feb. 1986.

....outperforms the polynomial kernel, and that constructing multi class classifiers from a set of one vs. one classifiers yields better performance than using a set of one vs. all classifiers. 4. TIMIT EXPERIMENTS To test the performance of SVMs on a more difficult task we used the TIMIT database [3]. Training was performed on the sx and si training sentences. These create a training set with 3696 utterances from 168 different speakers. For testing we chose the core set. It consists of 192 utterances from 24 different speakers not included in the training set. All utterances contain ....

W. Fisher, G. Doddington, and K. Goudie-Marshall. The DARPA Speech Recognition Research Database: Specifications and Status. In Proceedings DARPA Speech Recognition Workshop, pages 93--99, 1986.

....in continuous speech recognition has been the lack of sufficient speech material for the study of speech events and for training, development, and testing of algorithms and systems. A major effort in this area has been undertaken under the auspices of DARPA, with the production of the TIMIT[1, 2] and Resource Management[3] speech corpora. The availability of these corpora has enabled speech recognition systems to be evaluated on a common ground, which has stimulated research in many laboratories, both within and outside of DARPA projects. As a step in providing comparable data for the ....

W.M. Fisher, G.R. Doddington, and K.M. Goudie-Marshall, "The DARPA Speech Recognition Research Database: Specifications and Status," Proc. DARPA Speech Recog. Workshop, 1986.

....As of this writing, these include 463 responses to the what city are you calling from question, 1305 responses to the what city and state did you grow up in and 101 responses to the what is your last name question. The set of phonemic labels is similar to that used with the TIMIT corpus [2,3], with the following distribution. #h 3589 n 1923 pau 1687 ah 1434 l 1212 s 1134 ih 1051 k 956 tcl 837 ao r 763 t 750 ae 696 kcl 682 eh 668 iy 667 q 579 ow 577 m 568 d 529 dcl 484 ao 464 ax 453 er 441 b 440 uw 435 r 414 y 412 g 400 w 393 v 385 gcl 355 ls 341 p 334 ix 313 f 303 aa 274 ey 269 sh ....

Fisher, W., Doddington, G., and K. Goudie-Marshall, "The DARPA Speech Recognition Research Database: Specification and Status.", Proceedings of the DARPA Speech Recognition Workshop, pp. 93--100, 1986.

....and dynamic time warping. The available data was split into a training and test set. We used 80 of the data for training (ISOLET1 4) and 20 for tests (ISOLET5) as recommended by the originators of the dataset. TIMIT Database The DARPA (Defence Advanced Research Project Agency) TIMIT database [15, 18] is an acousticphonetic database consisting of data, that is phonetically transcribed and time aligned. TIMIT contains a total of 6300 sentences, 10 sentences spoken by each of 630 speakers from 8 major dialect regions of the United States of America. Each speaker utters 2 calibration sentences ....

W. Fisher, G. Doddington, and K. Goudie-Marshall. The DARPA speech recognition research database : Specification and status. In Proceedings Speech Recognition Workshop, 1986.

....the domain, size and importance of each database. The conditions for each of the ARPA benchmark tasks are also described. All the data is from native American English speakers. The task is continuous read speech. C. 1 TIMIT The DARPA (Defence Advanced Research Project Agency) TIMIT database (Fisher et al. 1986; Lamel et al. 1986) is an acoustic phonetic database consisting of data that is phonetically transcribed and time aligned. The text corpus design was a joint effort among the Massachusetts Institute of Technology (MIT) Stanford Research Institute (SRI) and Texas Instruments (TI) TIMIT ....

Fisher, W., Doddington, G., and Goudie-Marshall, K. (1986). The DARPA Speech Recognition Research Database: Specifications and Status. In Proceedings Speech Recognition Workshop.

No context found.

W. M. Fisher, G. R. Doddington, and K. M. Goudie-Marshall. The DARPA speech recognition research database: specifications and status. In Proceedings of the DARPA speech recognition workshop, 1986.

No context found.

W.M. Fisher, G.R. Doddington, and K.M. Goudie-Marshall, "The DARPA speech recognition research database: Specifications and status," Proceedings of the DARPA Speech Recognition Workshop, pp. 93-99, 1986.

No context found.

W. Fisher, G. Doddington, and K. Goudie-Marshall, "The DARPA speech recognition research database : specifications and status," in Proceedings of the DARPA workshop on speech recognition, 1986, pp. 93--99.

No context found.

W.M. Fisher, G.R. Doddington, and K.M. Goudie-Marshall, "The DARPA speech recognition research database: Specifications and status," Proceedings of the DARPA Speech Recognition Workshop, pp. 93-99, 1986.

No context found.

W. M. Fisher, G. R. Doddington, and K. M. GoudieMarshall, "The DARPA Speech Recognition Research Database: Specifications and Status," in Proceedings of the DARPA Speech Recognition Workshop, 1986.

No context found.

W.M. Fisher, G.R. Doddington, and K.M. Goudie-Marshall, "The DARPA Speech Recognition Research Database: Specifications and Status," Proc. DARPAWorkshop on Speech Recognition, Palo Alto, Calif., Feb. 1986, pp. 93--99.

No context found.

W.M. Fisher, G.R. Doddington, and K.M. Goudie-Marshall, "The DARPA Speech Recognition Research Database: Specifications and Status," Proc. DARPA Speech Recog. Workshop, 1986.

No context found.

Fisher, W.M.; Doddington, G.R.; Goudie-Marshall, K.M. (1986) The DARPA speech recognition research database: specifications and status. Proceedings DARPA Speech Recognition Workshop, 93-99, 1986.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC