Results 1 - 10 of 53,575

Table 1. Three approaches to integrating out-of-handset (OOH) rejection into the speaker verification system.

in Divergence-Based Out-Of-Class Rejection For Telephone Handset
by Identification Chi-Leung Tsang, Chi-leung Tsang, Man-wai Mak 2002
"... In PAGE 4: ...01 Table 2. Equal error rates (in %) achieved by the baseline, cepstral mean subtraction (CMS), and the three approaches shown in Table1 .... ..."
Cited by 12

TABLE III Three different approaches to integrating out-of-handset (OOH) rejection into a speaker verification system.

in Stochastic Feature Transformation with Divergence-Based Out-of-Handset Rejection for Robust Speaker Verification
by Man-Wai Mak, Chi-leung Tsang, Sun-yuan Kung 2004
Cited by 8

Table 1: The CPU time spent to build and test the WCL-1 speaker verification system.

in Text-Independent Speaker Verification Based on Probabilistic Neural Networks
by Todor Ganchev, Nikos Fakotakis, George Kokkinakis 2002
Cited by 7

Table 1: The CPU time spent to build and test the WCL-1 speaker verification system.

in TextIndependent Speaker Verification Based on Probabilistic Neural Networks
by Todor Ganchev, Todor Ganchev, Nikos Fakotakis, George Kokkinakis 2002
Cited by 7

Table 1. Comparison of Equal error rate (EER) for speaker verification systems using wired phone speech (%).

in GMM AND KERNEL-BASED SPEAKER RECOGNITION WITH THE ISIP TOOLKIT
by Tales Imbiriba, Aldebaro Klautau, Naveen Parihar, Sridhar Raghavan, Joseph Picone

Table 1: Comparison of EER and DCF for the GMM and SVM system on the POLYCOST speaker verification task.

in Discriminative kernel classifiers in speaker recognition
by Marcel Katz, Martin Schafföner, Edin Andelic, Sven E. Krüger, Andreas Wendemuth

TABLE VIII SPEAKER VERIFICATION PERFORMANCE OF THREE SYSTEMS WITH AND WITHOUT INTERSESSION VARIABILITYCOMPENSATION. THE MLLR-SVM (ONE-SIDE) RESULTS CORRESPOND TO LAST COLUMN OF THE FORTH AND SIXTH ROWS OF TABLE VII

in Speaker recognition with session variability normalization based on MLLR adaptation transforms
by Andreas Stolcke, Senior Member, Sachin S. Kajarekar, Luciana Ferrer, Elizabeth Shriberg 1987
Cited by 2

Table 1. Equal error rates (EERs) and their relative reduction (Rel. Red.) with respective to equal-weight fusion achieved by the speaker and face verification systems using intramodal multi-sample fusion. Note that fusion takes place only within the audio and visual scores, not between them. Equal-weight+Znorm (Zero-sum+Znorm) means that equal-weight fusion (zero-sum fusion) was performed on Z-norm scores.

in Intramodal And Intermodal Fusion For Audio-Visual Biometric
by Authentication Ming-Cheung Cheung, Ming-cheung Cheung, Man-wai Mak
"... In PAGE 3: ... 4. RESULTS AND DISCUSSIONS Table1 shows the results of speaker verification and face verifica- tion using different types of intramodal multi-sample fusion tech- niques described in Section 2, and Fig. 1 plots the correspond- ing DET curves.... In PAGE 4: ...Table 2. EERs and relative error reduction with respect to the EER of speaker verification ( Table1 ) obtained by linearly combining the means of intramodal fused scores. The combination weight fl in Eq.... ..."

Table 1 summarizes the FRRs, FARs and verification time obtained by the SV systems using different normalization methods. All the results are based on the average of 106 male speakers.

in A Two-Stage Scoring Method Combining World And Cohort Models For Speaker Verification
by W. D. Zhang, M. W. Mak, M. X. He 2000
"... In PAGE 3: ... Table1 : Average error rates (in %) obtained and theoretical verification time taken by different normalization methods. Table 1 shows that the FAR of the speaker-world model is much smaller than the speaker-cohort model, suggesting that the former is more capable of discriminating the speaker from the general... In PAGE 3: ...Table 1: Average error rates (in %) obtained and theoretical verification time taken by different normalization methods. Table1 shows that the FAR of the speaker-world model is much smaller than the speaker-cohort model, suggesting that the former is more capable of discriminating the speaker from the general... In PAGE 4: ...applications, it is important to reduce FAR to make the SV system more robust to impostor attacks. The third and fourth rows of Table1 show the result of the two- stage decision-making approach. The results show that using the two-stage approach (with a=0 and b=0.... In PAGE 4: ... 2(c). Table1 also shows the theoretical verification time required by different scoring methods. If we combine the world and cohort models in a one-stage approach (as in [5]), the time taken will be the sum of the time required by the two models, i.... ..."
Cited by 4

TABLE IV SPEAKER VERIFICATION RESULTS USING BASELINE SYSTEM AND MLLR-SVM BASED ON 2+2 TRANSFORMS FROM FIRST RECOGNITION PASS. THE TOP VALUE IN EACH CELL IS THE EER, BELOW IT, THE MINUMUM DCF VALUE APPEARS IN NORMAL FONT. FOR SRE-05, THE ACTUAL DCF VALUES USING THRESHOLDS OPTIMIZED ON SRE-04 ARE SHOWN IN boldface

in Speaker recognition with session variability normalization based on MLLR adaptation transforms
by Andreas Stolcke, Senior Member, Sachin S. Kajarekar, Luciana Ferrer, Elizabeth Shriberg 1987
Cited by 2
Results 1 - 10 of 53,575