### Table 1: Final weights obtained by the adaptive weighted myriad ltering algorithms.

"... In PAGE 18: ...Table 1: Final weights obtained by the adaptive weighted myriad ltering algorithms. Table1 shows the nal weights obtained by the two algorithms (with = 0:05 and ~ = 1:0) using 1000 samples of the noisy observed signal x(n). The nal trained lters, us- ing both the adaptive algorithms, were successful in accurately extracting the high-frequency sinusoidal component.... ..."

### Table 1: System performance for GMMs with increasing numbers of Gaussians, using a matching MAP adaptation weight a0 .

2003

Cited by 7

### Table 5: Performances of the online unsupervised adaptation of the system, for a soft adaptation, as a function of the MAP adaptation weight a6 .

2004

"... In PAGE 4: ...dded to the 2 minutes of initial enrollment, i.e. more than dou- bling the training data. The unsupervised online soft adaptation, taking into ac- count the a posteriori probability of the target for weighting the adaptation, provides very similar results (cf Table5 ). Dis- tributions of impostor scores and of true speaker scores were estimated by histograms on the NIST 2001 cellular data, lead- ing to the estimation of the probability a55 a15 a3 a32 a13a41a20 given the score a14a16a15 a13a18a17a19a3a21a20 .... ..."

Cited by 1

### Table 2: Multiplierless 32d QMF banks found by DLM-98 with static and adaptive weights. (The objective is the reconstruction error Er.)

1997

"... In PAGE 15: ... Moreover, we would like to reduce the amount of oscillations and improve convergence time. The second and third columns of Table2 show the objective-function values of the designs found and the corresponding convergence times of DLM-98 with static weights. DLM-98 does not converge when the static weight w is large.... In PAGE 18: ... In this case, an appropriate decrease of w will greatly shorten the convergence time. Table2 illustrates the improvements in convergence times using adaptive weights. For all the initial weights considered, the adaptive algorithm is able to nd converged designs in a reasonable amount of time, although the solution quality is not always consistent.... ..."

Cited by 12

### Table 4: Performances of the online unsupervised adaptation of the system along with the missed adaptation (MAd) and false adaptation (FAd) errors, as a function of the MAP adaptation weight a6 and the adaptation threshold a78 .

2004

"... In PAGE 4: ... However the system is much less sensitive to this threshold than to the hard decision threshold a78 , this is an advantage for practical use. The Detection Error Tradeoff (DET) curves for both the on- line supervised and unsupervised systems are shown along with the DET curve of the baseline system (cf Table4 ). The unsuper- vised online adaptation of the system seems to provide perfor- mance balanced along the DET curve between the baseline and the supervised adaptation setup, except for low miss detections where it gets closer of the baseline system: the unsupervised adaptation did not improve performances for the target speak- ers with the lowest detection scores.... ..."

Cited by 1

### Table 2. Absolute error rates for gender (in-)dependent (GI/GD) adaptation with 10 or 100 eigenvoices (EV) and different MAP- adaptation weights (w, a164

### Table 6. Results of adaptive weighting schemes.

### Table 2: System performance degradation with an increased number of nearest impostor segment adaptations, using a MAP adaptation weight a6a51a7a54a9a71a11 .

2004

Cited by 1

### Table 1 summarizes the parameters of a typical in- terference scenario. Fig. 5 shows both simulated and theoretical learning curve of the algorithm adapting the weights in order to cancel the interference appear- ing at t = 0. The graphs display the two performance measures of interest. The resulting RFI suppression in steady-state exceeds 20 dB, which is su cient to avoid an AD converter overload and leaves several dB mar-

"... In PAGE 4: ...update-rate 1=T: 50 kHz; 100 kHz; 500 kHz lowpass lter: characteristic: elliptic, order: 4, bandwidth: 1=(2T) Table1 : Summary of simulation parameters. gin for imperfections in implementation, especially re- garding the analog components.... ..."

### Table 5, along with results by adapting the weights (second column) and by randomly selecting a transform for a class (third column).

2000

"... In PAGE 4: ... Table5 . Performance with and without weight adaptation The results reveal that indeed an important part of the gain is due to the transforms alone.... ..."

Cited by 2