Skip to main content

Table 4 Cepstrum ξ c for simulated AR processes, where the AR model is estimated from different sequences of hallo

From: Mean square error optimal weighting for multitaper cepstrum estimation

ξ c (K,M)

M 1(4−42)

M 2(2−17)

M 3(9−20)

F 1(11−49)

F 2(9−50)

F 3(7−49)

OPT098002

0.363 (6)

0.325 (6)

0.328 (6)

0.546 (6)

0.395 (6)

0.648 (6)

OPT098004

0.338 (11)

0.294 (11)

0.299 (11)

0.520 (11)

0.372 (11)

0.649 (11)

OPT098008

0.314 (21)

0.253 (21)

0.261 (21)

0.510 (21)

0.351 (21)

0.663 (21)

OPT093002

0.313 (6)

0.214 (6)

0.220 (6)

0.705 (6)

0.399 (6)

0.995 (6)

OPT093004

0.304 (11)

0.177 (11)

0.190 (11)

0.622 (11)

0.397 (11)

1.04 (11)

OPT093008

0.301 (21)

0.175 (21)

0.189 (21)

0.615 (21)

0.385 (21)

0.974 (21)

SINopt

0.316 (6)

0.200 (8)

0.210 (8)

0.627 (4)

0.3917 (5)

0.727 (3)

THOMopt

0.328 (6)

0.212 (8)

0.217 (7)

0.671 (3)

0.428 (5)

0.771 (3)

WELCHopt

0.302 (6)

0.203 (9)

0.201 (8)

0.624 (4)

0.422 (5)

0.716 (3)

HAMM

1.65 (1)

1.65 (1)

1.64 (1)

1.70 (1)

1.67 (1)

1.71 (1)

ARopt

0.428 (19)

0.361 (12)

0.171 (13)

0.722 (48)

0.663 (27)

0.635 (45)

  1. There were six different speakers (three males and three females). The range of the model orders are noted for the different speakers. The number of multiple windows K is also given after the value of ξ c for the different methods. For the AR estimator, the estimated model order M for the minimum error is presented.