One-Dimensional Missing-Data APES via Expectation Maximization - Spectral Analysis of Signals: The Missing Data Case - page 55

2.5

2

1.5

1

0.5

0

1st Missing Sample
2nd Missing Sample
3rd Missing Sample
True Value (1st)
True Value (2nd)
True Value (3rd)

− 0.5

− 1

− 1.5

2

−

− 2.5

0 0.1 0.2 0.3 0.4 0.5

Frequency (Hz)

(a)

2.5

2

1.5

1

0.5

0

− 0.5

1st Missing Sample
2nd Missing Sample
3rd Missing Sample
True Value (1st)
True Value (2nd)
True Value (3rd)

− 1

− 1.5

− 2

− 2.5

0 0.1 0.2 0.3 0.4 0.5

Frequency (Hz)

(b)

FIGURE 5.4 : Estimation of the first three missing samples versus frequency [ N = 128 ,

σ 01, 51 (40%) samples are missing, vertical dotted lines indicate the true frequency

locations of the spectral components with the closely spaced lines indicating the continuous
spectral component]. (a) Real part and (b) imaginary part of the missing samples estimated
via MAPES-EM2.

n

= 0 .