The information amount obtained through sampling measurement of a dynamic signal is not merely the sum of information amount of each sampling point, because there is a mutual dependency among each point. In this paper, the information obtained per unit time (
I0) is shown as a function of the signal power spectrum (
Ω), the sampling frequency
(fs) and the root mean square of measurement errors (σ
n), which is shown in eq. (15), under the assumptions that
(1) the signal is Gaussian,
(2) the measurement errors have a Gaussian distribution and are independent of the signal values. Eq. (15) is derived by calculating the entropies in frequency domain. It is also proved that in the range of
fs «signal bandwidth,
I0 increases by the order of
fs when
fs is increased and in
fs»signal bandwidth,
I0 increases in proportion to log
fs. An example is shown in Fig. 10 and compared with the case of continuous measurement.
抄録全体を表示