Digital Signal Processing Reference
In-Depth Information
Collecting data from four signals (with noise and frequency change )
Steu er ung
Sign alweich e
FF T00
Fr eq uenz -K00
Y/t- G r a fi k 0 0
Rampe/Noise
FM- Signale
Ad ditio n
Netzdaten 00
Y/t- G r a fi k 0 1
Zä hler 00
Digitalins0 0
Positionss 00
Multiplikat.
Rauschanteil
Block diagram and test : prognosis of 4 noisy FM-signals
Illustration 291:
Variable training data: noisy and frequency modulated signals
Both block diagrams for “collect data” and “prognosis” are identical up to the modules on the right side.
The training data and the lower test data are superimposed and frequency modulated by a noise signal of
0,045V, so that the frequency is able to more than double. Recorded are parameters of the frequency range.
Nevertheless, the signals are recognized more reliably than in the “common case” of Illustration 289. This
significantly supports the theses from before.
The test described in Illustration 288 can be varied so that slightly noisy signals can be
used for data collection respectively training. In addition, a permanent alteration of the
base frequency
(inside a certain range)of the four signals is carried out. Strictly speaking,
the four signals in Illustration 290 additionally are frequency modulated.
Furthermore,
parameters
of the
frequency range
are now being used. From a physical
point of view, here are certain advantages at hand. First, inside the frequency bands used,
the noise is less noticeable because it is distributed over the entire band, which leaves very
little for each of them. In addition, the spectral lines of the 4 signals quasi represent
singularities
, which are easy to recognize and evaluate. In spite of significantly more
difficult recognition characteristics, the neuronal network recognizes the signals better
than before (noise voltage 0,045V).
