Digital Signal Processing Reference
In-Depth Information
0s
1s
| Noise | Sine 100 Hz 4V | Sine 300 Hz 2V | Sine 400 Hz 3V | Sine 500 Hz 4V
Wide
Gaussian window
50,00
0s
36,75
23,50
10,25
1s
-3,00
25
50
75
100
125
150
175
200
225
ms
Narrow
Gaussian window
50,00
0s
36,75
23,50
10,25
1s
-3,00
25
50
75
100
125
150
175
200
225
ms
Illustration 59:
Cascade representation of windowed signal segments
The entire test signal with a duration of t = 1s is to be seen above. It contains five different segments each
of 0.2s duration. The GAUSSian window (vector length 512, overlap 450 readings or samples) is superim-
posed on this signal. At time intervals of c. 30ms this window cuts out a segment.
Above a wide GAUSSian window (parameter 3) is used and below a narrow GAUSSian window (parame-
ter 10) is used. After the noise, the fields of the 4 different frequencies in this “time landscape” can clearly
be recognised.
The windowed signal segments must clearly overlap because otherwise information could be lost. Infor-
mation is meaningful patterns of a certain duration which must not be “dissected”. How far this overlap-
ping can go depends on the highest frequency contained in the signal. It indicates the maximum speed at
which the signal - and therefore the information - can change. This will be taken up again in Chapter 9
(“Digitalisation”).
