208
|
5 Methods for Discrete Processing and Analysis of Biosignals
FIR-LP
Sampling points
ideal LP
Fig. 5.47: FIR low-pass 15 order according to Figure 5.45 where the sample at the filter edge at
200 Hz has been lowered to 0.4.
In general, however, the measured biosignals are not periodic and therefore do not
have a discrete frequency spectrum. If the FIR filter just designed is now applied to
these signals, an error occurs. To estimate this, the impulse response is determined. In
this case, the transfer function of the FIR filter is no longer discrete, but continuous in
frequency. In Figure 5.45 it can be seen that between the given values of the magnitude
frequency response very large deviations from the ideal lowpass can occur, which are
called overshoots.
This can be remedied by no longer attempting to sample the ideal lowpass, which
has a very large slope, but by replacing the steep slope with a smoother transition from
the passband to the stopband. As an example of this, the sample point with the value
1 at the edge at 200 Hz is to be replaced by a value of 0.4. The result shows Figure 5.47.
It can be clearly seen that the flatter edge at 200 Hz greatly reduced the overshoots in
the frequency response compared to the steep edge according to Figure 5.45.
5.4 Post-Reading and Exercises
Discretisation
1.
Explain the process of discretisation and quantisation in the context of analogue-
digital-conversion of signals.
2.
Explain the Shannon theorem.
3.
What effect of analogue to digital conversion is called aliasing?
4.
Explain the effect of an anti-aliasing filter.
5.
What is the purpose of a sample-and-hold amplifier before an analogue-to-digital
converter?