Digital Signal Processing Reference
In-Depth Information
Figure 4.13 Real-time spectrum
4.7.1 Vehicle Classification
The idea of performing a time average is very important. We can use it for
classifying automotive engines. Suppose we investigate a two-stroke engine and a
four-stroke engine. A sequence of data z
k
is collected as a
.wav
file using a
conventional media player for a sufficiently long time, say 10 s. We choose a record
x
i
having size m
¼
2
N
, part of z
k
, such that when we concatenate all the
x
i
we will
obtain the original sequence z
k
. We average the vectors
x
i
and compute the DFT as
M
X
M
i
¼
1
x
i
1
jj
DFT
jj
¼¼¼¼¼)
s:
x
¼
and
x
Here the block mean is replaced by the recursive mean using concepts in
Section 1.7. Figure 4.14(a) shows the spectrum of the averaged vector
x
for a
four-stroke engine and Figure 4.14(b) shows it for a two-stroke engine. Notice the
distinct difference in spectral characteristics, allowing us to obtain a parametric
signature. This averaging is essential
in order to obtain the features of each
spectrum.
4.7.2 Instrument Classification
This is a typical problem to upgrade the single micro-phone recording to a
sterio recording. Let the time series recorded using a single micro-phone in a
Search WWH ::
Custom Search