Digital Signal Processing Reference
In-Depth Information
10
4
2
×
(a)
0
2
−
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
Time (sec)
(b)
(c)
(d)
(e)
(f)
(g)
200
(h)
100
0
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
Time (sec)
Figure 6.12
Comparison of pitch contours of PDAs for (a) female speech. (b)
Reference; (c) TA, (d) WTA, (e) STA, (f) SS, (g) WSS, and (h) SS-SA-based PDAs
of the estimation. Several methods have been proposed to flatten the speech
spectrum in order to avoid the formant interaction effect [20, 5, 22, 23]. The
speech spectrum is first flattened by removing the formants (by either linear
or nonlinear methods) before the pitch estimation process can begin.
The linear spectrum-flattening method uses the LPC inverse filter to remove
the formants from the speech signal. The main drawback of this method is that
for high-pitched speech, like that of females and children, the first complex
Search WWH ::
Custom Search