The algorithm was next tested on co-channel simultaneous speech signals
spoken by two speakers. Each speech signal file was artificially created
by mixing two independent speech signals with 0 dB signal-to-signal
ratio.
To evaluate our algorithm objectively, we also applied cepstrum for
simultaneous speech signals which is not generally designed as a
multi-pitch detector.
Results with cepstrum are shown in table 2 and results with our
algorithm are shown in table 3.
An example of detected contours is depicted in
figure 5 where the reference is shown in figure 6.
Pairs of speech files by which concurrent speech signals are created are
shown in the first and second columns in table 2 and 3.
As the results, our algorithm significantly outperformed cepstrum as
well and showed high performance.
Some of the gross errors were found at the first process mainly because
of unvoiced consonants.
Since we focused only on harmonic structure, the gross errors
caused by them were difficult to avoid.
Meanwhile, when the two simultaneous speakers were male and female, male
rather resulted worse. At the second process stated in
Section 3, AIC rather prefers to be positioned in
as higher frequency as it can because the number of free parameters can
be lessen.
Accordingly, if both pitch and amplitude of one utterence was
specificially lower than another, it tended to be ignored.