Speech Data

次へ: Analysis condition of Acoustic 上へ: Speaker Classification Problem 戻る: Speaker Classification Problem

Speech Data

For the experiment, we made a pseudo speech using word utterance speech data (ATR 5240 word utterance speech data). The number of speakers was 4. Each speaker uttered 8 blocks, each of which consisted of 8 different words. Silences before and after each word were deleted. The words were randomly concatenated so each speech set consisted of 32 blocks and 31 speaker transitions. Fig.3 shows an example of the speech data. We use 8 sets of speech data in the experiment. Average set length was about 150 seconds.

**図 3:** Fig.3 Example of speech data (Speaker: 4 males)
$\begin{figure}\begin{center} \fbox{\epsfig{file=PS/seq.ps,height=25mm}}%% ,width=70mm}} \end{center}\end{figure}$

Jin'ichi Murakami 平成13年1月19日