| Japanese | English |

MEASUREMENT OF JAPANESE SYLLABLES /KA, KI, KU, KE, KO, SA, SI, SU, SE, SO/
(Analysis of the Japanese voice 7)

Date:	12:00, 30 Sep. 2002
Place:	Nagoya, Japan
Microphone:	SONY ECM-MS957
Amplifier:	SONY DAT WALKMAN TCD-D100
Personal computer:	DELL INSPIRON 7500
OS:	Windows 2000 Professional
Software:	DSSF3
WAVE sound file:	voice71.wav (44.1kHz / Stereo / 8sec / 1.34MB) voice72.wav (44.1kHz / Stereo / 7.48sec / 1.25MB)

This is the calculation conditions of SA.

As in the last time, the ACF of which t_e value is the minimum is shown later.

The graph below shows the time change of the sound level. Five utterances of "ka, ki, ku, ke, ko" can be seen.

Even though the same syllable "ka" is pronounced, there are many ways of utterances. Especially the Japanese language has different word with the same pronunciation. So, many templates are required for one syllable.

"ka" 0.68 second after measurement start, 40msec after utterance start
t₁=3.31msec / f₁=0.23 / t_e=4.5msec / F(0)(SPL)=-18.49dB

The new function is the zoom in/out display for each parameter graph.

This is the time change of the t₁ value. The utterance "ka" began at 0.64 seconds. Generally, finding the start/end point of the utterance is difficult in the speech analysis. But using this graph, we can catch the utterance correctly.

This is the time change of the t_e value. The utterance began at 0.64 seconds.

This is the time change of the f₁ value.

This is the time change of the F(0) value.

"ki" 2.35 second after measurement start, 35 msec after utterance start
t₁=0.18msec / f₁=0.42 / t_e=2.95msec / F(0)(SPL)=-25.74dB

This is the graph of t₁.

This is the graph of t_e.

This is the graph of f₁.

This is the graph of F(0).

"ku" 4.015 second after measurement start, 20 msec after utterance start
t₁=1.81msec / f₁=0.21 / t_e=2.10msec / F(0)(SPL)=-17.02dB

"ke" 5.585 second after measurement start, 20 msec after utterance start
t₁=0.32msec / f₁=0.33 / t_e=3.87msec / F(0)(SPL)=-28.86dB

"ko" 8.665 second after measurement start, 40 msec after utterance start
t₁=0.7msec / f₁=0.16 / t_e=3.22msec / F(0)(SPL)=-39.14dB

Next, another five utterances "sa, si, su, se, so" were measured. The graph shows time change of sound level.

"sa" 0.41 second after measurement start, 25 msec after utterance start
t₁=0.09msec / f₁=0.59 / t_e=2.86msec / F(0)(SPL)=-38.98dB

The four parameters are almost same values as previous analysis.

"si" 1.77 second after measurement start, 50 msec after utterance start
t₁=0.18msec / f₁=0.48 / t_e=3.56msec / F(0)(SPL)=-29.75dB

"su" 3.4 second after measurement start, 110 msec after utterance start
t₁=0.09msec / f₁=0.37 / t_e=2.79msec / F(0)(SPL)=-27.77dB

"se" 4.795 second after measurement start, 30 msec after utterance start
t₁=0.11msec / f₁=0.50 / t_e=2.51msec / F(0)(SPL)=-41.51dB

"so" 6.355 second after measurement start, 60 msec after utterance start
t₁=0.11msec / f₁=0.33 / t_e=3.28msec / F(0)(SPL)=-40.7dB

In the table below, the data analyzed is summarized.

	time (ms)	utterance(ms)	t1(ms)	f1	te(ms)
a tune do	0.625	35	1.13	0.27	6.80
a tune re	2.040	10	1.22	0.32	4.30
a tune mi	3.575	20	1.20	0.32	6.27
a tune fa	5.075	20	1.22	0.38	4.94
a tune so	6.605	25	1.20	0.66	7.76
a tune ra	8.140	5	1.18	0.49	6.45
a tune si	9.650	10	1.30	0.30	4.91
a tune do	11.255	20	1.25	0.50	4.45
a vowel compare		5	1.18	0.44	3.98
I		10	0.27	0.97	6.78
u		10	2.18	1.00	5.05
e		15	0.29	0.50	15.48
o		15	1.18	0.35	21.80
a		25	1.20	0.45	8.46
ka		10	1.86	0.46	6.26
sa		60	0.09	0.64	2.48
ta		10	1.93	0.43	5.07
na		25	4.35	0.98	21.94
ha		40	0.70	0.16	3.22
ma		15	4.08	1.00	8.86
ya		15	0.11	1.00	10.09
ra		20	2.59	0.18	7.29
wa		5	0.25	0.69	7.42
ka		40	3.31	0.23	4.50
ki		35	0.18	0.42	2.95
ku		20	1.81	0.21	2.10
ke		20	0.32	0.33	3.87
ko		40	0.70	0.16	3.22
sa		25	0.09	0.59	2.86
si		50	0.18	0.48	3.56
su		110	0.09	0.37	2.79
se		30	0.11	0.50	2.51
so		60	0.11	0.33	3.28

There are various pronunciations for the same character. A character can be told with a sound, but a sound is not told in a written form. The more it becomes intellectual and advanced communication, the more audio analysis may be difficult.

October 2002 by Masatsugu Sakurai