Table of Contents
Advertisement
5.1 Writing Message script
Because the fequency bandwidth of the reproduced speech
is narrower than that of natural speech, care should be
tak
e
n to write a script that avoids the use of ambiguous
expressions. Also, if messages that share common phrases
are used, ROM area can be used effectively.
5.2 Original Spee
c
h Recording
Because the effective dynamic range for synthesized
speech is narrow, the speaker (announcer) should take
care to suppress inflections and to speak in an level,
even tone of voice.
Because the synthesizing process sometimes emphasizes
noise in the original voice recording, care should be
taken to keep all noise (particularly hum) to the abso-
lute minimum. For this reason, the use of a professional
recording studio using open reel tapes is recommended.
Also, to permit selection of the sample with the best
speech quality and the least noise, you are asked to
provide several recordings of each message.
5.3 original Speech Evaluation and Editing
Selection should be made from among the recorded messages
taking into consideration lack of noise, even tone of
voice, sound quality, etc. The messages thus selected
should be edited onto an open reel or cassette tape. When
an open reel tape is used, either the full tape or two
tracks may be used. For a good
SiN
ratio when using a
cassette, the .use of metal tape is recommended. The
recording should be done at a somewhat high level with
peak values are in the range of +5 to +8dB. To ease the
analysis process, the format shown in Fig. 5-2 should be
used with the messages appropriately divided by blanks.
5
-
2
Advertisement
Table of Contents
