Preprocessing the Input Signal
E-6
E.6 Preprocessing the Input Signal
Preprocessing (that is,
small amounts of digital multiband processing and
analog compression, amplitude normalization, and graphic equalization) can
enhance the quality of the speech, reduce editing time, and reduce the
time-to-market for Texas Instruments synthesizers.
Amplitude Normalization—Be sure the wave file is fully amplified (100%).
Normalize (reduce) the amplitude by approximately 15% (~3 dB) with a good
wave editor (such as Cool Edit Pro
). It is a good idea to save this file under
a different name to indicate that it has been down-sampled and normalized.
This distinguishes the normalized file from the original wave file.
Down-Sampling—Using a wave-editing tool with good filtering, such as Cool
Edit Pro
or Matlab
, down-sample .WAV files from 48 kHz (preferred) or 44.1
kHz to 8 kHz or 10 kHz, respectively. A higher sampling rate improves the
quality but increases the size of the data.
Note:
Some wave-editing tools may introduce noise when down-sampling. The
quality of the results depends upon the quality of the
filter.
Editing the Wave File—Border-editing (i.e., deleting silence at the beginning
and end of the signal) must also be performed with a wave-editing tool. The
following editing steps can be applied:
-
Remove artifacts (such as transients, etc.) and extraneous noise from the
signal.
-
Replace unacceptable speech sounds (such as slurred or distorted
sounds) with correct sounds from somewhere else in the file.
-
Selectively boost the amplitude of weak fricative sounds
[f, v, th (voiced
and unvoiced), and z], if needed. Cut the amplitude of strong sibilants (s,
sh, ch, j), if needed).
Graphic Equalization—The synthesis process tends to boost the lower
frequencies and attenuate the higher frequencies in a signal. Sometimes the
effect is a
muffled or muddy quality. Therefore, wave files of such voices may
require graphic equalization to increase or decrease the gain for certain
frequencies. With the graphic equalizer of a wave editing program, the
amplitude of higher frequencies (~2–4 kHz and above) can be boosted while
the amplitude of lower frequencies (below ~100 Hz) can be attenuated.
Note:
Use graphic equalization as a last resort, since this process can cause the
voice to process more poorly.
Summary of Contents for MSP53C691
Page 1: ...MSP53C691 Speech Synthesizer December 2000 MSDS Speech User s Guide SPSU020 ...
Page 22: ...2 10 ...
Page 96: ...C 6 ...