SPECSYNTH creates a spectrogram from a digital sound waveform,
allowing copy synthesis through a tracing method. The synthesis
procedure is the same as that used by SYNTH. When the program SPECSYNTH
is started, it searches the contents of the working memory buffer for a
variable "y" which is assumed to represent the input sound waveform. If
the variable "y" does not exist, it is generated as a 200-ms segment of
a "neutral" vowel similar to schwa.
Entering Data/Track Drawing
Upon starting the program, a spectrogram
is generated from the user-defined waveform vector. The spectrogram is
generated with a Fast Fourier Transform (FFT) method. The lines
representing the formant tracks are superimposed on the spectrogram.
The fundamental frequency (F0) track is estimated from the waveform by
a sliding window cepstral analysis (Rabiner & Schafer, 1978). These
initial fundamental frequency estimates can be refined or smoothed with
the aid of the drawing functions. The overall amplitude (AV) is
estimated from the waveform by computing the rms energy in a sliding
rectangular window. Editing and track drawing are mouse controlled, as
described for the SYNTH program.
SPECSYNTH has two playback options. The program can play the synthesized utterance, or it can play the original sound for comparison purposes.
Back to Track Draw home page