This section defines the association between speech signals and file extensions. Typically speech signals are stored in binary form with a file header showing some information about the contents of the file. One such piece of information can be a name for the signal or signals contained in the file. For example ESPS (and the SSFF format used by Emu) format files contain a name for every signal in a file in the file header. In this case, Emu can search all files with a given basename to find a given signal or track, for example F0. Other file formats (such as Windows .wav files) do not have this information and so this search would not be possible. This section of the template defines an association so that the search is not needed.
The syntax of the track statement is:
where trackname is the name of the track and extn+ denotes one or more extensions separated by spaces. If the signal files do contain the track name in the header, then the track name listed here must match exactly. Where there is no track name stored in the file header, the track name used here can be any arbitrary name.track trackname extn+
In the current version of Emu, the track name samples always refers to the sampled acoustic waveform and is used, for example, in calculating spectrograms. This usage derives from the default behaviour of ESPS which calls any sampled signal samples in the file header. This assumption may be lifted in future but for now, the sampled speech data should be called samples regardless of the file format used.
The example shows track statements corresponding to a common set of ESPS derived signal files:
track samples sd track fm fb track bw fb track rms f0 track F0 f0 track ac_peak f0 track k1 f0 track prob_voice f0 |