3.4.1 Using isip_transform:
Command Line Interface
We use a single integrated tool, known as
isip_transform
for feature extraction. It has been briefly introduced in previous
sections. The process for extracting features entails two steps:
- Transform Builder: Use isip_transform_builder
to create and configure a signal flow graph. The signal
flow graph comprises input, algorithms, and output
components where the features are stored. When the signal
flow graph is complete, a
recipe
contains the information from each component. To learn more about
the isip_transform_builder go to
Section 3.3.
- Transforms: Use isip_transform using a file
containing your recipe to extract the features from audio data.
The recipe specifies what format to expect for the speech input,
what algorithms and their implementations to apply to extract the
features, and what should be the format of output file containing the
features.
Once a recipe has been created, running the isip_transform is
the final step in feature extraction.
When completed, it produces a feature file that can readily be
used by the speech recognizer. The command line and argument
formats are described in
Section 3.4.2.
|