Main Components

The CPQD ASR is formed by the components shown in Fig. 5. Not all the components are required and their installation depends on the needs of the customer’s application.

../_images/asr-deploy.png

Fig. 5 CPQD ASR installation structure

Each CPQD ASR component has its own installer:

CPQD ASR Engine

Its installation is required and does not depend on what language is used nor on the input audio sample rate. Its installer follows the format: cpqd-asr-engine-VERSION.run. Before installing, please check software and hardware requirements.

CPQD ASR Server

Its installation is required and does not depend on what language is used nor on the input audio sample rate. Its installer follows the format: cpqd-asr-server-VERSION.run. Before installing, please check software and hardware requirements.

Acoustic Model

Its installation is required and depends on the language used and the audio sample rate. For example, for Brazilian Portuguese and 8 kHz audio (phone applications), the installer follows the format: am-ptbr-8k-VERSION.run. For 16 kHz audio, the installer follows the format: am-ptbr-16k-VERSION.run.

Free Speech Model

Its installation is optional and depends on the installed acoustic model. If the application is going to use free speech recognition, then it must be installed. Before installing, please check memory requirements. For each acoustic model, there is a corresponding free speech model installer. For example, for the acoustic``am-ptbr-8k-1.0.6-b3.run``, there will be a free speech model following the format slm-ptbr-VERSION-am-ptbr-8k-1.0.6-b1.run, where slm-ptbr-VERSION identifies the free speech model, while am-ptbr-8k-1.0.6 is the associated acoustic model.

Built In Grammars

Their installation is optional and only depends on the adopted language. If the application will use the built in grammar feature, then they should be installed. Their installer follows the format grammar-ptbr-builtin-VERSION.run.