Train Audio Matching

Before using Media Server to recognize audio clips, you must train your IDOL Speech Server. HPE recommends training Speech Server by sending actions similar to the following example:


Media Server performs audio matching using the acoustic landmark approach, by sending Speech Server the request action=AddTask&Type=AfpMatchWav. This means that you must train the Speech Server using the actions AfpAddTrackStream, AfpAddTrackWav, and so on. Do not use the actions that begin with Afpt, such as AfptAddTrackStream and AfptAddTrackWav, because these relate to acoustic template matching.

Media Server can only use audio fingerprint databases that exist in the default audio fingerprint database directory. This directory is specified by the AFPDir parameter, in the [Paths] section of the Speech Server configuration file. The example request (above) adds audio fingerprints to a database (named Music) in the default directory. To ensure that training is added to the default directory, HPE recommends against setting the PackDir action parameter in your training actions, and recommends using the AfpDb parameter only when the specified database exists in the default directory.

For more information about training IDOL Speech Server, refer to the IDOL Speech Server Administration Guide.