Basic Speech Recognition Parameters

The basic speech recognition parameters are used system-wide and are shared by all speech recognition applications hosted on the NuPoint UM system.

To configure basic speech recognition parameters:

1. In the navigation tree, click Auto-Attendant. The Auto-Attendant menu items appear.

2. Click Misc. Parameters. The Auto Attendant Parameters window appears.

3. In the Category field, select Basic Speech Recognition from the drop-down menu.

4. Configure the following parameters (or click Use Default Valueto apply the default.)

5. Primary Language: By default, this value is set to the System Language. For Bilingual language systems, prompts are presented in this language first and then in the secondary language. You can switch primary and secondary languages to change the order of presentation.

   Note: NOTE: If you change Primary and Secondary language, you must perform activation.

   • Secondary Language: For Bilingual language systems, select the secondary language, or select Disable to configure a unilingual system.

   • Low Recognition Confidence Level: This value specifies the relative confidence level below which speech recognition results are rejected. The minimum value is 0 and the maximum value is 1, with a two-digit precision. The default value is 0.5.

   • High Recognition Confidence Level: This value specifies the relative confidence level above which speech recognition results are implicitly confirmed. The minimum value is 0 and the maximum value is 1, with a two-digit precision. The default value is 0.8.

   • No Speech Timeout (in milliseconds): This value specifies the length of silence that will trigger a help prompt to be played to the user. The minimum value is 0 and the maximum value is 60,000 milliseconds (60 seconds). The default value is 20,000 milliseconds (20 seconds).

     Note: You may need to change the default No Speech Timeout value to a higher value if external callers are consistently slow to respond. As a rule of thumb, the No Speech Timeout value should be about 5 to 8 seconds longer than the recorded external greeting.

   • Post-Speech Silence (in milliseconds): This value specifies the length of silence that must follow an utterance before the speech recognition engine begins to process it as a complete sentence. The bigger this value is, the longer the pauses are allowed from the users. The minimum value is 0 and the maximum value is 60,000 milliseconds (60 seconds). The default value is 1000 milliseconds.

   • Output Volume: This value specifies the linear volume that is applied to the output signal before prompts are played back to the users. You can use this parameter to adjust the volume when the signal played to the users is too weak or too loud. The minimum value is 0.0 and the maximum value is 100.0. The default value is 100.0.

   • Speech vs. Accuracy: This value specifies the relative priority of speech vs. accuracy. Using high accuracy algorithms requires more CPU resources and thus limits the number of concurrent speech recognition sessions. The minimum value is 0.0 and the maximum value is 1.0. Entering “0.0” places the emphasis on speed, while entering “1.0” places the emphasis on accuracy. The default value is 0.5.

   • Sensitivity: This value allows you to configure the level between background noise and speech, and thus it controls the sensitivity of the speech detector. The minimum value is 0.00 and the maximum value is 1.00. Values approaching 1.00 improve the detection of speech but also increase the detection of background noise and thus utterances need to be spoken with a strong voice so as not to be mistaken with background noise. The default value is 0.5.

   • Barge-In: This field allows you to enable or disable barge-in for the entire speech recognition engine (that is, the Speech Auto Attendant and every other installed speech recognition application). Barge-in is enabled by default.

   • Note: Even if barge-in is enabled, speech recognition applications can choose to disable barge-in during portions of their dialogs. If barge-in is disabled, no application can use barge-in.

   • Prompt Normalization: All greeting prompts recorded into the Speech Auto Attendant will be normalized to ensure good cohesion with the rest of the Auto Attendant prompts. Prompts will be normalized to the RMS value of 2200.

   • Prompt Trimming: All greeting prompts recorded into the Speech Auto Attendant will be trimmed to ensure good cohesion with the rest of the Auto Attendant prompts. Prompt trimming will drop silence (signal below a power of 300) at the beginning and end of the prompts.

6. Click Save. The configured parameters will now be set to the selected values.