Example of how voice interrupt detection works

This example illustrates the stages that Blueworx Voice Response goes through in interpreting a voice interrupt.

Figure 1 shows the audio signal level of the two-syllable word “cancel” being spoken to a Blueworx Voice Response application. We follow the stages which Blueworx Voice Response goes through for this word to be interpreted as a voice interrupt.

Figure 1. How voice interrupt detection parameters act on a voice sample
The graphic is in the form of a graph with time on the x axis and the audio signal energy level on the y axis. The voice interrupt detection level is marked on the y axis. A curve is drawn which twice exceeds the detection level, as detailed in the succeeding text.
  1. The energy level of the audio signal exceeds the level set in the voice interrupt detection level parameter.
  2. The value of the voice interrupt detection on time parameter is reached and the audio signal will qualify as a voice interrupt.
  3. The sample energy level of the voice sample falls below the voice interrupt detection level parameter. Blueworx Voice Response counts (in milliseconds) until the value of the off time parameter is reached. If the off time elapses without the audio signal exceeding the voice interrupt detection level parameter, Blueworx Voice Response interprets this as a voice Interrupt.
  4. The sample energy level rises above the voice interrupt detection level again. Blueworx Voice Response registers this and the off time is reset until the level drops below the line again, and no voice interrupt detection is reported to the application.
  5. Again, the sample energy level falls below the threshold. Blueworx Voice Response begins counting milliseconds for the off time parameter to expire.
  6. If the second peak of the energy level of the sample was not here, the Blueworx Voice Response would report a voice interrupt at this point. However, as the peak is there, the off time has not been reached, so nothing is done here.
  7. Blueworx Voice Response registers that the off time has elapsed and the energy level of the audio signal has remained below the voice interrupt detection level. A voice interrupt is reported, and the play action stops (unless the prompt has been force played).

In Figure 1, when the off time is less than the difference between points 3 and 4, the off time criterion is satisfied before point 4, and a voice interrupt is detected. Blueworx Voice Response acts on the next action in the state table and if this is a record or speech recognition action, then the second half of the word in this example may be used.

Similarly, if the voice interrupt detection level is set higher, the time difference between points 3 and 4 increases. If this time difference exceeds the off time, then as with the case above, the second half of the word in this example may be used.