Answer Machine Detection Custom Server functions

Description of the custom server functions and associated parameters.

The custom server provides two functions, SingleSimple and SingleComplex.

SingleSimple supports the following input parameters

SingleSimple supports the following output parameters

SingleComplex supports all input parameters

SingleComplex supports the following output parameters

Table 1. AMD Input Parameters
Parameter Name Description Type Default Value (Used when parameter set to 0)
adaptive_threshold This controls which algorithm is used for the detection of voice energy. There are three supported algorithms to calculate the threshold. Any energy less than the threshold is considered silence/background, anything above the threshold is considered voice/beep.
  • 0 - Use a fixed energy (threshold argument)
  • 1 - Ratio, the threshold is set to a little more than the minimum energy detected
  • 2 - Offset, the threshold is set to the minimum energy + threshold argument
long 0
threshold This controls the threshold used in the calculations when adaptive_threshold is 0 or 2 long 10000
voice_on_time Minimum duration of voice to be considered initial voice. This is used decide if "real" voice has started and not a click. long 100 ms
voice_off_time Duration of silence after voice to consider that the speaker has finished talking. If less than human_time of voice has been detected so far then the detection result is Human. If more than human_time of voice has been detected then the detection result is Machine. long 1000 ms
human_time Duration of voice to be considered a human. If the voice is less than this time its a human, otherwise its a machine. long 5000 ms
run_time How long to run the detection algorithm for. This can happen if silence is on the line, or the speaker is still speaking. This results in the Unknown detection.
Note: The SendData custom server function time MUST be longer than the supplied run_time.
long 10000 ms
beep_on_time Duration of steady (beep) energy to be considered a potential beep. long 120 ms
beep_off_time Duration of silence after a beep to be considered a completed beep. This results in the Beep detection. long 120 ms
beep_modulation Percentage difference between samples to be considered steady energy and thus a potential beep. long 3%
event_stop_mask Mask of events to cause the detection algorithm to stop. If an event is not present in mask it will be ignored. The events must be added together for this argument.
  • 1 - Unknown (always on)
  • 2 - Human
  • 8 - Machine
  • 32 - Beep

So to stop only on human and machine the mask would be 1 + 2 + 8 = 13.

long all
Table 2. AMD Output Parameters
Parameter Name Description Type Values
type Detection type. long
  • 0 - Unknown
  • 1 - Human
  • 3 - Machine
  • 5 - Beep