ACED – Acoustic Call Event Detection

ACED uses industry-leading and patent-pending Voice Activity Detection speech technologies with accuracy (speech vs. silence) of over 93%, to automatically partition a call into:

  • Agent Speech
  • Customer Speech
  • Speech Talkover
  • Silence/Noise/Music
  • Telephony acoustic events: Beep, Ring, DTMF, etc.

It can further break speech regions down into:

  • Foreground Speech
  • Background Speech
  • Laughs and Sighs
  • Disfluencies (such as “um”)
  • Backchannels (such as “uh huh”)

Frequency and trend statistics of these different acoustic events provides:

  • A clear picture on what occurred during the call (Did the call go to voicemail? Was the call transferred?)
  • Valuable insight into agent-customer interaction dynamics (Which speaker was more dominant?)
  • An objective way to estimate the overall quality of the conversation (Did one speaker frequently interrupt the other?)