Hello everyone! I am a voice biometrics enthusiast and I am currently exploring the possibility of integrating an external biometrics engine with Genesys Cloud for customer identity verification. We want to use the ‘AudioHook’ API to stream the live call audio to our biometrics server and then receive a ‘Match’ or ‘No Match’ result to update the agent’s script. Has anyone successfully implemented a real-time voice biometrics check using AudioHook, and what are the latency implications for the customer experience?
Hey Arj23! I am a speech analytics manager and I love the idea of using biometrics for verification! We have tested this with a few AppFoundry partners. The latency with AudioHook is actually very low, usually sub-one hundred milliseconds. However, you must ensure that your biometrics engine can handle the ‘Dual-Channel’ audio from AudioHook. You need to separate the customer’s voice from the agent’s voice for the biometrics check to be accurate. If you send a mixed stream, the accuracy will drop significantly.
I manage the schedules for our verification team. Arj23, from a process perspective, you should consider the ‘Enrollment’ phase. You need a way to capture the customer’s initial ‘Voice Print’ and store it securely. We found that the best time to do this is during a standard IVR interaction where the customer is already providing their account details. You can use the Recording API to export those snippets to your biometrics engine for training.
Hello. I am also a biometrics enthusiast. To follow up on Ana20, when you use AudioHook, you should also implement a ‘Confidence Score’ threshold in your Architect flow. Do not just rely on a simple yes or no. If the biometrics engine returns a low confidence score, your flow should fall back to a traditional security question. This ensures that you do not have a high rate of false rejections for customers with background noise or poor cell phone reception.