Pindrop Introduces Deep Voice 3
Pindrop, a provider of voice security and authentication technology, has launched Deep Voice 3, the latest version of its voice recognition technology.
Built on advanced machine learning and deep neural networks, Deep Voice 3 is designed to more accurately recognize the voices of callers into contact centers with less speech.
Deep Voice 3's technical advances include a rearchitected neural network, an advanced training program to teach the system based on real-world learnings, and improved accuracy and speed.
Deep Voice 3 allows for strong authentication with brief snippets of speech. This enables numerous use cases, especially in interactive voice response and Internet of Things applications where speech is conveyed in the form of small commands. Specifically, in IoT this enables speaker identification in as little as three syllables, allowing for personalization on utterances as short as wake words (for example; "OK Google," "Hi Bixby," and "Alexa").
Deep Voice 3 has been optimized to perform across a variety of acoustic and noise environments. Adapting to session variability improves authentication accuracy when handling differences in background noise conditions that exist between callers' enrollment and authentication calls.
"In today's contact center, companies are working to deliver the kind of seamless customer experience that consumers get everywhere in their digital life. This means confidently authenticating more callers and doing so at a significantly faster rate," said Vijay Balasubramaniyan, CEO of Pindrop, in a statement. "Deep Voice 3 is a crucial part of the technology that is making that possible and taking us a step closer to frictionless identity in a voice-first world."
Deep Voice 3 will be generally available this quarter to new customers via the Pindrop Passport authentication product. In the coming months, Deep Voice 3 will be incorporated into Pindrop's anti-fraud Protect product.
Pindrop's new technology incorporates machine learning into voice biometrics.