V2.0 Feature Tour

December 15, 2021
Blog Thumbnail

V2.0 releases of Porcupine wake word and Rhino Speech-to-Intent engines achieve much higher accuracies than the predecessors and add official support for French and Spanish languages.

Porcupine Wake Word Engine

Accuracy

Porcupine V2.0 is twice more accurate than V1.9. This improvement comes at no extra computational cost and with the same ease of type-and-click model training paradigm using Picovoice Console.

Figure 1.

Rhino Speech-to-Intent Engine

Accuracy

Rhino V2.0 is 30% more accurate than its previous version, V1.6. This improvement is more pronounced when noting that V1.6 was already surpassing the performance of marquee cloud-based alternatives such as Amazon, Microsoft, Google, and IBM.

Figure 2.

Overlapping Speech

One of the challenges in spoken language understanding is guessing when the user is done. This problem is also known as endpointing. This is fairly challenging especially when there is noise even. A special kind of noise is a second speaker or additional speakers. Rhino V2.0 comes with an optional feature that allows voice command recognition even in presence of overlapping speech. To learn how to activate this feature please refer to your SDK documentation.

Language Support

V2.0 brings official support for Spanish and French languages besides existing support for English and German. Support for even more languages is underway. Stay tuned!