On this website we use cookies, even from third parties, to improve your visit. 

PerVoice announces new engines for speech recognition

neural network

 

 

(Trento) - PerVoice announces the introduction of two new technologies that significantly improve the performance of its Audioma©-Box transcription system, both in terms of accuracy and speed.


The technology "Deep Neural Network" improves the performance of transcription accuracy due to a higher precision in the identification of the phonemes of the spoken language.
At the same time, to improve the speed of transcription, PerVoice introduces a technology that allows the use of graphic cards (GPU - Graphical Processing Unit) that can be found aboard of many computers.

The results, which depend on the language and the application domain, are significant: for example, for the Italian, the Word Error Rate decreases by 31.3% relative and this means that up to an error on three is recovered by the system. Together, the speed increases of 29.4% using the on board GPU, making it less expensive to purchase the hardware necessary for the use of the system.

Starting from the first week of May, Pervoice will make available to the market the version 1.7 of Audioma©-Box which includes the technologies mentioned above.

PerVoice - Viale Verona 190/1 38123 Trento Tel. +39 0461383960 - Email:info@pervoice.it - Part. IVA 02019100227          ©2014 PerVoice SpA

PerVoice is a company of