Details
Presenter(s)
Display Name
MOURRANE, SOUFIANE
- Affiliation
-
AffiliationSTMicroelectronics
- Country
Abstract
In the context of speech detection and recognition, this paper presents an event-driven low power feature extraction unit, based on continuous-time digital signal processing, that will be combined with a neural network classifier to perform keyword spotting. This system allows the recognition of up to 10 keywords with an accuracy higher than 90% which is at par or better than state of the art.