A VCO-based Voice Activity Detection System for IoT applications
Dante Loi, Nordin Zbida, Luis Hernández, Andreas Wiesbauer · 2022
This paper shows a system level implementation of a Voice Activity Detector aimed at edge computing applications. The system is based on a voice band filter followed by a Manhattan norm calculator. It works with time encoded signals instead of charged capacitors or analog voltages. The advantages of this approach span from ultra low power consumption, to the suitability for deep sub-micron implementations, together with the possibility to fully synthesize the circuit using digital designing tools.