Neuromorphic Computing Architectures for Ultra-Low Power Edge Keyword Spotting
Views: 3
Downloads: 0
Nur Susanto
1
,
Ilham Prasetyo
2
1, 2Department of Forestry and Environmental Science, IPB University
Downloads
Always-on acoustic sensing devices demand sub-milliwatt power budgets to operate sustainably on miniature button batteries. We construct a Spiking Neural Network (SNN) mapped onto neuromorphic hardware chips, achieving 95.7% keyword detection accuracy with a power draw of merely 180 microwatts.
