Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection

Abstract: 

Recent advances in Voice Activity Detection (VAD) are driven by artificial and Recurrent Neural Networks (RNNs), however, using a VAD system in battery-operated devices requires further power efficiency. This can be achieved by neuromorphic hardware, which enables Spiking Neural Networks (SNNs) to perform inference at very low energy consumption. Spiking networks are characterized by their ability to process information efficiently, in a sparse cascade of binary events in time called spikes. However, a big performance gap separates artificial from spiking networks, mostly due to a lack of powerful SNN training algorithms. To overcome this problem we exploit an SNN model that can be recast into a recurrent network and trained with known deep learning techniques. We describe a training procedure that achieves low spiking activity and apply pruning algorithms to remove up to 85% of the network connections with no performance loss. The model competes with state-of-the-art performance at a fraction of the power consumption comparing to other methods.

up
0 users have voted:

Paper Details

Authors:
Flavio Martinelli, Giorgia Dellaferrera, Pablo Mainar, Milos Cernak
Submitted On:
27 May 2020 - 8:49am
Short Link:
Type:
Presentation Slides
Event:
Presenter's Name:
Flavio Martinelli
Paper Code:
SS-L5.5
Document Year:
2020
Cite

Document Files

icassp2020snn_slides.pdf

(37)

Subscribe

[1] Flavio Martinelli, Giorgia Dellaferrera, Pablo Mainar, Milos Cernak, "Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection", IEEE SigPort, 2020. [Online]. Available: http://sigport.org/5440. Accessed: Jul. 13, 2020.
@article{5440-20,
url = {http://sigport.org/5440},
author = {Flavio Martinelli; Giorgia Dellaferrera; Pablo Mainar; Milos Cernak },
publisher = {IEEE SigPort},
title = {Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection},
year = {2020} }
TY - EJOUR
T1 - Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection
AU - Flavio Martinelli; Giorgia Dellaferrera; Pablo Mainar; Milos Cernak
PY - 2020
PB - IEEE SigPort
UR - http://sigport.org/5440
ER -
Flavio Martinelli, Giorgia Dellaferrera, Pablo Mainar, Milos Cernak. (2020). Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection. IEEE SigPort. http://sigport.org/5440
Flavio Martinelli, Giorgia Dellaferrera, Pablo Mainar, Milos Cernak, 2020. Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection. Available at: http://sigport.org/5440.
Flavio Martinelli, Giorgia Dellaferrera, Pablo Mainar, Milos Cernak. (2020). "Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection." Web.
1. Flavio Martinelli, Giorgia Dellaferrera, Pablo Mainar, Milos Cernak. Spiking neural networks trained with backpropagation for low power neuromorphic implementation of voice activity detection [Internet]. IEEE SigPort; 2020. Available from : http://sigport.org/5440