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

TL;DR

This paper presents a novel training method for spiking neural networks that enables low-power voice activity detection suitable for neuromorphic hardware, achieving state-of-the-art performance with significant power savings.

Contribution

It introduces a training procedure for SNNs that reduces spiking activity and network connections without performance loss, advancing low-power neuromorphic VAD.

Findings

Achieves state-of-the-art VAD performance.

Reduces network connections by 85% without accuracy loss.

Demonstrates significant power efficiency over existing methods.

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 an RNN-like model and trained with known deep learning techniques. We describe an SNN training procedure that achieves low spiking activity and pruning algorithms to remove 85% of the network connections with no performance loss. The model achieves state-of-the-art performance with a fraction of power consumption comparing to other methods.