Authors: Xiang Hao, Chenxiang Ma, Qu Yang, Jibin Wu, Kay Chen Tan

Published on: October 07, 2024

Impact Score: 8.6

Arxiv code: Arxiv:2410.04785

Summary

• What is new: Introducing an ultra-low-power speech enhancement system using spiking neural networks (SNN), specifically designed for edge devices.
• Why this is important: Current speech enhancement methods have high computational costs, making them unsuitable for devices like headsets and hearing aids.
• What the research proposes: The proposed Spiking-FullSubNet, which uses a full-band and sub-band fused approach and a novel spiking neuron model, efficiently enhances speech with low power consumption.
• Results: Spiking-FullSubNet demonstrated superior performance in speech quality and energy efficiency, winning the Intel Neuromorphic Deep Noise Suppression Challenge.

Technical Details

Technological frameworks used: Spiking neural network (SNN) framework

Models used: Spiking-FullSubNet, novel spiking neuron model

Data used: Intel Neuromorphic Deep Noise Suppression (N-DNS) Challenge dataset

Potential Impact

Headset and hearing aid manufacturers, companies working on edge computing solutions, and potentially any industry requiring efficient speech enhancement technology.

Want to implement this idea in a business?

We have generated a startup concept here: Neuromorphic Audio Solutions.