MarbleNet: Deep 1D Time-Channel Separable Convolutional Neural Network for Voice Activity Detection

TL;DR

MarbleNet is a deep neural network utilizing 1D time-channel separable convolutions, achieving comparable voice activity detection performance to state-of-the-art models with significantly fewer parameters, and demonstrating robustness in real-world scenarios.

Contribution

The paper introduces MarbleNet, a novel deep residual network with 1D separable convolutions, reducing parameter count while maintaining high VAD accuracy.

Findings

Achieves similar performance to state-of-the-art VAD models

Uses approximately 1/10th the parameters of comparable models

Demonstrates robustness through extensive ablation studies

Abstract

We present MarbleNet, an end-to-end neural network for Voice Activity Detection (VAD). MarbleNet is a deep residual network composed from blocks of 1D time-channel separable convolution, batch-normalization, ReLU and dropout layers. When compared to a state-of-the-art VAD model, MarbleNet is able to achieve similar performance with roughly 1/10-th the parameter cost. We further conduct extensive ablation studies on different training methods and choices of parameters in order to study the robustness of MarbleNet in real-world VAD tasks.