Encoder-Decoder Neural Architecture Optimization for Keyword Spotting

Abstract

Keyword spotting aims to identify specific keyword audio utterances. In recent years, deep convolutional neural networks have been widely utilized in keyword spotting systems. However, their model architectures are mainly based on off-the shelfbackbones such as VGG-Net or ResNet, instead of specially designed for the task. In this paper, we utilize neural architecture search to design convolutional neural network models that can boost the performance of keyword spotting while maintaining an acceptable memory footprint. Specifically, we search the model operators and their connections in a specific search space with Encoder-Decoder neural architecture optimization. Extensive evaluations on Google's Speech Commands Dataset show that the model architecture searched by our approach achieves a state-of-the-art accuracy of over 97%.

arXiv commentAccepted for Interspeech2021

References in corpus (12)

• Neural Architecture Search with Reinforcement Learning
• Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition
• Hello Edge: Keyword Spotting on Microcontrollers
• SNAS: Stochastic Neural Architecture Search
• A neural attention model for speech command recognition
• Streaming keyword spotting on mobile devices
• Depthwise Separable Convolutional ResNet with Squeeze-and-Excitation Blocks for Small-footprint Keyword Spotting
• Neural Architecture Search For Keyword Spotting
• Temporal Convolution for Real-time Keyword Spotting on Mobile Devices
• Noisy Differentiable Architecture Search
• MatchboxNet: 1D Time-Channel Separable Convolutional Neural Network Architecture for Speech Commands Recognition
• AutoKWS: Keyword Spotting with Differentiable Architecture Search