Encoder-Decoder Neural Architecture Optimization for Keyword Spotting
This work addresses keyword spotting for audio processing applications, offering a specialized model design that improves accuracy while maintaining memory efficiency, though it is incremental as it builds on existing neural architecture search methods.
The paper tackled the problem of keyword spotting by using neural architecture search to design convolutional neural networks, achieving a state-of-the-art accuracy of over 97% on Google's Speech Commands Dataset.
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%.