A New Lightweight CRNN Model for Keyword Spotting with Edge Computing Devices

Abstract

AbstractKeyword Spotting (KWS) is a significant branch of Automatic Speech Recognition (ASR), which has been widely used in edge computing devices. The goal of KWS is to provide high accuracy at a low false alarm rate (FAR) while reducing the costs of memory, computation, and latency. However, limited resources are challenging for KWS applications on edge computing devices. Lightweight models and structures for deep learning have achieved good results in the KWS branch while maintaining high accuracy, low computational costs, and low latency. In this paper, we present a new Convolutional Recurrent Neural Network (CRNN) architecture named EdgeCRNN for edge computing devices. EdgeCRNN is based on a depthwise separable convolution (DSC) and residual structure, and it uses a feature enhancement method. The experimental results on Google Speech Commands Dataset depict that EdgeCRNN can test 11.1 audio data per second on Raspberry Pi 3B+, which are 2.2 times that of Tpool2. Compared with Tpool2, the accuracy of EdgeCRNN reaches 98.05% whilst its performance is also competitive.KeywordsEdge computingKeyword spottingConvolutional recurrent neural networkFeature enhancementLightweight structure