Hello Edge: Keyword Spotting on Microcontrollers (2017-11-20T00:00:00.000000Z)

TL;DR

It is shown that it is possible to optimize these neural network architectures to fit within the memory and compute constraints of microcontrollers without sacrificing accuracy, and the depthwise separable convolutional neural network (DS-CNN) is explored and compared against other neural network architecture.

Abstract

Keyword spotting (KWS) is a critical component for enabling speech based user interactions on smart devices. It requires real-time response and high accuracy for good user experience. Recently, neural networks have become an attractive choice for KWS architecture because of their superior accuracy compared to traditional speech processing algorithms. Due to its always-on nature, KWS application has highly constrained power budget and typically runs on tiny microcontrollers with limited memory and compute capability. The design of neural network architecture for KWS must consider these constraints. In this work, we perform neural network architecture evaluation and exploration for running KWS on resource-constrained microcontrollers. We train various neural network architectures for keyword spotting published in literature to compare their accuracy and memory/compute requirements. We show that it is possible to optimize these neural network architectures to fit within the memory and compute constraints of microcontrollers without sacrificing accuracy. We further explore the depthwise separable convolutional neural network (DS-CNN) and compare it against other neural network architectures. DS-CNN achieves an accuracy of 95.4%, which is ~10% higher than the DNN model with similar number of parameters.

Authors

Liangzhen Lai

2 papers

V. Chandra

2 papers

Yundong Zhang

1 papers

TL;DR

Abstract

Authors

References36 items

CMSIS-NN: Efficient Neural Network Kernels for Arm Cortex-M CPUs

The Microsoft 2017 Conversational Speech Recognition System

Dual Path Networks

ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices

MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

Convolutional Recurrent Neural Networks for Small-Footprint Keyword Spotting

Deep Convolutional Neural Network Inference with Floating-point Weights and Fixed-point Activations

Max-pooling loss training of long short-term memory networks for small-footprint keyword spotting

Xception: Deep Learning with Depthwise Separable Convolutions

Model Compression Applied to Small-Footprint Keyword Spotting

Layer Normalization

TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems

Personalized speech recognition on mobile devices

Going Deeper with Embedded FPGA Platform for Convolutional Neural Network

Throughput-Optimized OpenCL-based FPGA Accelerator for Large-Scale Convolutional Neural Networks

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Adam: A Method for Stochastic Optimization

Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling

Sequence to Sequence Learning with Neural Networks

Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

Small-footprint keyword spotting using deep neural networks

Do Deep Nets Really Need to be Deep?

Low-rank matrix factorization for Deep Neural Network training with high-dimensional output targets

Discriminative keyword spotting

An Application of Recurrent Neural Networks to Discriminative Keyword Spotting

Learning Precise Timing with LSTM Recurrent Networks

Bidirectional recurrent neural networks

Long Short-Term Memory

Automatic recognition of keywords in unconstrained speech using hidden Markov models

A hidden Markov model based keyword recognition system

Experimental, limited vocabulary, speech recognizer

Speech commands: A public dataset for single-word speech recognition

Convolutional neural networks for small-footprint keyword spotting

Compressing deep neural networks using a rank-constrained topology

Long short-term memory recurrent neural network architectures for large scale acoustic modeling

Recurrent neural network based language model

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names