DTW-NN: A novel neural network for time series recognition using dynamic alignment between inputs and weights

Brian Kenji Iwana, Volkmar Frinken, Seiichi Uchida

Research output: Contribution to journalArticle

Abstract

This paper describes a novel model for time series recognition called a Dynamic Time Warping Neural Network (DTW-NN). DTW-NN is a feedforward neural network that exploits the elastic matching ability of DTW to dynamically align the inputs of a layer to the weights. This weight alignment replaces the standard dot product within a neuron with DTW. In this way, the DTW-NN is able to tackle difficulties with time series recognition such as temporal distortions and variable pattern length within a feedforward architecture. We demonstrate the effectiveness of DTW-NNs on four distinct datasets: online handwritten characters, accelerometer-based active daily life activities, spoken Arabic numeral Mel-Frequency Cepstrum Coefficients (MFCC), and one-dimensional centroid-radii sequences from leaf shapes. We show that the proposed method is an effective general approach to temporal pattern learning by achieving state-of-the-art results on these datasets.

Original languageEnglish
Article number104971
JournalKnowledge-Based Systems
Volume188
DOIs
Publication statusPublished - Jan 5 2020

Fingerprint

Time series
Neural networks
Feedforward neural networks
Accelerometers
Neurons
Alignment
Warping
Coefficients
Product standards
Neuron

All Science Journal Classification (ASJC) codes

  • Management Information Systems
  • Software
  • Information Systems and Management
  • Artificial Intelligence

Cite this

DTW-NN : A novel neural network for time series recognition using dynamic alignment between inputs and weights. / Iwana, Brian Kenji; Frinken, Volkmar; Uchida, Seiichi.

In: Knowledge-Based Systems, Vol. 188, 104971, 05.01.2020.

Research output: Contribution to journalArticle

@article{af722f2cea774ca7a23dc4986804784d,
title = "DTW-NN: A novel neural network for time series recognition using dynamic alignment between inputs and weights",
abstract = "This paper describes a novel model for time series recognition called a Dynamic Time Warping Neural Network (DTW-NN). DTW-NN is a feedforward neural network that exploits the elastic matching ability of DTW to dynamically align the inputs of a layer to the weights. This weight alignment replaces the standard dot product within a neuron with DTW. In this way, the DTW-NN is able to tackle difficulties with time series recognition such as temporal distortions and variable pattern length within a feedforward architecture. We demonstrate the effectiveness of DTW-NNs on four distinct datasets: online handwritten characters, accelerometer-based active daily life activities, spoken Arabic numeral Mel-Frequency Cepstrum Coefficients (MFCC), and one-dimensional centroid-radii sequences from leaf shapes. We show that the proposed method is an effective general approach to temporal pattern learning by achieving state-of-the-art results on these datasets.",
author = "Iwana, {Brian Kenji} and Volkmar Frinken and Seiichi Uchida",
year = "2020",
month = "1",
day = "5",
doi = "10.1016/j.knosys.2019.104971",
language = "English",
volume = "188",
journal = "Knowledge-Based Systems",
issn = "0950-7051",
publisher = "Elsevier",

}

TY - JOUR

T1 - DTW-NN

T2 - A novel neural network for time series recognition using dynamic alignment between inputs and weights

AU - Iwana, Brian Kenji

AU - Frinken, Volkmar

AU - Uchida, Seiichi

PY - 2020/1/5

Y1 - 2020/1/5

N2 - This paper describes a novel model for time series recognition called a Dynamic Time Warping Neural Network (DTW-NN). DTW-NN is a feedforward neural network that exploits the elastic matching ability of DTW to dynamically align the inputs of a layer to the weights. This weight alignment replaces the standard dot product within a neuron with DTW. In this way, the DTW-NN is able to tackle difficulties with time series recognition such as temporal distortions and variable pattern length within a feedforward architecture. We demonstrate the effectiveness of DTW-NNs on four distinct datasets: online handwritten characters, accelerometer-based active daily life activities, spoken Arabic numeral Mel-Frequency Cepstrum Coefficients (MFCC), and one-dimensional centroid-radii sequences from leaf shapes. We show that the proposed method is an effective general approach to temporal pattern learning by achieving state-of-the-art results on these datasets.

AB - This paper describes a novel model for time series recognition called a Dynamic Time Warping Neural Network (DTW-NN). DTW-NN is a feedforward neural network that exploits the elastic matching ability of DTW to dynamically align the inputs of a layer to the weights. This weight alignment replaces the standard dot product within a neuron with DTW. In this way, the DTW-NN is able to tackle difficulties with time series recognition such as temporal distortions and variable pattern length within a feedforward architecture. We demonstrate the effectiveness of DTW-NNs on four distinct datasets: online handwritten characters, accelerometer-based active daily life activities, spoken Arabic numeral Mel-Frequency Cepstrum Coefficients (MFCC), and one-dimensional centroid-radii sequences from leaf shapes. We show that the proposed method is an effective general approach to temporal pattern learning by achieving state-of-the-art results on these datasets.

UR - http://www.scopus.com/inward/record.url?scp=85071263027&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071263027&partnerID=8YFLogxK

U2 - 10.1016/j.knosys.2019.104971

DO - 10.1016/j.knosys.2019.104971

M3 - Article

AN - SCOPUS:85071263027

VL - 188

JO - Knowledge-Based Systems

JF - Knowledge-Based Systems

SN - 0950-7051

M1 - 104971

ER -