Presentation 2010/3/19
Detection of Abnormality by Using Waveform Time-series Classifiers on Low-Quality Sensor Data
Ken Ueno, Topon Kumar Paul, Makoto Sato,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) To detect abnormality in a target object, multiple sensors are widely used in various application domains. However, when many sensors are necessary for simultaneous monitoring of the target object, use of high-quality sensors may not be feasible due to the high costs of the sensors; instead, low-quality sensors are widely used in many industrial applications. In this paper, we propose a decision fusion technique called k-Nearest Neighbor-based Bayesian Classifier(k-NNBC)that enables accurate anomaly detection using multivariate sensor data. In our empirical evaluations on the angular bump related anomaly detection problem using a 3-axis accelerometer, we have found that the k-NNBC outperforms the widely-used k-Nearest Neighbor(k-NN)classifier with Dynamic Time Warping(DTW)distance measure; especially on low-specification configurations, the k-NNBC improves the accuracy by approximately 12.5% on the average, which shows the effectiveness of the proposed method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Anomaly Detection / Decision Fusion / k-Nearest Neighbor-based Bayesian Classifier / Dynamic Time Warping
Paper # SSS2009-35
Date of Issue

Conference Information
Committee SSS
Conference Date 2010/3/19(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Safety (SSS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Detection of Abnormality by Using Waveform Time-series Classifiers on Low-Quality Sensor Data
Sub Title (in English)
Keyword(1) Anomaly Detection
Keyword(2) Decision Fusion
Keyword(3) k-Nearest Neighbor-based Bayesian Classifier
Keyword(4) Dynamic Time Warping
1st Author's Name Ken Ueno
1st Author's Affiliation System Engineering Laboratory, Corporate Research & Development Center, Toshiba Corporation()
2nd Author's Name Topon Kumar Paul
2nd Author's Affiliation System Engineering Laboratory, Corporate Research & Development Center, Toshiba Corporation
3rd Author's Name Makoto Sato
3rd Author's Affiliation System Engineering Laboratory, Corporate Research & Development Center, Toshiba Corporation
Date 2010/3/19
Paper # SSS2009-35
Volume (vol) vol.109
Number (no) 473
Page pp.pp.-
#Pages 4
Date of Issue