Sensor2Vec: An Embedding Learning for Heterogeneous Sensors for Activity Classification

Junpei Joni Zhong, Ahmad Lotfi

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

4 Citations (Scopus)

Abstract

Based on the idea of word2vec embedding method in NLP, this paper presents a novel idea called sensor2vec which captures the contextual information of the heterogeneous sensory information in the ambient assisted living setting. The contextual information is essential in order to classify and understand the human activity using multi-modal sensory data. In the activity classification, the sensor2vec embedding method is able to do the pre-processing which produce the embedding layer which represents the semantic value in the high-dimensional space. The preliminary experiment based on LSTM shows that the sensor2vec performs better classification result than the one-hot inputs.

Original languageEnglish
Title of host publication2020 International Symposium on Community-Centric Systems, CcS 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728187419
DOIs
Publication statusPublished - Sept 2020
Externally publishedYes
Event2020 International Symposium on Community-Centric Systems, CcS 2020 - Hachioji, Tokyo, Japan
Duration: 23 Sept 202026 Sept 2020

Publication series

Name2020 International Symposium on Community-Centric Systems, CcS 2020

Conference

Conference2020 International Symposium on Community-Centric Systems, CcS 2020
Country/TerritoryJapan
CityHachioji, Tokyo
Period23/09/2026/09/20

Keywords

  • Activity Classification
  • Ambient Assisted Living
  • Embedding Learning
  • LSTM

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Information Systems and Management
  • Mechanical Engineering
  • Control and Optimization
  • Health Informatics
  • Instrumentation

Fingerprint

Dive into the research topics of 'Sensor2Vec: An Embedding Learning for Heterogeneous Sensors for Activity Classification'. Together they form a unique fingerprint.

Cite this