@inproceedings{a1eaf456d0444de6ad11bf87f7586f8d,
title = "Dense bag-of-temporal-SIFT-words for time series classification",
abstract = "The SIFT framework has shown to be effective in the image classification context. In [4], we designed a Bag-of-Words approach based on an adaptation of this framework to time series classification. It relies on two steps: SIFT-based features are first extracted and quantized into words; histograms of occurrences of each word are then fed into a classifier. In this paper, we investigate techniques to improve the performance of Bag-of-Temporal-SIFT-Words: dense extraction of keypoints and different normalizations of Bag-of-Words histograms. Extensive experiments show that our method significantly outperforms nearly all tested standalone baseline classifiers on publicly available UCR datasets.",
keywords = "Bag-of-Words, BoTSW, D-BoTSW, Dense features, SIFT, Time series classification",
author = "Adeline Bailly and Simon Malinowski and Romain Tavenard and Laetitia Chapel and Thomas Guyet",
note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2016.; 1st ECML PKDD Workshop on Advanced Analysis and Learning on Temporal Data, AALTD 2015 ; Conference date: 11-09-2015 Through 11-09-2015",
year = "2016",
month = jan,
day = "1",
doi = "10.1007/978-3-319-44412-3\_2",
language = "English",
isbn = "9783319444116",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "17--30",
editor = "Ahlame Douzal-Chouakria and Pierre-Fran{\c c}ois Marteau and Vilar, \{Jos{\'e} A.\}",
booktitle = "Advanced Analysis and Learning on Temporal Data - 1st ECML PKDD Workshop, AALTD 2015, Revised Selected Papers",
}