Presentation | 2014-01-23 Early recognition for improving classification performance in the initial stage of the sequences Jun SUGIMOTO, Yasukuni MORI, Ikuo MATSUBA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | The early recognition is a problem that outputs a recognition result without waiting for the sequential end in the recognition of the time sequence. The early recognition is used for a system required for the forestall processing, and it is demanded to have high performance at the final stage of sequence, as well as at the early stage of the sequence. Therefore, instead of building an classifier for the whole sequence, an inherent weak classifier for each time is built and put together. This method has high performance for notonly the whole sequence but also for the sub sequence. However, this method assumes a shorter data set than the value setted at the time of learning, so it can not be used for practical early recognition, since the length of the sequence is defferent for each sample data. Therefore, this study proposes a method that applies time series modeling to the original method, so it can be applied when the sequence length is unknown and can perform the most suitable recognition for a sub sequence. In addition, an early recognition experiment of the online letter is performed to prove usefulness of the suggested method. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Early recognition / Time sequence / Boosting / Hidden Markov Model |
Paper # | PRMU2013-100,MVE2013-41 |
Date of Issue |
Conference Information | |
Committee | MVE |
---|---|
Conference Date | 2014/1/16(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 | Media Experience and Virtual Environment (MVE) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Early recognition for improving classification performance in the initial stage of the sequences |
Sub Title (in English) | |
Keyword(1) | Early recognition |
Keyword(2) | Time sequence |
Keyword(3) | Boosting |
Keyword(4) | Hidden Markov Model |
1st Author's Name | Jun SUGIMOTO |
1st Author's Affiliation | Graduate School of Advanced Integration Science, Chiba University() |
2nd Author's Name | Yasukuni MORI |
2nd Author's Affiliation | Graduate School of Advanced Integration Science, Chiba University |
3rd Author's Name | Ikuo MATSUBA |
3rd Author's Affiliation | Graduate School of Advanced Integration Science, Chiba University |
Date | 2014-01-23 |
Paper # | PRMU2013-100,MVE2013-41 |
Volume (vol) | vol.113 |
Number (no) | 403 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |