-
公开(公告)号:KR101260728B1
公开(公告)日:2013-05-06
申请号:KR1020110078436
申请日:2011-08-08
Applicant: 국민대학교산학협력단
Abstract: 본발명은전자코 시스템에서특징되먹임을이용하는시간구간선택방법에관한것이다. 이를통해, 처리시간, 메모리, 전력소모를개선할뿐만아니라특히입력데이터에서불필요한부분을제거하고분류에중요한데이터만남도록데이터를재구성함으로써원본데이터전체를사용했을때보다도오히려더 높은인식률을나타낼수 있는기술에관한것이다. 본발명에따른시간구간선택방법은, 주성분분석(PCA) 과정; 선형판별식분석(LDA) 과정; 주성분분석과정과선형판별식분석과정의순차적진행이후에, 전체입력된원본데이터중 주요시간구간을선택하되, 선택된시간유닛들을합쳐서연속된시간구간을생성한뒤, 그연속된시간구간을이용하여원본데이터를재구성하는시간구간선택과정;을포함하여구성된다.
-
公开(公告)号:KR1020130016478A
公开(公告)日:2013-02-18
申请号:KR1020110078436
申请日:2011-08-08
Applicant: 국민대학교산학협력단
CPC classification number: G06K9/6232 , G06F17/10 , G06K9/6234 , G06K9/6247
Abstract: PURPOSE: A method for selecting a time section using feature feedback in an electronic nose system is provided to minimize the processing time, memory, and power consumption by reforming data through the selection of a main time section from the entire input data. CONSTITUTION: A consecutive time section is generated by selecting a main time section from an entire original data inputted after the sequential progress of LDA(Linear Discriminant Analysis) and PCA(Principle Component Analysis) processes and combining the selected time units. The original data is reformed using a consecutive time section. The PCA process generates a first projection vector for the original data, a first feature mask for the first projection vector based on a first threshold value, a single mask by merging the first feature masks, and transforming data in an area having reduced noises based on the single mask.
Abstract translation: 目的:提供一种使用电子鼻系统中的特征反馈来选择时间段的方法,以通过从整个输入数据中选择主时间段来重构数据来最小化处理时间,存储器和功耗。 构成:通过在LDA(线性判别分析)和PCA(原理分量分析)处理的顺序进行之后输入的整个原始数据中选择主时间段并组合所选择的时间单位来生成连续的时间段。 原始数据使用连续的时间段进行重组。 PCA处理产生用于原始数据的第一投影向量,基于第一阈值的第一投影向量的第一特征掩码,通过合并第一特征掩模的单个掩码,以及基于基于 单面膜。
-