[논문]Kernel Fisher Discriminant Analysis for Natural Gait Cycle Based Gait Recognition

Huang, Jun; Wang, Xiuhui; Wang, Jun

doi:10.3745/jips.02.0115

[국내논문] Kernel Fisher Discriminant Analysis for Natural Gait Cycle Based Gait Recognition 원문보기

Huang, Jun (College of Modern Science and Technology, China Jiliang University) , Wang, Xiuhui (College of Information Engineering, China Jiliang University) , Wang, Jun (College of Information Engineering, China Jiliang University)

This paper studies a novel approach to natural gait cycles based gait recognition via kernel Fisher discriminant analysis (KFDA), which can effectively calculate the features from gait sequences and accelerate the recognition process. The proposed approach firstly extracts the gait silhouettes through moving object detection and segmentation from each gait videos. Secondly, gait energy images (GEIs) are calculated for each gait videos, and used as gait features. Thirdly, KFDA method is used to refine the extracted gait features, and low-dimensional feature vectors for each gait videos can be got. The last is the nearest neighbor classifier is applied to classify. The proposed method is evaluated on the CASIA and USF gait databases, and the results show that our proposed algorithm can get better recognition effect than other existing algorithms.

* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.

제안 방법

The paper presents a new gait recognition method based on KFDA, as shown in Fig. 1. First, the GEI is calculated from gait cycle.
The results of experiment prove that our method is effective for gait recognition. The main contributions of this paper include (1) a new feature extraction algorithm is proposed based on natural gait cycles, and the observation vector set is constructed using the extracted features. In order to improve the algorithm robustness, the algorithm also adopts the feature combination of gait image and its region bounded by legs.
The main contributions of this paper include (1) a new feature extraction algorithm is proposed based on natural gait cycles, and the observation vector set is constructed using the extracted features. In order to improve the algorithm robustness, the algorithm also adopts the feature combination of gait image and its region bounded by legs. (2) KFDA method is used to refine the extracted gait features, and low-dimensional feature vectors for each gait videos can be got.
3, we can see the outermost outlines of the human body have changed when human is walking at condition of carrying backpacks and other objects, which can lead to poor gait recognition robustness [19]. Since changes on top part of the body is very little and lower part of the body (such as legs ) has obvious change in normal, carrying and clothing conditions, the paper adopts the method in [20] to represent the gait feature by both gait silhouettes image and energy image in the local outline of the legs. Fig.
Aiming at this problem, we add μI (I represents a unit matrix and μ is a coefficient) to the Q matrix, and Qμ = Q + μI, which can let Q become a nonsingular and use a generalized eigenvalue equation to solve.
Our experiments use 60 subjects at 90º and normal conditions that is each subject has 6 sequences, each sequence contains 2 gait cycles and every sequence have 12 gait cycles. In the experiments, 2 gait sequences and 4 gait cycles are selected for training and another 4 gait sequences and 8 gait cycles for testing. We carried out 15 experiments on CASIA B and calculate the average.
Due to the small number of samples, to obtain the unbiased estimation of the correct recognition rate, leave-one-out cross-validation method is adopted to conduct the experiments. The paper adopts cumulative match score (CMS) [22] to evaluate the experiment results and presents recognition rates of Rank 1 and Rank 5. In order to evaluate the GCI-GEI feature extraction effect and the KFDA dimension reduction and classification ability, the recognition rate of our method is compared to other exiting algorithms, and the results are showed in Table 1.
A database consisting of 1,870 sequences of the 122 subjects based the USF Human ID database [23] is divided into 1 set for training and 10 probes labeled from A to J for test, which based on 3 covariates: normal, walking, and carrying condition. Being different from experiment on CASIA, this experiment adopts weighted mean recognition rate to evaluate the experiment results. To demonstrate the advantages of KFDA, we choose the traditional algorithms LDA, PCA and DCV, and manifold earning algorithm LPP for comparison.
This paper proposes a new algorithm for gait recognition based on GEI and KFDA. The proposed algorithm firstly proposed a new feature extraction algorithm based on natural gait cycles, and the observation vector set is constructed using the extracted features by combination of gait image and its region bounded by legs. Then the expansion vector of the GEI by column is used as the input to get the optimal subspace W_opt and α_opt, the projection on α_opt is calculated by the GEI observation vectors.

이론/모형

Due to the small number of samples, to obtain the unbiased estimation of the correct recognition rate, leave-one-out cross-validation method is adopted to conduct the experiments. The paper adopts cumulative match score (CMS) [22] to evaluate the experiment results and presents recognition rates of Rank 1 and Rank 5.

성능/효과

Finally, the algorithm is tested on CASIA and USF gait database. The results of experiment prove that our method is effective for gait recognition. The main contributions of this paper include (1) a new feature extraction algorithm is proposed based on natural gait cycles, and the observation vector set is constructed using the extracted features.
In order to improve the algorithm robustness, the algorithm also adopts the feature combination of gait image and its region bounded by legs. (2) KFDA method is used to refine the extracted gait features, and low-dimensional feature vectors for each gait videos can be got. (3) Experimental results on CASIA and USF gait database show that even when human body have changed in walking at condition of carrying backpacks and other objects, the proposed method can still get better gait classification rates than other existing methods.
(2) KFDA method is used to refine the extracted gait features, and low-dimensional feature vectors for each gait videos can be got. (3) Experimental results on CASIA and USF gait database show that even when human body have changed in walking at condition of carrying backpacks and other objects, the proposed method can still get better gait classification rates than other existing methods.
As can be seen from Table 1, the average recognition rate of our algorithm is best. We can find that our proposed algorithm can get more than 90% recognition rate in Rank 1 and Rank 5, which testify that the method based on GCI-GEI and KFDA can identify better the different human gait and get better recognition effect than other algorithms in the small sample gait database. Furthermore, the experiment data shows the recognition rate of GEI+LBP+DCV is lower than GEI+PCA, which explains reducing the dimension by PCA will lose important gait recognition information after LBP feature extraction from GEI.
We can find that our proposed algorithm can get more than 90% recognition rate in Rank 1 and Rank 5, which testify that the method based on GCI-GEI and KFDA can identify better the different human gait and get better recognition effect than other algorithms in the small sample gait database. Furthermore, the experiment data shows the recognition rate of GEI+LBP+DCV is lower than GEI+PCA, which explains reducing the dimension by PCA will lose important gait recognition information after LBP feature extraction from GEI.
Comparing with other algorithms, the GCIGEI+ KFDA approaches can improve the recognition rate by about 6%. In addition, since the experiment is based on normal, walking and carrying condition, the experiment results demonstrate the proposed algorithm can eliminate effects of walking and carrying conditions and have great robustness.
At last, the feasibility of the algorithm is verified by experiments on CASIA and USF. Comparing with other 5 existing algorithms, the results prove the proposed algorithm can get better recognition effect and have better robustness than other existing algorithms in the small sample gait database.

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증

[국내논문] Kernel Fisher Discriminant Analysis for Natural Gait Cycle Based Gait Recognition 원문보기

Abstract ▼ AI-Helper

Keyword

표/그림 (7)

표/그림 (7)

AI 본문요약
AI-Helper

제안 방법

이론/모형

성능/효과

이 논문을 인용한 문헌

활용도 분석정보

활용도 Top5 논문

관련 콘텐츠

원문 보기

원문 URL 링크

오픈액세스(OA) 유형

이 논문과 함께 이용한 콘텐츠

AI-Helper ※ AI-Helper는 오픈소스 모델을 사용합니다.

선택된 텍스트

연합인증

[국내논문] Kernel Fisher Discriminant Analysis for Natural Gait Cycle Based Gait Recognition 원문보기

Abstract ▼ AI-Helper

Keyword

표/그림 (7) 모든 표/그림 보기

표/그림 (7) 슬라이드로 보기

AI 본문요약 엑셀 다운로드 AI-Helper

제안 방법

이론/모형

성능/효과

이 논문을 인용한 문헌

활용도 분석정보

활용도 Top5 논문 더보기

관련 콘텐츠

원문 보기

원문 URL 링크

오픈액세스(OA) 유형

이 논문과 함께 이용한 콘텐츠

AI-Helper ※ AI-Helper는 오픈소스 모델을 사용합니다.

선택된 텍스트

표/그림 (7)

표/그림 (7)

AI 본문요약
AI-Helper

활용도 Top5 논문