USF Libraries
USF Digital Collections

Gait-based human recognition at a distance

MISSING IMAGE

Material Information

Title:
Gait-based human recognition at a distance performance, covariate impact and solutions
Physical Description:
Book
Language:
English
Creator:
Liu, Zongyi, 1975-
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
LDA
Eigen-Stance
Population HMM
baseline
gait biometrics
Dissertations, Academic -- Computer Science and Engineering -- Doctoral -- USF
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: It has been noticed for a long time that humans can identify others based on their biological movement from a distance. However, it is only recently that computer vision based gait biometrics has received much attention. In this dissertation, we perform a thorough study of gait recognition from a computer vision perspective. We first present a parameterless baseline recognition algorithm, which bases similarity on spatio-temporal correlation that emphasizes gait dynamics as well as gait shapes. Our experiments are performed with three popular gait databases: the USF/NIST HumanID Gait Challenge outdoor database with 122 subjects, the UMD outdoor database with 55 subjects, and the CMU Mobo indoor database with 25 subjects. Despite its simplicity, the baseline algorithm shows strong recognition power. On the other hand, the outcome suggests that changes in surface and time have strong impact on recognition with significant drop in performance.To gain insight into the effects of image segmentation on recognition -- a possible cause for performance degradation, we propose a silhouette reconstruction method based on a Population Hidden Markov Model (pHMM), which models gait over one cycle, coupled with an Eigen-stance model utilizing the Principle Component Analysis (PCA) of the silhouette shapes. Both models are built from a set of manually created silhouettes of 71 subjects. Given a sequence of machine segmented silhouettes, each frame is matched into a stance by pHMM using the Viterbi algorithm, and then is projected into and reconstructed by the Eigen-stance model. We demonstrate that the system dramatically improves the silhouette quality. Nonetheless, it does little help for recognition, indicating that segmentation is not the key factor of the covariate impacts. To improve performance, we look into other aspects.Toward this end, we propose three recognition algorithms: (i) an averaged silhouette based algorithm that deemphasizes gait dynamics, which substantially reduces computation time but achieves similar recognition power with the baseline algorithm; (ii) an algorithm that normalizes gait dynamics using pHMM and then uses Euclidean distance between corresponding selected stances -- this improves recognition over surface and time; and (iii) an algorithm that also performs gait dynamics normalization using pHMM, but instead of Euclidean distances, we consider distances in shape space based on the Linear Discriminant Analysis (LDA) and consider measures that are invariant to morphological deformation of silhouettes. This algorithm statistically improves the recognition over all covariates.Compared with the best reported algorithm to date, it improves the top-rank identification rate (gallery size: 122 subjects) for comparison across hard covariates: briefcase, surface type and time, by 22%, 14%, and 12% respectively. In addition to better gait algorithms, we also study multi-biometrics combination to improve outdoor biometric performance, specifically, fusing with face data. We choose outdoor face recognition, a "known" hard problem in face biometrics, and test four combination schemes: score sum, Bayesian rule, confidence score sum, and rank sum. We find that the recognition power after combination is significantly stronger although individual biometrics are weak, suggesting another effective approach to improve biometric recognition.The fundamental contributions of this work include (i) establishing the "hard" problems for gait recognition involving comparison across time, surface, and briefcase carrying conditions, (ii) revealing that their impacts cannot be explained by silhouette segmentation, (iii) demonstrating that gait shape is more important than gait dynamics in recognition, and (iv) proposing a novel gait algorithm that outperforms other gait algorithms to date.
Thesis:
Thesis (Ph.D.)--University of South Florida, 2004.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Zongyi Liu.
General Note:
Includes vita.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 147 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001498175
oclc - 57716599
notis - AJU6778
usfldc doi - E14-SFE0000529
usfldc handle - e14.529
System ID:
SFS0025220:00001


This item is only available as the following downloads:


Full Text

PAGE 1

!" #$ # ## % % & #' ()* #++ #++ #++ #++ #,#++ / 012113 ,((% # 2113

PAGE 2

. 4

PAGE 3

.# ## !5'36721 11011899:+ #* ) + # # # 8 -# # # # + / ; %% &' # # # + ( < *# #* #=### #+ .# & + #*+ +>### +

PAGE 4

.<'. % .<''&% !.! = !.%0 /. <&.
PAGE 5

8+8+8 .# = 3D !.%3 ( .<'%(%/. .
PAGE 6

7+8+0 ( 000 7+8+2 ( 008 7+3 008 7+@ 00D !.%D
PAGE 7

. 2+0 !# # + 6 2+2 / $ % # 5 :#5! : B-5 : 5'/: ?# # + 0D 2+8 .#% ##%= + 09 2+3 .#% #%= (&(-#2@)+ 22 8+0 #%= &' + 8@ 8+2 *% ##+ 87 8+8 B B G 0H !# B 9 + 89 8+3 (& ( C !#+ 86 8+@ (". 01 -. # # # < "1+1@+ 32 7+0 (' + 018 7+2 .#. "#%= # &' #IJ .+ 017 7+8 "#B %= + 000

PAGE 8

. 7+3 / )E' E % #( # # + 00@

PAGE 9

'0+0 %= #<+ 2 '0+2 "!#? *+ 3 '2+0 #&'!$+ 0@ '2+2 5: # 5: ## 5: # 5:## + 07 '2+8 &( -# # ) ? !.# # <+ 06 '2+3 (& ? #(#5: -? 5:' ? *5: -? *+ 21 '2+@ #(&(!$+ 20 '2+7 &+#-# 5011:+ 20 '8+0 .#' ##!##B 50 2:+ 23 '8+2 =B5: 5:# 5: 5:# +2@ '8+8 .# -# -#' -#B %+ 29 '8+3 K#/ = ###+ 81 '8+@ # 02 %= &' #( + 83

PAGE 10

'8+7 # 02 %= &' # <&' !21H+ 83 '8+D .## # B + 30 '8+9 C!+ 38 '8+6 ) 5:5:% 5: 5: # ( " 5: 5: + 33 '3+0 ( # <-# #!# + @0 '3+2 -# -## C B-+ @8 '3+8 %=D + @@ '3+3 B -# / ( &.C D0)+ @6 '3+@ #'% #( + 71 '3+7 .#. -# F # # -# -##+ 70 '3+D # <5:5:! + 72 '3+9 # 5: = # / = 5: / = # / =+ 78 '3+6 # = 5L :' B 5L :# + 7@ '3+01 .##$-#81 B-!C#&# %(+ 7D '3+00 !' #. #<+ 79

PAGE 11

'3+02 5:<5:# < <)#. #+ 76 '3+08 5: < 5: # < 01 ) # #+ D1 '3+03 #! # "*0B !0H-#( # -#5&: # #+ D0 '3+0@ "5 :*0B "5 : 0H' !5 :-# -#! -# % = % 5 < !:#&# !#+ D8 '3+07 "5 :*0B "5 : 0H' !5 :-#-#! ##!#+ D3 '3+0D = 5L : B 5L : # ##(.#+ D7 '@+0 %= # # < )M % # C + D6 '@+2 @,%= #&'A/. (55: 5:: < 55:5::-#022 )+ 90 '@+8 %= /E < )#5: # ##5:B-5:# 5:5:5:.5= ( #:+ 93 '@+3 .# B # % !-#% ### %(+ 9@ '@+@ # #@,%= &'A/.+ 97

PAGE 12

'@+7 .# # # !# /E+ 9D '@+D %=#! #+ 60 '@+9 #/E !##.%= # #5-#022):+ 63 '@+6 # /E 5/-: !# -# <#!#+ 6@ '@+01 # %= 5B: 5#. : 5: 5: 5.: # < # 5D0)( :+ 67 '@+00 .#. "#&(5%= 0@@):+ 6D '@+02 .#. "#(&(5%= 8+02@):+ 69 '@+08 .# # @ # %= ## /E!#+ 011 '@+03 .# .( ) % #' !## 88)#+ 010 '7+0 #( #+ 01@ '7+2 5< 0211: # %( # #'%%.2111+ 017 '7+8 .#2' /( #+ 019 '7+3 .#' C + 016 '7+@ < 5%= : 5%= : N' 5%= : -#B #+ 002 =

PAGE 13

'7+7 <' 5%= : 5%= :N' 5%= :* ( # !-#B #+ 002 '7+D &C + 003 '7+9 B "' !@H ( + 00@ '7+6 .## !@H+ 007 =

PAGE 14

# # # # # + # # # + # ## + ? # -# # # #E #+ < = -# # # &'A/. # -#022)#&(-#@@)# (&(-#2@)+ ## # + <##### # # -#" + # # C K # # (* ( 5((: -# # -#% E# !5!:# # #+ # # D0 )+ $ # # # # ((#B##) # % + ? # # # =

PAGE 15

#$ + /### #* # + #+ ### 5: # # # #E -# # # -###M5: ##E((#% K# M 5: # # E (( % # # 5 !: # # #+ .# # # + -## # # "5E 022):# 22H 03H 02H + # -# + ? # I* -J # # + ? "# # ##* #C # + .# #*5:##I#J 5: ##= # 5:# # #5: ####+ =

PAGE 16

!" "#$% &"% .# # =+ .###*++# #+ .#)>#O89P -# # # # # # # + ,E -* O0DP # # # =+ ### = # -## E OD33PO890682P # OD8P+ # # # = O92P -# # # # # # -# # # # =+ # + = O79P= C # + .## # #" ####*# # # C # #*+ # + O2D 20P # C # # "+ .# # # = # E 0

PAGE 17

5: 5: 5: 5: 5: 0+0+ %= # < 5: # 5:
PAGE 18

#++-## -##QI J+ ### # E# O2267P+ = # # O80P+ ? # # + -# -##O2PM-##28 = O23P+ = O78P* 081 # # # E59 : 539 : 573 :507:+ .#-"### C #5:5:-# *5:-# *+ .#* )## + # # C # # ) ODDP+ ## "5:# O7@ @3 90 D 9@ D6 81 7 @8 370@68963@93079786P+ -# # ## #$-# $)+ < C *=C # # ) $ # + # # + 8

PAGE 19

'0+2+ "!#? *+ 3

PAGE 20

' () "*( &&#& $$&&$ ) "+& "# '+0+2 # 5: 5:#$ 5: # + #2-##-# # 8 # #+ ? #&'A/. -# 022 ) &( -# @@ ) (& ( -# 2@)&+#5<.
PAGE 21

+ ? # # # # $ #5 : 5 :+ # # # + < # # # # # =# #$* ## *+ # # # # I#J = ##$ + ##* ##+ #@# + ? # #+ # #+ -####E+ # + < # E (* ( 5((: % # + .# # # # I#J #+ .### (( C -## + #% # 5 !: # -# # -#) # #E C )+ .# # -#-#)-#C## + ? # -# # #" # -# #+ #&'A/. E -## # #&( 5 C:# (& (5 C:+ # 7 # # -# + 7

PAGE 22

E-#%+ # % ## ( # # 5%(: ?* O71P+ % # # # # O77P+ .# = # -#I#J #+ ? = -# # *+ .## -####" # # # -# #+ # # 5N: # 5N N:# # + #" ## 0+ ? # # # I#J + .# C # # + = + B81 # 2 # # --+ 2+ ? # # $ # C + F # *# + 8+ ? -## # # #+ ? #) # KE#+ 3+ ? #C # + # # -#" + D

PAGE 23

' # # + 2+0 E # # # # #C -# ## -+ !-# + .# # $ # + # "+ " # # -# ++ #+ -## # # #E ++ #+ # # ( ##5(:O30P-# # 5 : -# + # # <#5 <:+ < # # ####+ --!( +& # # # # # #21# 5: 5:# #5: #+ 9

PAGE 24

. 2+0+ !## + !# E *0 5): 5&' 0:O3832P < < < B# D6H 77H 26H D0 D0 D0 5&': O60 62P < < < B# 79H 70H 02H D0 D0 D0 # 5!:O6863P < < < .5 : B# 60H D0H @6H 83H 29 D0 D0 D0 ( 5<.
PAGE 25

' .-!/ /0).) !&$ --!( +& .# # # $ # + .# # = # -# # #-# # # + .# # $## $ #+ .###+ # #* 5&'A/.:# #0-### O3832P+ .##-##+ $ # 5# : $ = #+ % #$-###$5#: # = + .# # ##. ++ # =# ## # #+ .# # = # $ # # + .#* # $# # # $ #+ ## # + O6062P5&': # 5':+.## 5: $ # # 5: = K # # 5: 5':# !5: ) #+ O6863P5!:## #-# ## # #### + 01

PAGE 26

.# #E-##E+ .# # !+.$# E% + .#&(R ###-# #O373DP+ .# -# "# #-# # -+ .## ## E "=#+ .## =# = -# # + $ # (* + # ## $ + .##C # # # ##+ .# (( # ? # # # $ #(( #B #+ .# # ((# # = # + # # O61P # # -# # + # = # # E+ .#-### #((-## + ? (( *+ # C+ # ((M (( -# # * + # (( M ) # $+ .# # + O@8P## + .## D#C + % # + ## ##$# 00

PAGE 27

# # # *$ + O@2P # # = -# (( $-# + %O@1P" $+ ## # + .## # #E+ #5B(: ""+ '' +!!&$ --!( +& .# ##E### + < # # # $ + .# # # # ## O@9P+ )# % # # # $+ .# # # + # #-### O26P+ ) # # # # $ #! 5 !:+ .# # #E = # #+ # + /= 5<.
PAGE 28

#O98010P+ )R -# + .# # $ + !# # # 9@H "00@+ !# # # # $ # $ ) # #O0@96P5(&:+.##" E#E #" + .##-## # ## # ## # # # + .# # # -$ #$M#+ ##### + <#5EN !N#: # # # #E # # # + # # # #" + #(& ## #* #(& #=# #+ '1 "!"( !!."& .## ## #E # # O86P+ # # E ) -##O869@97P+ # ## ## #8+ 08

PAGE 29

'' !" !"!2!& $ # $ # #" + # ) -+ #01 -#E 7)022)+ !## # + '' 3 #.!) !" +!//)0 !"!2!& .#&'A/.# # 022)82C + .# # -# #+ # # # + # #+ #+ '' !"!&" .# # & # ( 2120 / 0@07 2110+ # + .# -# ) 5: + .# # # ) *# E #+ .# '+ 2+0+ .# = 0@ # ) = @ # =+ # + < Q + .# #-+ % #-## =81 ##-# ) # # + ?# # ##-= + '+2+2# 03

PAGE 30

15.85 mElliptical Path 1.7 m CC13.5 m 13.5 m 4.7 m Calibration Board '2+0+#&'!$+ ### + .# # ) = # + !## # # # = # # #+ ) # # # # C #+ ##)* C # + )* *# # 5 = 7 *:+ ( ) # #+ ## # # # 0: M2: # M 8: # M 3: / M @: # $ ( / 0@

PAGE 31

5: 5: 5: 5: '2+2+' 5:# 5:# # 5: # 5: # # + ( /+ .# 88 ) -# # ( / ## #=# + 2+2# # $ )# # C + ''' !/#!") &( +. .# #= "#+ .# = # C + .#" # + .# # # # = # #= + 07

PAGE 32

. 2+2+/ $% # 5:# 5!:B-5 :5'/: ?# # + .# / ?# # .+ # (/ / / 5 : 020 88 / 5 : 71 5 : 020 5 : 71 / 5 : 022 88 / 5 : @3 5 : 021 5 : 71 ? # # # '%%. O30P+ # # # # E+ #$+ = # = "= # + .#=#CC + .#$ -# # # # / ( -# # # ) / + .# # # + .# # 2+8+ .# = # + .# ### $)####+ ? # ## #)# # = # # # # # )M ) I# !J # ( # # #I# !J# / + .# -# 2+8 !A 0D

PAGE 33

. 2+8+ .# % # # %= + .# # %= 5!/( 5! '(/: 021 B. /# "# "% 00 % 5!A//: 88 +.5##: # = + ##-## ( -##/ + .##= -# OD@P+ ) #* ##+ '1 !"!2!& .# &( 0 $ 2@ ) 2 $ @@ ) .# # + ## 2* 5#8% :## 3+@D+ '+2+8# + % # $ # = 01 + .## + ( # #C C-## 09

PAGE 34

' 2+8+ &( -# # ) ? .# # <-##B&+ + .# C -# # # 5: # # &( 21 5A: ## #81A5:#3+@D ##&(0+7@### #+ .# &( 2 C -# # 5 : C @@ )+ # &( 1 &2' -# #$*C + # # = # 3// /+/ + '4 2 !"!2!& &* #&(#(&( -# #* # + # #+ Q+ .#(&(O2@P$2@)* #+ '+2+3# -+ % #) C -*52+17A#:* 06

PAGE 35

5: 5: 5: 2+3+ (& ? # ( #5: -? *5:' ? *5: -? + 52+92A#: -*#*+ % #$ 0081+ =* -*-'+2+@# -+ .#"#O27P 0+ -##-# #S 2+ -##-# #-S 8+ -##EC #-S 3+ -##EC -S # = "+ 2+3# #= -# #-## #"=+ 21

PAGE 36

'2+@+#(&(!$+ 5: 5: 5: '2+7+&+#-# 5011:5: /. *5:/. 5:/<+ 20

PAGE 37

. 2+3+ .# % # %= (& ( -#2@)+ .#%= /?#/ "=+ %=+ C ? ? BB0+0 0+2 ' 0+8 2+0 2+2 ! 2+8 ' 8+0 8+2 ! 8+8 ' 8+3 8+@ ! 8+7 ' 3+0 NB' N N 3+2 NBN N 3+8 NB' N 3+3 NBN '5 ) &" % #"+!.-") 67 !"!2!& .# # ) + Q # + 3// 4/ / + .# 011 )+ #2110+ !'+2+7# -)8 + .# -#C $##7# )+ .# $ 5:-## -# 5: # # #C # ) #"+ 22

PAGE 38

1 8 8 6 '7 .# # + E # # *S -#= C S #$ C # I J #+ # # # #52:+ #$ # $ + #E #+ & # # # &'A/. # = # @ # # (& ( = ##C*+ ## # = + ? #C #+ < )+ 1 !!."/&& !&/) /0"+. 6'7 .##2OD@P #"0O3832P+ .# $ # A ## E + .###" K+ '+ 8+0 # Q # # # # #+ 28

PAGE 39

Silhouette Computation: Silhouette Estimation: 1.Compute the Mahanalobisdistance of each pixel within the bounding from the mean background image. 2.Smooth the distance image by two-pass 3x3 average filter. 3.Threshold the smoothed distances based on an user specified D MAHA 4.Use Expectation Maximization (EM) to estimate the silhouette from the distances 5.Filter out small regions (< N SIZE ) 6.Keep just the largest connected region 7.Center the silhouette in the horizontal direction by consideringthe upper half of the silhouette 8.Size-normalize so that the silhouette occupies the whole length of the image and save it as 128 by 88 sized PBM images. Probe Gallery Similarity Computation: 1.Break up probe sequence into K subsequences of Nprobeor Ngaitcontiguous frames each. 2.For each probe subsequence, estimate the maximum correlation with the gallery sequence. 1). Shift probe sequence with respect to gallery sequence 2). Compute distance between frame pairs #pixels in AND-ed silhouettes / #pixels in the OR-ed silhouette 3). Add the distances 3.Pick the median of the maximum correlations of the probe subsequences as the similarity measure. Gait Period Detection:1.Consider the number of silhouette pixels mostly from the legs (bottom half of the silhouettes) vs. time. 2.Detect the local minima in the above plot 3.Compute the median of the distances between minima, skipping every other minimum --two possible medians, depending on whether we skipped the first one or not. 4.Take the average of the medians as the gait period (Ngait). 8+0+ .# # # !# # B 50 2:+ .# &$# E!#50:& &$ # !#52: # #+ 23

PAGE 40

5: 5: 5: 5: 8+2+ = B5: 5:# 5: 5:# + 2### #+ .#" = # # $+ .# = # # =+ .# # # # #+ .# #$+ .## # $+ # = # + # # =# $+ .# = # # # # # # =+ .# =*-## -# # # = #+ '+ 8+2 # =## =+ .# = *=# # =+ .# =#-##+ 2@

PAGE 41

1 /+#"" 9"!(") .# # # = # # # =+ # # # = + = # # # + # ## $ # # # 5 : = # = #+ ? # 5 : # T 5 : # # = + = -# # = # # (# # = # + # (# = *+ # # O38P # "= ##+ .# # # # * ## = =E5%(: + # = = # K # (# + ? # G G (= ( 5((: 5 : G 5 : 5 : -# # # 5 : G + # = # 5 :+ ? # # %( $ -O21P+ .#C ## + 5 : G 5 : G 5 : 5 : G 5 :5 : 5 : 5 : G 5 5 : 5 :: 5 : 5 : 58+0: 27

PAGE 42

.#%(E ## M##(#=## #+ 5 :G 50 1 2@@: 5 :G0 5 : 58+2: ? # -##E # E$ *-# 0@+ # # + ? # ###(# 5:6 6 # "$ 8 8 "#$ ## + .# # B + .# # %( -####-# ##(#+ .# ## %(+'* #-##+ # ### 029=#-### #029 99= E # + .# # # C # + ? # # ##E # # =+ .# # # # #E # # ###33+ # $ ## -# #+ '+8+8# # + 5:# -#$5: #) -###" *5: ) # ## Q# $ 2D

PAGE 43

'8+8+ .# # -#' -# B %+ .# # # # # -+ # $ # # 5: B ##+ .# # # # # ## #+ <* # $#+ ###* #)+ # ##### # -# # # # ) E+ .# ####* #)+ <# # # # -# # # # # # -# ## = #+ 29

PAGE 44

1' !" $ "(") .# = # # + + ? # =# # # 5 :+ .# # = -# # #5: -## 5## :+ # # = # -# # # # # #+ '+ 8+3 # # 5 :+ / # + ? # # # + &# #-##*#" + ? # # # + / # # --# ### -+ # -# + .# # -# # # # M81 + 11 ./!" % .-#"!") .# # # $+ + $ G 50: 5 : $ G 50: 5 : + .# #) #-## $+ # $ ) $ -# # # $ # + / # # + # # $ G 5 : 55 N0: : + .#$ G 50: 5 : 26

PAGE 45

0 50 100 15 0 1400 1600 1800 2000 2200 2400 2600 2800 3000 3200 Frame NumberNumber of foreground pixels '8+3+ K#/ =# ##+ # = # $ # # #+ / # $ + ? # # #$ -## $ + .## # $$ + ## ## # 5 5 : 5 :: -# # ## =# #+ .# # O21P+ .# # = # /5 : # # 5 5 : 5 ::G /5 5 : 5 :: /5 5 : 5 :: 58+8: 81

PAGE 46

/ # # # # # -## + .#=# 5 :5 :G 5 5 N : 5 N :: 58+3: # # # # = # $ -# # # $+ <# # # # = + .# # $ $ # -# # *E# *+ 5 :G( = 5 :5 : 58+@: 1' .!)( !!."/&& !&/) /0"+. 6'7 .# ## # # = # I J = #+ ? # # # # #(&(2+3+ .#### = &' -# # ) # # C + # # = # # = + ? # # # # # # ) #+ 80

PAGE 47

1' !& &#/"& .# # # = -+ ? # #$#$# = -#E## $ #E+ -# # '%%. O30P # " # # 5(: # 5 <: + # # # + $ # -## + # =# # # =+ # # -# # ## "* + # # # "5 :#* + # # ) -# # + < -# + .# # # -# # $ # # + .## # # ###-)+ .# # ## # -# ### # + = -# # 5' B. 2112 O31P: # E ## # # + .# # # + ' B. 2112 # E + 5 5 : ) # # # 82

PAGE 48

+ ? E+ .# O31P-# # 5 :G 5 : ( 5 : ++ 5 : 58+7: -# ++ + # # E -# # # = # # -# E + .#(!-# # (! 5 :G 5 : ( 5 : ( 5 : ( 5 : 58+D: -####5(!:# + .#(! + ##(!E # E + # E ### "5 : + ## ## # # + .# #5 <:+ 8+0 E # # 5 : *0@# "5 :0H01H+ B E (! + "8HD9H*0 02H68H @+ .# # # # E # + 0H # # (! 01H # E #$ + # # 0H "# E+ ?## "0H7H92HM 01H 23H63H+ .# 88

PAGE 49

0 5 10 15 2 0 0 10 20 30 40 50 60 70 80 90 100 Rank Identification Rate Exp. A Exp. B Exp. C Exp. D Exp. E Exp. F Exp. G Exp. H Exp. I Exp. J Exp. K Exp. L 8+@+ # 02 %= &' #( + 0 5 10 15 20 0 10 20 30 40 50 60 70 80 90 100 False Alarm RateVerification Rate Exp. A Exp. B Exp. C Exp. D Exp. E Exp. F Exp. G Exp. H Exp. I Exp. J Exp. K Exp. L 8+7+ # 02 %= &' # <&' !21H+ 83

PAGE 50

. 8+0+ # %= &' # 0 @ # B 0H 01H &E5&/: / 5/: (!/ 5(!:+ + %=+ C 5H: 5H: 5H: G0H G01H 0 @ &/ (! / &/ (! / BD8 99 @2 91 92 90 63 63 # D9 68 39 91 9D 92 63 63 #B39 D9 82 @D 7@ 76 96 63 82 77 23 87 33 70 91 91 % # 22 @@ 07 88 8@ @2 D7 D7 B0D 32 01 22 21 3@ @6 71 #B0D 89 02 23 29 31 @D @@ 70 9@ 37 79 D2 91 61 60 # @D D9 39 71 7D D7 9@ 9@ > B87 72 22 3@ 39 73 D@ D7 .## 8 02 1 8 7 0@ 2D 23 .## 8 0@ 1 8 7 09 2D 23 # # # #+ .# # = #+ '+ 8+@ 8+7 # ( <# #= + 8+2 # ## ## #&'+ # # # # + ? # 5: #* #C = -## # 5: # # # -# #####+ ? *# # = # " ## ##+ %= -# # C ) # # # %= -# # C # # + .#*= -*# 8@

PAGE 51

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

PAGE 52

# # # -+ = # 5! /(/:##= !#5!/( /:+ ? ## -9 5!:5! : 5: 5 : 5!: 5: 5! : 5 : -# # ++."=/(/+ 8+8E# 0H # # = + .# # # # C + = # N #= -# # # C # + .# # #+ .## # #= + .## #= =# #= + .# 5++: # + .# ## + ###= # # -# # # # + $ # # # = #C #' -# # # # = + .# # # # # # + # 8+8 #-# ##,R 1+67M # = + .# 1 1110 -# # # # # # )+ ) # # ### "#= C # #C + .# -# # = )#+ *#= #-# ? = O6DP+ 8D

PAGE 53

. 8+8+ B B G 0H !# B 9 M # '= # /#/++ /(/+ %= B# B-N # N # N BN B-N # N # N B. .N % > 5!: 92 9D 7@ 33 8@ 21 29 D2 7D 39 7 7 5! : D7 92 @6 33 8@ 2@ 01 D@ 72 31 8 7 5!: @3 97 33 82 07 21 03 D@ @D 23 7 8 5! : 78 99 36 8D 29 0D 21 D2 @6 26 8 7 5: 60 92 70 83 23 09 02 76 @7 39 5 : 96 9D @3 83 80 21 21 76 71 37 5: 79 62 30 29 29 27 22 D9 7D 26 5 : D8 98 @8 83 8@ 07 0D 7D @7 8@ ( D@ 9@ @8 87 26 21 09 D2 70 8D @ @ ( D@ 9D @3 83 81 21 06 D2 71 89 @ 7 ++ 02+9 8+3 9+8 @+7 7+7 8+7 @+6 8+D 3+@ 6+@ 0+7 0+3 # # # # # $+ # -#" O8@P G1 1@ 56@H ": # - 5%=K# %=!KB-: 5%=!KB%=K: %=K %=,K.+ 1'1 .!)( "+ !&/) /0"+. ) 2 !"!&" # E # # #+ .# ### # $ # # + % ## ### # # # # # # -###+ #*## 89

PAGE 54

. 8+3+ (& ( C !#+ -? ? -? -? ? ? (&5#:O0@P 011H 011H 011H D7H &(5((: O3@37P D2H D0H 67H 80H #+5 : @1H (.5( :O@8P 011H 67H 67H @3H 5+. + : 62H 67H 67H D2H "#$%& () #C -# ## C #+ .# # # # (& ( -# # # ##= #*## #+ 2+3 # $ 2@)*##7 C + # = # -+ 8+3 # "C ## = + .# # # #+ ? ###-##+ ? ## ## = $ #+ #= $ -#* ######+ 1'4 !!" :("& ?# # # # S # ####%= # + #+ .# 86

PAGE 55

# # # # # + .# # ### # ##+ .# # # C C # # + .# C )# #+ ? $ # C # # # # C-# #+ .##C##%= -# #C-#)+ # # %= !M#%= %M # %= M # %= !#" 5!A //:+ # #) ##0 2 5 : 5 : + .# # ) L 5 :G 5 : 5 : 5 : $ # C ) + .# # L 5 : # ) # # # # C # + # E # + ##### # ## -## #-# $ + '+8+D# -## # C-# # + / # # ## C+ .# # 1+93 0+@7 2+D8 3+2@ 7+@@ + .# ? = 31

PAGE 56

15 10 5 0 5 10 1 5 0 5 10 15 20 Percentage Score Difference Count Median =1.56 15 10 5 0 5 10 1 5 0 5 10 15 20 Percentage Score DifferencesCount Median =0.84 5:B5:# 15 10 5 0 5 10 1 5 0 5 10 15 20 Percentage Score DifferencesCount Median =6.55 15 10 5 0 5 10 1 5 0 5 10 15 20 Percentage Score DifferencesCount Median =2.73 5: 5: 15 10 5 0 5 10 1 5 0 5 10 15 20 Percentage Score DifferencesCount Median =4.25 5:. 8+D+ .# # # B L 5 : C-#5:B5:#. 5: 5:5:.+ 30

PAGE 57

. 8+@+ (". 01 -. # # # < "1+1@+ + + + + , + + + + / & .0 1 1 1 111'2 11132 1142! 11356 11356 1177! 1!52' $80 11144 11144 1113! 1115 1112' 1111 11!4 1133 11!41 11411 9" : ; ; ; ; ; : : : : : O6DP # # # # # # # 1+ # ##* ##. # + & # # )# # #### # # 1 5-# 1+110: ++ # # # C E+ ? #"## -? =*+ # # M## ## + ? # " # -O8@P+ .#-# ##01 + # 1+1@ # # *C 501 N0:+ 8+@ -# ## # # )+ # # # #--## # + .# 5-#: 5:+ .#-. 8+@### C ### + .# ## + =# 32

PAGE 58

10 5 0 5 1 0 0 10 20 30 40 50 60 70 Difference in gait cycle lengthNumber of sequences Across View Across Shoetype Across Surface Across Briefcase Across Time '8+9+ C!+ # # # # # + '+ 8+9 # # # C # ) # + # # # # # E+ # # # -# # C + .# # # 5 :#*### + 1'5 "#$% !/#& ## "S !#)-# ES #IEJ)S !#$-# ##$*+ # $*# #)C = + # )-# # # # = -# # ) "+ / 38

PAGE 59

5: 5: 5: 5: 5: 5: '8+6+) 5:5:% 5:5:# ( " 5:5: + ) = + ? ) #)" #91H # = # #A#M # 02 # )+ ? ) # # ) # 31H # = M # @7 ) # + .##)EM # @3 ) # + '+ 8+6 # # + # # # # -# ) # E+ .# $ # ++ -###)+ # # + #" #-# + 11 #..!% # # # + ? # # %(" -# ## + .## #&'A/.-# ## )82 #@ + ##(&(+ 33

PAGE 60

11 0)*(!) )$)0& ? # E + < # E -# # # # # (! #+ $ C # E 0HE7H E5= :+' = # # # 39HE9DH E + # # # # # # C # C # + .## # + C # # = 5 8+2+3:+ .# C # # # # # + = #C * ## #) #+<### #-# #$+ .## # $# 7 #C+ ?## C 5#:5#: # # "60H 6@H O68810@P -# # 81H 3@H -# # C # # O@8 07 0@P E + < # # ## $ + .# ###) ### # + # $ # ##* # + # 3@

PAGE 61

# # -# ## + .# # -# + ?# # ) # $ # $ 82H+ # # # ## + .# C # # # -## # # -# + = ##* S # # # # S 11' !" & !( <#$# "+ ? #$ -#+ < ## + # #E # #' B 5' B.: 2112 O31P+ E C # # B.2112" 0H+ &* "E+ # #* + B. 2112 # + # # + # "#* # # + B C @3H@H-# 83H+ 8+0 9DH 21H # = -## $ # + .# 37

PAGE 62

@DH+ # #* C ## M#C@ #+ B 3DH 1H C -# 22H+ %= # -# # # 7 # #+ .##= 7H+ #+ .#' B. 2112 $ + .# # + < # # # # # # # # #+ .# + & # # ##0:# -#2: -#+ 111 + !" ) "9" ## = + <## # + # #E # + # # # + .# #-# # # #+ # # # + .# # + < # # #-#*-#-# # $# #+ 3D

PAGE 63

$ # # = + .# # # #$ # #+ .# # # + .# #C ###+ !# ## ## -# # # + .# # C # + .# # = # ##+ = # C # # # # # C + # # # # # # # + 39

PAGE 64

4 # E # ### + F# -#C+ !# # # S < # S # ## ##-# ## #+ .#= *M # + <####-# # =+ 5 # # # #+: .# # = -* -# Q* + ##$ # + .# $ # # # # 5):+ ?# $ # C # C # $+ .#$ = #C =# -##$ C-## *+ # $ # = # # # # *+ .# ### -#* #$ + <# # # 4 36

PAGE 65

6 #S ## # #-#IJ #" 5: "### 5:-#### IJ((% #+ 4 !) #!/ /+#""& ( # #+ ( # $ # O70 @@P+ ### + !D0)# 5( : # #"+ .#$ #D0 ) # 5: 5$ -# # # -: 5: 5 -# # # -: 5: 5 -## !# -:5:5-##!# -:,5:+ ? "# # # =81 31 + .#### #* ##*## #=# #* + ? *##8 # $ -# + = # # # *-# =-### -# #$ ## *+ ? ) = # ) + ? = # # # # # C + '+3+0# -= ## # # -+ F # * = # # @1

PAGE 66

'3+0+ ( # <-## !# + + # # $ # $ * ) -# # $ # # # # -# # # + # $ #) #) #)+ @0

PAGE 67

'+ 3+2 # # # ) C C + # # ## E # ## # # 029 =+ .# -'+3+2# -### ##* #+ # -##O32P## ## #E # # ###+ .# -+ # = # # # #+ # # = ++ -# R # -# # # + # # 5 # # -: ###E # 5## -# # -:+ .# # ## + # ## # # = # # ##) + 4' $/ !&$ /+#"" ()&"#(") # 5:# -5: #) -### 5: ) # # # Q#$ #$# # 5: # # -# # + '+ 8+8 # =#*+ .##= ## 8+0+0 # # #5 :# 5 :# T 5 : @2

PAGE 68

'3+2+. -# -## C B-+ .#( -# -# ( #+ !# -# -###,& !#+ # # = + & # (# = # = %= (=E5%(:-#=+ # = # # # + # # # # # # # = ) -# # + ? # # # (* (5((: -# %#+ .##(( # # # # ) + .#(( # "# D0 )+ # @8

PAGE 69

# "### # # # #+ .# -# # # 72 $ # 5!: #+ % # $ # # # (* (5((:# ) + % # ## # ## ) # #%+ 4' .)0 "!)( 9.-/!& .# #((##" + .# # # = -# # # # $+ .# # + < #-# #" -+ #E 5'+3+2: # + .## # 5 :G N 53+0: /##-#= )10# # # = # # # # =#* -#. + < # # #M 5 : G 0 5 :+ # # 5 : #. 5 :+ #= # $ + ? 5 :#++ C -# # -## + ? @3

PAGE 70

' 3+8+ %= D + # -# # = # # + # # # # #=C )*+ %= # + .# = # # # 5 : G 0 5 : -# + G0 # =+ .# G 0 53+2: G 5 : 5 5 : 5 :: 53+8: /# ## = M# = + ? # =# #$+ ? -# M01#= + '3+8# -## =D=+ 4'' -#/!") $$) !, $/ 6-7 ? -(( $$ # + (* ( 5((: # 0 # G 5 !" : # = + .#= -# # 5 :G 5 G G : @@

PAGE 71

## -)# = + .# # # G $ 5 : G 0 -# $ 5 : G 5 G : ++ # # # ## + ? # # = # # ##= + $ 5 :G 0 53+3: .# # E # + .# (( -# # O93P + C = # # "M # # + 4'' $/ !!." &".!") ? *$++ G $ + ##=# + #=# # $ ) # # #=+ G 5 : 53+@: .##=#=# R#-# $ ? ##+ # 5 :G U 53+7: @7

PAGE 72

? # = # OD0P # R #+ ) # $+ # $ + # $ # # = # $ # G % % % + .###= & + .# # # 5 :G 5 : # 5 : $ 5 : 5 : 5 : 5 : 53+D: -# G 5 : # # # # *' -+ 5 : G 5 G : G $ 5 : G00 G 5 : # 5 : $ 5 : 2 & 00 53+9: G 5 :G 5 : 53+6: 5 : G 5 G : G 0 G & 0 G # 5 : $ 5 : 5 : G & 0 00 53+01: .# $5%$+3+D3+01: #E#? #$ #* ## =E+ .# = #E # ## # + ? ## # -# #*+ @D

PAGE 73

4''' $/ ; ".)!") ? # #!* 5!:O8P -# #* ##"E !G 2 5 :N2 53+00: -#5 :#(( # $ # #+ .# # # # + '+3+3# !-## C D0) *## *+ ###"21# #$+ # ## C + 4'1 0)"!)( !" $/ .##%# ## # + ? # -# # # = # -# #((+ ? 5!: #+ # # 5-# : # V5 : #E # # + # # = ## + # # $ ## + -# # # =+ # # #+ ? #### @9

PAGE 74

0 5 10 15 20 25 3 0 1.15 1.2 1.25 1.3 1.35 1.4 1.45 x 104 Number of StatesAIC Values Grass Concrete 3+3+ B -# / ( & .C D0 )M < ? 5 : # <#? *5:+ = # + / # # ##V5 :R-#$ + # C %-## -(( #-## # &' + '+ 3+@ # % #+ .# ##91H# + 4'4"!)( !"( +)0 #&)0 )# $ ## # # #((+ .# B # # OD1P+ # @6

PAGE 75

'3+@+#'% # + .#. -&# # ? $ !# # ? $+ #+ # # #$ $## -# # # # = # ###+ /###$ ##((M##((##+ 4'5 ()&"#(") # # # (( ) #V5 :G # + G N 5 : 53+02: .# # # 1 0 # ## #+ #### #E#C-# #* # # #+ ? # # ## ####### 71

PAGE 76

5: 5: 5: 5: 5: '3+7+.#. -# F ## -# -##+ ## # -+ 5 :G 5 : & 5 :G0 5 : & 5 : #53+08: #= # & G1 2 & G1 9+ 41 <#!/" % ()&"#("$ /+#""& ?# # $ ##S #=# IJ =S = S .# $ #+ .# -## ## 2 + # # # 5: # # = 5: # (# = # = 5:##(#6 6-"# ) 5: ## # ==E5%(:-## # 70

PAGE 77

5: 5:! '3+D+ # <5: 5:!+ 72

PAGE 78

0.4 0.5 0.6 0.7 0.8 0.9 1 0 20 40 60 80 100 120 Ratio of correctly added pixelsNumber of sequences 0.4 0.5 0.6 0.7 0.8 0.9 1 0 20 40 60 80 100 120 Ratio of correctly removed noisesNumber of sequences 5: 5: '3+9+ # 5: = #. / =5: /=#. / =+ 5: # + .## # #-### # O@2P+ + ##*# + -# = # "# ## -# 8 0 M## ++ + .# # ( #)# 50: #)# 52: # ( #)# 50:G ( ( #)# 52:G -# # # ( ( + ? # #)# 50: #)# 52:+ # # -# 8 8 # # ##+ .# # # $ # # # -#+ + .# # # + .#%# 78

PAGE 79

#+ '+3+7# -=#$ 5 -: $ # 5 -:+ 5: 5: '+ 3+7 # -## -M5:# --###" M 5: W 5: # = # # *+ '+3+D# -= + ? # # *C #+ 41 9/ / <#!/" % < #% E# =# 5# :# + .# # =# =# =# =+ ###+ ? # $ #-# ## "# # $+ '+3+9 # -# ## # $ -# ## #+ ? ### + .# ###+ ? ##$ = # 5 :5 :+ .# # # = # *+ / # #C -# ### *#* + # + .# -# # # + .## #="+ #-# ## + ? # #$ # $ #$+ 73

PAGE 80

50 40 30 20 10 0 4 3 2 1 0 1 2 3 4 PPPV / PPPV(Before) as % PD / PD(Before) as % Grass Concrete Grass Carrying 3+6+ # = 5L : B 5L :#+ .# # $ # # $ # #$+ '+ 3+6 # -# -# = L G011 L G011 ? # # $ # # C *+ ? # # $ -# + # $ L 1 L 1 -# # #+ ? ##### # 0H # # 21H81H+ 41' 2#&")&& "+ -) !!") # % # C # ##-# 7@

PAGE 81

C + .## #M # C -## # 81 + .# #-# ## % = # $ # # + ? # $ # #+ # # #$-#$ ) + ? # 01 # $ '+ 3+01+ # # # # -+ !# # # -#*+ 411 )!/;!2/" % :) !"!&"& #E # ##. # + !'+3+00# -21)* + ? #"#=# 3+8 -#-# ## # #%+ '+3+02 # -# # )+ '+ 3+08 # # 01 C + ? # # $ ###= + # # #-# ## #% ##+ 44 .-!(" ) !" (0)") ? # ##% ## $+ # # C S < # # # # # ( 77

PAGE 82

5:<# 5:# '3+01+.##$-#81B-! C # & # % (+ % # -#$+ 7D

PAGE 83

'3+00+!' #. # <+ ## -+ ##5:# "# # 5: # # # -# + ? $ ###8+ 44 (0)") !) #!/ /+#""& # # #+ ###((-## #* # # + ) # # # #+ # # # + .# ## + ? # # ## $# # # # + .# ## ##* # # -# + #=#$ 79

PAGE 84

5:< 5: '3+02+ 5: <5: # < < )#. #+ 76

PAGE 85

5:<# 5:# '3+08+5:<5:# < 01)#. #+ % # -)+ D1

PAGE 86

B D H K 0 20 40 60 80 100 PI at rank 1 Manual Auto B D H K 0 20 40 60 80 100 PV at PF = 1% Manual Auto '3+03+ #!# *0 B 0H -# ( # -#5&:!# # + = 5#: 5: 5:,5.:+ ##$## $###$+ < # ###+ # # = # # # + .##= #$ -# ## #+ -# #) #-#5:# # ## #+ '+3+03# # # # # # 0H + ? # # -# # # ## = # # = + .# 5:(/ + .# D0

PAGE 87

# 1 6 + <###" #= 521H*:+ .#-# ## @*# # M # ## -# # # ++ # -#+ ## ## 5: # C # # #5:# =5# : + ? # # -## =29H 2@H -# # #* # # #+ 44' (0)") ()&"#("$ /+#""& ? -# # M # + # C S # # #+ # # # -# # # $ # # # + '+3+0@E#-# -# #* #= + ? = !5-: 5#: 5:+ .# -## # #+ .#$## = -# # # + .# = !5:-# # # C + .# # # # ) # $ -# # ## -# # #+ ## #C)#=## D2

PAGE 88

A B D H K 0 20 40 60 80 100 PI at rank 1 Raw Reconstructed Error A B D H K 0 20 40 60 80 100 PV at PF = 1% Raw Reconstructed Error 3+0@+ 5 : 0 B 5 : 0H 5 : -# # -# % = % 5 < !: # & # #+ # @ %= 5B: 5#: 5:5:!,5.:+ ++ # + ? ##=+ ## ###=5# :=5'+3+93+6:+ ? # #=+ !# '+3+0@# #=$##-#$) -## ##+ = #$ 7 ##=# + # # C # # # # (( E #% + .### #5. 8+2:# @+2+ '+3+07# -# "## ##@* = -## -#+ ? # 5: # # # # # # # #+ 5: # -# # + ? "# D8

PAGE 89

A B D H K 0 20 40 60 80 100 PI at rank 1 Raw Reconstructed A B D H K 0 20 40 60 80 100 PV at PF = 1% Raw Reconstructed 3+07+ 5 : 0 B 5 : 0H !5 : -# -#!## !#+ # @ %= 5B: 5#: 5:5:,5.:+ ## -= #= *# ##+ .# -###C #E88-#-##E ##= + 45 #..!% # # # -+ .# % ### #-#((# +? # # # $ # # + ? C # ((%##* # -#81 + &# # $ # # C # #+ D3

PAGE 90

O@2P # 5# # (. #: ## # #+ .# (($ + # C C+ ## #((# 72 = # # # = -# (. ##= + .# ) # # + "#-# # + .#=# -(.## =## #=5!##:+ .#= -### = + .# (. 5: 5: # $"((+<#### ## ((-##+ .# $ ## *=+ .# # = #* + .## # + .#$"(( #= + !#* C #(. # # # # # $ (. ## =###+ '+3+0D# # = 5L G 0115 # $ # 0:: # # # # (( # 5L G0115 # $ # 0::+ ? # # $ ## C *+ # D@

PAGE 91

50 40 30 20 10 0 1 0 5 4 3 2 1 0 1 2 3 4 5 PPPV / PPPV(Hp) as % PD / PD(Hp) as % Grass Concrete '3+0D+ = 5L : B 5L : # # # (. #+ .# # $ # #$+ $ L 1L 1-# # # + .#C 5 1+1@ #:#$+ .# # # # # $ ## $ # $5 1+1@:+ # =# ### # = #$ + ? #+ <=## -###=)#+ .# ## # C *+ .# # # O@2P+ .# # + # # # # ##+ <=# O97P-# # # # # # D7

PAGE 92

##+ ?#-## =# # # ##+ DD

PAGE 93

5 # ## #+ ? # # # = # #$ + # ###* + # # # + # ## 5: ### #E M 5: # # E (* (5((: %-# M 5: # # E (( ## !5 !: -# #-#) C# -# # #-#)-#C+ ? # -# #"# # # #+ .# # # # + # # # # C + # # -# # #* # 5022 ): = C ++ # + .##" # ##= # + # #&(#= @@ )+ .# # (& ( # # D9

PAGE 94

' @+0+ %= # # < )M % # C + C + # # # # &((&+ 5 !0$ /+#"" -&) "!") !&$ /0"+. <"## # -# #C ### #+ .# 8 # # + .# # =+ ? # #$ 8+0+0 # (# # = # %( "+ .# # + 8+0+2 # = # # $+ .# # # + $ # G 50: 5 : $ # G 5 : 5 N : + #$## # 5 : G0 $ + 5 :G 0 5 : 5@+0: '+@+0# -=# # $+ /# # ###= # D6

PAGE 95

+ .# ## # # + # $ $+ # # 5 : G 0 5 : G 0 + .# # # # % # ### + 5 :G ( 5 : 5 : 5@+2: ?### # # + # # "5(: "5 <:+ #@* = -# #=## + '+ @+2 ( # 21 < 21H # # -# # # # # # #+ ? # # # = ++ 5-: 5#: 5: -# #+ .# # # = =5:5,:+ (/R# -## *0 " 5 1+1@:+ <# # ## =###-# 31H#*+ < ##+ <911(E *3+78 $ 1+03 # #M 81 + 91

PAGE 96

0 5 10 15 2 0 0 20 40 60 80 100 RankIdentification Rate Experiment A Experiment B Experiment D Experiment H Experiment K 0 5 10 15 2 0 0 20 40 60 80 100 RankIdentification Rate Experiment A Experiment B Experiment D Experiment H Experiment K 5: 5: 0 5 10 15 2 0 0 20 40 60 80 100 False Alarm RateDetection Rate Experiment A Experiment B Experiment D Experiment H Experiment K 0 5 10 15 2 0 0 20 40 60 80 100 False Alarm RateDetection Rate Experiment A Experiment B Experiment D Experiment H Experiment K 5: 5: '@+2+ @,%= #&'A/. (55:5:: <55:5::-#022)+ # #<#!#&#' ## ##! #+ 90

PAGE 97

5' %)!.(& .!/;!") "+ #(/$!) &"!)( !)$ "!)( /(") /0"+. .# # = # #+ .# -## *O610@P## -# ##-# ####$ #+ ? O@DP+ % # # $ E "= + .# E # + $4 $ +$$% ) + / # -# # C # # # 8 + ?## #(( + # # -# # # % + 5' -#/!") $$) !, $/ 6-7 3+2+2# #(* (+ R + (( # (* ( 5((:" # 0 -# # # G5 !" : #= + .# -# # 21#!* 5!:O8P+ # # # # # # = #-## *#"+ .# #)? ##+ % # ### #* ##*## 92

PAGE 98

#=# #* + '+3+2# -=# #+ ? E#E ##5## -'+3+2: #C )+ 5'' %)!.(& .!/;$ !" %(/ # (( E # $ G # # # # # E G 0 0 + .#E # + ? # # 0 # 2 + .# # # B#OD0P -# ## # + # # $ $ # # -# # # # = ### #+ / ## $ ##((+ '+ @+8# ) C + / # # # C $ -# # ## # #B#+ 5'1 ./!" % .-#"!") # # # + .# # + # ##E#C )+ ## # #Q #"# -##%+ 98

PAGE 99

5: 5: 5: 5: 5: 5: '@+8+ %=/E' <)#5: # # # 5: B-5:#. 5:5:5:.5=( #:+ 93

PAGE 100

2 4 6 8 10 12 14 16 18 2 0 1 500 2 000 2 500 3 000 3 500 4 000 4 500 5 000 State Largest 2nd largest '@+3+ .#B # % !-# % ###%(+ .# '+@+3+ ? # # 5 0 8 09 21: # 56 02: # # # # #*# + .## + # + % # ##$ #$ + 5 :G 5 % : 5 % : 5@+8: # # # @ = &' + '+ @+@ # ( <+ #5:#### ## ++## 5= :5= :5= ,:M 5: # # # -# # -####+ 9@

PAGE 101

0 5 10 15 2 0 0 20 40 60 80 100 Rank Identification Rate Experiment A Experiment B Experiment D Experiment H Experiment K 0 5 10 15 2 0 0 20 40 60 80 100 False Alarm RateVerification Rate Experiment A Experiment B Experiment D Experiment H Experiment K 5: 5: @+@+ # # @ %= &'A/. 5: ( 5: < !21H+ 51 %)!.(& .!/;!") "+ !)$ -+/0(!/ .!") '+ @+7 # # Q # # # # $+ .# C # # -# = + .# # $ -# # = $ + ? # # # O@7 @6P-# #) # # # # + .# # -" #C# -# + # # @+2 # # # # $ E "= ((+ # %+ #'#R !5 !:# = ## ##C + # -# # # 97

PAGE 102

Similarity based on Euclidean Distance S1S2S20 Stance Inference using Population HMM Exemplar Exemplar Exemplar Project Project S1S2S20Linear Discriminant Subspace Emphasizing Inter-subject differences for each stance Probe Silhouettes Probe Stance-Frames Gallery Silhouettes Gallery Stance-Frames Similarity Score Morphological Erosion/Dialation Maximum Similarity Change Morphological Parameter @+7+ .# # # !# /E+ 9D

PAGE 103

# =E + .# # # # #+ 51 )! &(.)!) )!/%&& 6 7 E "= -# # "# $+ + ? # # + % # 5 !: =E # C -C ) E# #)C + ? #!N # # O@P # # !+? )##+ # -+ .# # 5 % :+ # # 0 5 # =: C # # + .# =# # G % 5 :5 : 5@+3: #-#= & G % 5 0 :5 0 : 5@+@: #= G 5 0 :5 0 : 5@+7: 99

PAGE 104

-# # # # + & ## G #)+ = & 5@+D: G # E & # 5 :++ G & G0 2 -# # # # & + <)#-# ## & + .#!5 !: #O@P+ G '" (" 5@+9: -# '" G #)+ = (" G #)+ = '" '" '" & '" '" #* ## # # & + # # E '" 61H = + ? # # # # # + .# $ (" 0 EE -# ) ) = # # + E $ # ") # 0 # 96

PAGE 105

(" + .# # % # 5 :+ (" 5 : G 0 (" 0 (" G 5 0 0 : (" 5 (" : 5 0 0 : 5@+6: 51' ./!"% #)$ /+#"" .!")& '+@+8##I-#J## C + ##-# # + .# # # # + .# # + !# -#-# # #I-#J+ .#& #= # # # $+ .# # # -#+ & 5 :G #)+ = $ $ 5 .) 5 0 : 0 : (" 5 (" : 5 .) 5 0 : 0 : 5@+01: -# .) 5 0 :G # 5 0 : 1 )./ 5 0 : 1 5@+00: .# 7 # ( ) *O6P+ '+ @+D # # # # # '+ @+8 # =E # + ? # ++ 03# # # '+ @+8 # # + !# 61

PAGE 106

5: 5: 5: 5: 5: 5: @+D+ %= # # 5: (# < # ( # 5: B5: # 5: 5:5:.'+@+8+ 60

PAGE 107

=## -##87#D##+ .# -# #### # # # -## # + .# ## #+ 511 9-.) "& !)$ )!/%&& ? # C # # # &( # (& ( -# # 2+2+ ? # # # # + ? E "+ 511 !))0 !)$ &" "& !# ##+ .# (( # # D0 ) # #+ # D0 #$ ) # # # ( + # # # # #= "# #+ /# ##((+= # C#= # #+ # ## "M # ) + = -##&((&#((# + 62

PAGE 108

. # C ) C + # # # C = -# C C + .# # 88 ) -# 07 $ ) # # # + .##@0) / -# -## ( )+ # ) # 9 $ # + ? # # # ) # + .# # ### ## OD@P+ ### )# #$+ !# -'+2+25:+ !-#=$-# # + .#"$022)#88 ) # # 4 # )+ .# $ # # ( )#+ .### # &( (& + .# C )C -#C ##+ 511' !" +!//)0 2/.& + # !( !% '+ @+9 # # ( 5 @: < 5@H:#02= + '+@+6#* # -#### # 68

PAGE 109

1 2 3 4 5 0 20 40 60 80 100 RankIdentification Rate Exp. A Exp. B Exp. C Exp. D Exp. E Exp. F Exp. G Exp. H Exp. I Exp. J Exp. K Exp. L 0 1 2 3 4 5 0 20 40 60 80 100 False Alarm RateVerification Rate Exp. A Exp. B Exp. C Exp. D Exp. E Exp. F Exp. G Exp. H Exp. I Exp. J Exp. K Exp. L 5: 5: @+9+ # /E !# # .%= # # 5-#022 ): 5: (# -*@5: <' !@H+ + -# # # # -# # OD@P&(R((O61P&RN O26P+ ? ##-E# # #= + # -= !5-: 5-N#:+ < # = # # 5= % : # 5= >:+ .# # # = 7 # C 5= :+ 5/ # &( 2 ### @@ ) *+ ? # +: ? = -## #-# # )M ) + ? #$)# / -##( + .#$ D0)#( $#)/ + 63

PAGE 110

0 10 20 30 40 50 60 70 80 90 100 ABCDEFGHIJKL Experiments (Gallery = 122 subjects)Identification rate Baseline UMD UCR New @+6+ # /E 5/-:!#-# <#!# # !# OD@P &(R (( !# O61P &R !# O26P #' 5022):+ 6@

PAGE 111

0 10 20 30 40 50 60 70 80 90 100 A (view)B (shoe)D (surf)H (carry)K (time) Experiments (Gallery = 71 subjects)Identification rate (%) UMD-1 UMD-2 CMU MIT CAS Baseline New @+01+ # %= 5B: 5#. : 5: 5: 5.: # < # 5D0 ) ( :+ .# !# ?# B 5&(0: O07P .? 5&(: O33P # # 5(&:O96P((5(.:O@2P#5!:O63P5&':OD@38P+ .# # /!# & C @0 )#/ + <#= ". 2+8#)# ( $5(:=#/ $-# + # # = # # # O38P -# # # # # OD@P+ '+ @+01 # -##@* = !-# ,#-## #+ ? # E" -# )+ 67

PAGE 112

50 55 60 65 70 75 80 85 TimeNormBaselineUMDMITIdentificaton rate (55 subjects) @+00+ .# # &( 5%= 0 @@ ): # /E !# !# &(R(( !#O3DP(.R!#O@0P+ 5111 !"!2!& .# &( 2 C -# # 5 : C @@ )+ # &( 1 & 2' -# # $ C + # # = # 3// /+/ +.# # # #+ ? # # # + '+ @+00 # # # E # 93H # D0H # # @@H &(R (( #O3DP D1H (.R#O@0P+ .# #E# + 6D

PAGE 113

60 65 70 75 80 85 NewBaselineUMDMITCMUIdentification rate (25 subjects) @+02+ .# # (& ( 5%= 8+0 2@ ): # /E !# !#&( #O61P(.!#O33P(&!#O0@P+ 5114 !"!2!& -$ .# (& ( # -# + = 8+0 (&5 2+3: C + .#E # #* -## ##E + #= -##&(# E#+ !'+@+02# -#-# E##-#-### OD@38P&(O61P(.O33P(&O0@P+ 69

PAGE 114

54 #..!% !)$ )!/%&& # # ##+ .#" # # + # -###+ .# E (* ( 5((: # $ # + # -# # % #+ ? # ## # # + C + .# # # E ((+ # # % -# !##EC# )C# )C + .# # -# I# J I#J # ##+ &* # (( # O@2 61P # (( # E+ $ # (( # # (( $ ((+ = = 5 #&((&(:#E -# C + .# # # #+ E C ) C -# + ? # E+ .## # !# # # ####+ '+@+08# -# 66

PAGE 115

0 10 20 30 40 50 60 70 80 90 100 A(View)B(Shoe)D(Surf)H(Carry)K(Time) Experiments (Gallery = 122 subjects)Identification rate No LDA No Morph Full Algo '@+08+ .#. #@,#%= ##/E!# 5: % # # 5/ !: 5: -# # 5/(#:5:-### 5' !:+ #* = #-#-## + ? ## !###+ % E # # #+ .# # B # O61P O0@P -# # ##+ # #*# # # ) # !#### E+ '+@+03# -## -## # # + # =# + ? #5: 5:* 011

PAGE 116

5:( 5:( '@+03+ .# .( ) % # # # 88 ) # + ) B 07 B B. + 5: # # # + # # # # + = #* O9DP -# # C ##+ 010

PAGE 117

= 8 # @ # #+ # ## -#+ ## C # + O88P # + #OD7P# ###+ < OD2 6@ 01 D9 91 36 87 83P ++ -# O0011710128D7672018P ++ # # C #O0308P8O0002 08P# + 7+0E#* + ## OD2P 0+ .#= -# # O0901P+ 2+ .# # -# # OD2 71 8D 87 0 6@ 012 D9 91 36P+ .# C # #+ .#E # #= + 012

PAGE 118

. 7+0+ (' + ? + ' % # # (&OD2P (&O71P (&O8DP (&O87P (&O83P (&O76P &+O0P !W(&O6@P &/ W&'O01P #O012P (.OD991P &+O36P W O72P W &(O39P &/ O03000208P 8+ .# -## #"* -" #"O7683367203000208P+ .# ) + # ## -## C # # + ? # #* -I#J+ %= # -# #-# + # #O91D9 39P+ # # # ## # -# 4 + ? # #" 018

PAGE 119

# -# # # + = (0)") /0"+.& .## # # + # # # # # # # # + .#### + = !( (0)") /0"+. ? # % # # ( # 5%(: O69P #+ ### ## #! + ? # & # # + .# # # # # $ ###*#* 5*):+ ? ##+ #& -# #D1)+ ?# # ## O9P 5 0 0 / :G # # -. 5 1 5 1 N / :: # # 57+0: -# 0 0 # ) 2 # 0 0 # #) # 1 # # / # # # # #$$#+ < # ( 013

PAGE 120

5: 5: 5: 5: 5: 5: 5: 5#: 7+0+ # ( #+ 5:5: 5:5#:+ # %$+ 7+0+ # # %/ #5%/ #:-# # ##" #+ # '%%. O7DP # %( # # + '+ 7+2 E # 5-#E0211: # = #5: 5: 0 C 5: #0 C+ %(##*" ## 8 = + 21H#=#+ !# # + =' !" (0)") /0"+. ? # # @+2+ R # # # # #$ # # 5 8+0+0:M # #(( 5 3+2+2: $ M # 01@

PAGE 121

0 10 20 30 40 50 60 70 80 90 FcDup. IDup. II EBGM Best of the other algorithms Mean '7+2+. 5<0211:#%( #' !#'%%.2111 O7DP+ .#%= ( #B # 9 %= C-#0 # 99 %= C #0; + # M # #% # #M# # -# # + / # # # # # + # + 7+2 + 7+2+ .# #%= # &' # IJ .+ .# E 022 )+ !0 !2 !8 !3 5%=: 82 88 3@ 06 28 89 5%=,:. 8 0@ 23 8 7 23 017

PAGE 122

=' #&) ( +.& # + ##EE-# #' B. 2112 O77P+ # ) # 5 :+ .# # 5 : 5 : # + .# # E .), 5 :G 5 : 57+2: .# E # # ##+ ## # # #+ .## ### # "+ ? = -# + 0+ # # # (.,$ 5 :G .), 5 :N .), 5 : 57+8: 2+ .# # # # + )# # # + .# # # 5 : # 5 : + .# # # -# # # + '+7+8# -2# # #5:+ 5 :G 0 2 T ) * 57+3: 01D

PAGE 123

4 2 0 2 4 5 0 5 0 50 100 150 Face Gait Histogram '7+8+.#2' /( # + 5 :G 0 2 T ) * 57+@: .#C## # + (.,$ 5 + :G 5 : 5 : 57+7: 8+ .# # # # : # & / + .## ) # + .# # # #"-* 3 %5 :G % 5 :50: % 5 :52: % 5 :52: % 5 :58: 57+D: 019

PAGE 124

5: 5: 5: 5: '7+3+.#' C + .## 5:%=-#(# # 5:%=< # # + .# < -# 5: %= B-5:%=/B-+ -# % 5 :5 :# -## + .# # (.,$ 5 :G 3 5 : 5 :N 3 5 : 5 : 57+9: 3+ # # 4 + # # ""# (.,$ 5 + :G 5 4# 5 :N 4# 5 :: 57+6: !# #### )# -# + # # # # #"+ =1 &#/"& 2)!")& ? # $ # = 016

PAGE 125

0+ ?##N # # S -# -# S 2+ ?# # ##S 8+ # # ++ N NS !# $ $ + # 5 5 $+ : D1 -# = # # + .# 5 5 ,: # -# # # + .# # + '+ 7+3 # = # + .#-# + # # # # # + < 5 5 ,+ 5 ,+ : # # M # 86 # )+ # # 5 5 ,+ 5 ,+ : 8 # M #20 #)+ ###&' + .#5 6 :$D1* ## + .#5 6 ( : # $ # -# = 81 # # -+ # C + .# # # C ++ # # 5 6 '%. 6 '%( : + .#-'+2+2+ # 5 6 6 ( :* 7 # + .#E #### + #= # -. 7+8+ .#"" = -# #=" -# 001

PAGE 126

. 7+8+"#B %= + / %= 5 U: 5U: 5 $+ D1 5 ,+ 86 A< 6 D1 6 '%. 86 N 5 $+ D1 5 ,+ 86 A< 6 D1 6 '%. 86 N 5 $+ D1 5 ,+ 86 A< 5 ,D1 5 ,+ 86 A< N 6 D1 6 '%. 86 6 ( D1 6 '%( 86 5 $+ D1 5 ,+ 20 A< 8( #! 6 D1 6 20 7( #! 5 $+ D1 5 ,+ 20 A< 8( #! 6 D1 6 20 7( #! 5 $+ D1 5 ,+ 20 A< 8( #! 5 ,D1 5 ,+ 20 A< 8( #! N 6 D1 6 20 7( #! 6 ( D1 6 ( 20 7( #! #8 #+ % #" = = -# + # "+ =1 ) "$!/ 2)!") 5%= :5%= :N5%= : -# # # '+ 7+@ -# # #( *@ < @H+ != # 31H 86H 0+ # # # # + D0HD1H @9H"# 79H*+ !'+7+7 # 5%= : 5%= :N5%= :* #+ 000

PAGE 127

1 2 3 4 5 0 20 40 60 80 100 Rank Identification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Face Only Gait Only 0 1 2 3 4 5 0 20 40 60 80 100 False Alarm RateVerification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Face Only Gait Only 5: 5: 7+@+ < 5%= : 5%= : N' 5%= : -# B # 5: "5:B "+ 1 2 3 4 5 0 20 40 60 80 100 Rank Identification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Face Only Gait Only 0 1 2 3 4 5 0 20 40 60 80 100 False Alarm RateVerification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Face Only Gait Only 5: 5: 7+7+ < 5%= : 5%= : N' 5%= :* ( #!-#B # 5: "5:B "+ 002

PAGE 128

=1' ) "!$!/ 2)!") .# #)"# = + #C + .# #)" # =+ ## O, P+ # = # # + # # = 7+8+ .# = # )# #+ % # ## # = ## #+ .# # #7+2+ '+ 7+D # < # @H+ % # # # -# # + ? # # + '+ 7+9 # # ## "@H+? # N # # # N N + .# = # # = # # + 1+D # 1+1@ # + # # # -# O99P+ =4 &(#&&) 7+3# )#C # + # )-#E # # E + # )-# 008

PAGE 129

0 1 2 3 4 5 0 20 40 60 80 100 False Alarm RateVerification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Face Classifier1 Face Classifier2 0 1 2 3 4 5 0 20 40 60 80 100 False Alarm RateVerification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Gait Classifier1 Gait Classifier2 5: 5: 0 1 2 3 4 5 0 20 40 60 80 100 False Alarm RateVerification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Face Classifier1 Face Classifier2 0 1 2 3 4 5 0 20 40 60 80 100 False Alarm RateVerification Rate Rank Sum Weighted Sum Score Sum Bayesian Rule Gait Classifier1 Gait Classifier2 5: 5: '7+D+ &C + .# < # -5:' N' 5:N 5:' N' ( #!5:N( #!+ % ## -# -# # + E # # + ? ## # ) -# # + .# # ) -# # # + < # # # # ) " + .## + # # # # # = @H 003

PAGE 130

0 10 20 30 40 50 60 70 80 90 100 Rank SumWeighted SumScore SumBayesian Rule Face+Gait Face+Face Gait+Gait 0 10 20 30 40 50 60 70 80 90 100 Rank SumWeighted SumScore SumBayesian Rule Face+Gait Face+Face Gait+Gait 5: 5: '7+9+B "' !@H ( 5: 5: ( #+ 7+3+ / ) E E % # ( # # + .# / )86+ # U U < < # < < # 0@ 02 3 @ 0 1 ? # 9 02 1 0 3 1 03 03 8 2 0 1 03 03 8 0 8 1 00@

PAGE 131

2 0 2 4 3 2 1 0 1 2 3 Gait Score(ZNorm)Face Score(ZNorm) Score Sum Gaussian Bayes NonMatch Match 4 2 0 2 3 2 1 0 1 2 3 Gait Score(ZNorm)Face Score(ZNorm) Score Sum Gaussian Bayes NonMatch Match 5: 5: 4 2 0 2 3 2 1 0 1 2 3 Gait Score(ZNorm)Face Score(ZNorm) Score Sum Gaussian Bayes NonMatch Match 4 2 0 2 3 2 1 0 1 2 3 Gait Score(ZNorm)Face Score(ZNorm) Score Sum Gaussian Bayes NonMatch Match 5: 5: 4 2 0 2 3 2 1 0 1 2 3 4 Gait Score(ZNorm)Face Score(ZNorm) Score Sum Gaussian Bayes NonMatch Match 3 2 1 0 1 2 3 2 1 0 1 2 3 Gait Score(ZNorm)Face Score(ZNorm) Score Sum Gaussian Bayes NonMatch Match 5: 5: '7+6+.## @H 5: 5: N C 5: 5: N' C 5:5:NC + 007

PAGE 132

'+ 7+6+ .# = # 5 # + ? #5:# # # -# #=###-#) #+ !5:# # + .# + #+ =5 #..!% !" !)$ !( 2)!") "#$% # # C I#J -# + 5 #:+ ? # # C # + ? # # ++ N # # # ++ N N+ .# # = # I#J C+ 00D

PAGE 133

> # + ? -### # E # # #E # + ## ##&'A/. # &( # (& ( 5: # # # 5: # 5: # ##5:= -#+ ? -# # C # + # # # #+ # # # # # # # C # C # + .# # # + C # # =58+2+3:+ .#C# ## # # + .#C + > :(" ) !" <### #-# #$ + .## # $#7 #C+ ?## 009

PAGE 134

C 5#:5#: # # "60H6@HO68810@P -## #81H -## C # # O@8 07 0@P E + # # -#022 ) -#7 # # 01H #+ < # # # # $ + .# # # # ) # # # # + # $ # # # # + ? # -# # # + .# ## # E # # # #* "01H20H 5022):D1H9@H&(5@@):+ >' :(" # !( ) !" .# # -# + ?# # )*#$#$5 #= -#022):*82H+ # # # # # + ##C##+ #@# ##82H73HE + >1 0.) "!") ) !" (0)") ? # # # # # = # $ + ? 006

PAGE 135

# = # # -# # # =)#+ .# ## # C *+ .# # # O@2P+ .# # + # # # # # # + < = # O97P -# ## # # # # #+ ?# -## = # # # # #+ >4 .)0 (0)") +!%)!.(& #5# #: + ? # # # #E # -# #* # #+ # # # 5: # # # #E 5: # # E # % 5: # # E # 5 !: # + = = 5 # &( (&(: # E -# C + .# # # #+ EC )C -# + ? # 021

PAGE 136

E+ .# # # # # # ####+ % E # # #+ .# # B # O61P O0@P -# # ##+ # #*# # # ) # # # # # E+ ? #'+@+03#5: 5: 5: # # # + >5 !" )$ !( < # $ # "+ ? # $ -# + < # # + ##E # # B 5' B.: 2112 O31P+ E C # #' B.2112" 0H+ &* " E+ # # *+ B.2112 # + # # + # # # # + B C @3H @H -# 83H+ '+ @+9 5.# #: 69H8@H#= -## 020

PAGE 137

$ # + .# D1H+ ## C ##M#C @ #+ B 3DH1HC -#22H+ %= # -# ## 7 ##+ .# "# = 20H+ #+ .# B.2112 $+ .# #+ &# # ##0:# -#2: -#+ ? # C # + 5 #:+ ? # # C## 5'+ 7+6:+ ? # # ++ N # # # ++ NN+ .# # = # I#J C+ >= #"# &!( + (")& ?## # ## #+ # # # + = # * O9DP -# # C # #+ !# 8### #$ 022

PAGE 138

# B. 2112O31P+ -###+ # # # S .# ##+ = # C * # # # ) # + # # ## #+ < # ##*##*?O66P-# # + .#C # # -## #-# + = ##*S # # # # S .#E #-#) #++ + ## $ )R # # )R S 028

PAGE 139

O0P +! #+* + "+ 9 0667+ O2P >+ !F+ + -+ 9 ; D858:329K331 ( #0666+ O8P +!* + #=#= #+ 9 9 27DK290 06D8+ O3P ++ >+%+ +.+,E -*+ # Q+ 2803@K0@2 06D9+ O@P + /+ # >+ + # + >+ ,+ % + "# ")+ 9777 $ 9 06D00KD21 >066D+ O7P + !* + + + ( + 9 27DK2D2 2112+ ODP !+ *!+>#+ "+ 328K381 2110+ O9P + + + % ## #+ & 2118+ O6P B++B++ + 9 3 $ 9 @3 2@3K2@9 ( 0662+ O01P ,+#,+?+ +* +B+ + 9777 $ 9 0071K007@ 2118+ O00P ,++#,+?+ +>+'+ 28 + :4 $ ; 2@K82 2118+ 023

PAGE 140

O02P ,+ + # ,+ ?+ + >+ '+ ( 2 8 + 9777 9 :4 $ /' <+2118+ O08P ,+ + # ,+ ?+ + >+ X+ #+ ( # + 1 9777 9 ,( 0D062113+ O03P X+# +>+',+?+ + B# + :4 $ ; 39K@@ !&!+2118+ O0@P + + >+ #+ # # #+ 9 2 877K8D0 2112+ O07P /+ !+,+#+ + 97779 2118+ O0DP >+%+ +.+,E -*+ #*+ # 68@8K8@7 06DD+ O09P &+ +* .+? + + 92DK988 066D+ O06P ?+ + #+ # + 2 220@K220668+ O21P + ++*+ + ? 2110+ O20P + + %++ + + 03D@K0392 2110+ O22P +' # #+#* + !# + 0D89DK31D 0690+ O28P >+ + /+ (+ + ++ !+ + 232396K236D 2118+O23P + + .# # + 9 ; D850:92K69 > 0666+ O2@P +>+#+ .# 5:+ #+(& .1009 (& >2110+ O27P + >+ #+ .# 5: + # (& 2110+ 02@

PAGE 141

O2DP %+ (+ + + B+ ( + /# + + + < D00KD21 2111+ O29P >+E +' + #+(+#>+ +"S 9 08DD@KD9@ 2110+ O26P >++# + # + 932K93D >2113+ O81P >+ !$# (+ /= >+ + + 9 2 2# # 2D2K2DD 2110+ O80P !+ +' + ( R# + 9; 9058:22DK281 ( #2110+ O82P + + C + %+ # #+ .# + # 06@K213 0692+ O88P +!+>+ + S 9777 :4 9 @6K73 0666+ O83P +!+,+>+ "+ 9777 $ 9 21502:026@K081D 0669+ O8@P >+ > + ,+ ? + !97! 9 78 $ + 0667+ O87P !+ ,+> + ;+ ,* + #+ 9= 2 2 # 092K09D ( #0666+ O8DP !+ ,+ > + #* + #+ # ## "+ 2150008:08D0K 08D6 0666+ O89P +>#+ B+ 282D@K99 >06D7+ O86P !+ ># !+ *+ -# + 9 2 2# # 810K800 2110+ O31P + ># # + (+ + # + ( # %+. + + 3// / 2112+ O30P + ># # + ( + E + + .# '%%. # #+ 9777 $ 9 22501:0161K0013 2111+ 027

PAGE 142

O32P + ># # + + + # ,+ + # # + 9 08DK032 2112+ O38P + ># # + + + # ,+ + .# # #+ 9 89@K899 2112+ O33P !+ + + + ; !+ /+ ) + #+ # "+ 9 2 2 # # 2118+ O3@P !+ /+ + #+ # "+ 9 2 2112+ O37P !+ !+ ) /+ B+ ,+ # ((+ 9 887K830 2112+O3DP !+ !+ /+ ) !+ /+ !+ + # -# + "# + 9777 9 + O39P !+ !+ # -# + #+ # + 9 2113+ O36P >+ (+ + + >+ (+ < "+ 9777 $9 2158: 0669+ O@1P ++ !+ %+ + 9 03DK0@2 ( 2113+ O@0P + + + # # ( # # >2118+ O@2P + + ,+.+ #" + 9 2118+ O@8P + ?+ + "+ 92 0@@K072 2112+ O@3P >+ >+ +E # .##+ 052:0K88 0669+ O@@P + +( !+< +#* +* +. # + 97 2 # + 9 06@K21@ 2113+ 02D

PAGE 143

O@7P + +( +* + #$ + D13KD00 >2113+ O@DP + + + < -# + # ? #++ 2113+ O@9P + + + #+ 9 3 200K203 2113+ O@6P + +* +%C#$ #+ 9777 $ #% + O71P X+ ;+ ? !+ ,+ >+ + 9777 9 $ 71 808K07 >2118+ O70P ++( + # #+ ( R#& #'2118+ O72P <+(*;+B +#+(=* " + 2118+ O78P .+ ( %+ + # + 9 ; 9058:280K279 ( #2110+ O73P (+ ( !+# +, + ? + < # < 37!88@K871 0673+ O7@P +/ %+!+ !E-#+ 0663+ O77P +>+# +#+>+( #+(+ *%+. (++ 2112+ 3// ( #2112+ O7DP +>+#+(+!+E +>++.#'%%. # #+ 9777 $ 9 22501: <+2111+ O79P '+ %+ + (+ !+ !+ >+ + C + @0K70 2110+ O76P +#* !+,+>+ "+ 8@53:970K9D3 2112+ OD1P + + + >+ # + 9777 $5 3K070697+ OD0P +++>+ + 0668+ OD2P !+ !+ ,+ >+ + 23200@K202@ 2118+ 029

PAGE 144

OD8P + + *#+ B # + < 71 3 + DD88K D310690+ OD3P + + *#+ %= + < 71 3 002 @9@K70@ 0698+ OD@P +* +>+#+ + +#,+ + .# # + 9777 $9 + OD7P + #+ "S 9 0D7K0D6 2112+ ODDP (+##+>5%+:+ $2# + #066D+ OD9P + #*# # .+ + < "+ 9777 9 2112+ OD6P +#*# # + .++ -+ 386K3372110+ O91P +#*# # + .++ -# -+ 2110+ O90P >+ # (+ /= + + + ) 9777 9 9 260K26@ 2111+ O92P +B+ (++/=B+,+ B + 0857:@08K@27 + 0666+ O98P + (+ +/+(*+ )+ # @77K@D8 2118+ O93P !+ !+ ,+ # -# + #+ #* # $+ 9777 92 9 2118+ O9@P +. !+ *+E) )#*+ D27KD80 2110+ O97P +. !+ *+ C + 9 2 2# # 2118+ 026

PAGE 145

O9DP + !+ *+ ( # C + D98KD61 >2113+ O99P + (+ = (+ + + >+ ,+ S 8856:03D@K039@ 2111+ O96P +. ++ # "+ 9 2 2# # 2118+ O61P !+ B # !+ +# -# + #+ # # + ? #++&!>2113+ O60P ++B + $ 3 + ##& #'2112+ O62P + + B + + # # # + 9777 $ 9 2@0828K0829 2118+ O68P +? ?+.+. + !-# "+ 9 000@K009 2112+ O63P +? .+. +/?++ # # "+ 9777 $ 9 2@0@1@K0@09 +2118+ O6@P ;+ ? .+ !+ ,+ >+ "+ 9= 2 2# # 91@K908 >2118+ O67P >+!+? >+,+!+ ) )+ 9 0601DK081 0692+ O6DP '+? =+ *#+ # 091K98063@+ O69P + ?* >+(+ /+ + (+ # # #+ 9777 $ 9 065D:DD@KDD6 >066D+ O66P (+ ?+ B # C + # ? #++ 2113+ O011P ,+? ?+#, >+,+ + + 9777 $ 9 0653: 066D+ 081

PAGE 146

O010P + ;+ ; (+ + /= >+ /+ + < # # # + 9 29DK261 2118+ O012P >+#+ #+ #+ 92 36DK@11 2112+ O018P ;++ +.# # + 2 9/* 7 217K201 0667+ 080

PAGE 147

# #E#& 066D # ( & % ##2111+ #& # '+ # +


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001498175
003 fts
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 041209s2004 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000529
035
(OCoLC)57716599
9
AJU6778
b SE
040
FHM
c FHM
1 100
Liu, Zongyi,
d 1975-
0 245
Gait-based human recognition at a distance
h [electronic resource] :
performance, covariate impact and solutions /
by Zongyi Liu.
260
[Tampa, Fla.] :
University of South Florida,
2004.
502
Thesis (Ph.D.)--University of South Florida, 2004.
504
Includes bibliographical references.
500
Includes vita.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 147 pages.
520
ABSTRACT: It has been noticed for a long time that humans can identify others based on their biological movement from a distance. However, it is only recently that computer vision based gait biometrics has received much attention. In this dissertation, we perform a thorough study of gait recognition from a computer vision perspective. We first present a parameterless baseline recognition algorithm, which bases similarity on spatio-temporal correlation that emphasizes gait dynamics as well as gait shapes. Our experiments are performed with three popular gait databases: the USF/NIST HumanID Gait Challenge outdoor database with 122 subjects, the UMD outdoor database with 55 subjects, and the CMU Mobo indoor database with 25 subjects. Despite its simplicity, the baseline algorithm shows strong recognition power. On the other hand, the outcome suggests that changes in surface and time have strong impact on recognition with significant drop in performance.To gain insight into the effects of image segmentation on recognition -- a possible cause for performance degradation, we propose a silhouette reconstruction method based on a Population Hidden Markov Model (pHMM), which models gait over one cycle, coupled with an Eigen-stance model utilizing the Principle Component Analysis (PCA) of the silhouette shapes. Both models are built from a set of manually created silhouettes of 71 subjects. Given a sequence of machine segmented silhouettes, each frame is matched into a stance by pHMM using the Viterbi algorithm, and then is projected into and reconstructed by the Eigen-stance model. We demonstrate that the system dramatically improves the silhouette quality. Nonetheless, it does little help for recognition, indicating that segmentation is not the key factor of the covariate impacts. To improve performance, we look into other aspects.Toward this end, we propose three recognition algorithms: (i) an averaged silhouette based algorithm that deemphasizes gait dynamics, which substantially reduces computation time but achieves similar recognition power with the baseline algorithm; (ii) an algorithm that normalizes gait dynamics using pHMM and then uses Euclidean distance between corresponding selected stances -- this improves recognition over surface and time; and (iii) an algorithm that also performs gait dynamics normalization using pHMM, but instead of Euclidean distances, we consider distances in shape space based on the Linear Discriminant Analysis (LDA) and consider measures that are invariant to morphological deformation of silhouettes. This algorithm statistically improves the recognition over all covariates.Compared with the best reported algorithm to date, it improves the top-rank identification rate (gallery size: 122 subjects) for comparison across hard covariates: briefcase, surface type and time, by 22%, 14%, and 12% respectively. In addition to better gait algorithms, we also study multi-biometrics combination to improve outdoor biometric performance, specifically, fusing with face data. We choose outdoor face recognition, a "known" hard problem in face biometrics, and test four combination schemes: score sum, Bayesian rule, confidence score sum, and rank sum. We find that the recognition power after combination is significantly stronger although individual biometrics are weak, suggesting another effective approach to improve biometric recognition.The fundamental contributions of this work include (i) establishing the "hard" problems for gait recognition involving comparison across time, surface, and briefcase carrying conditions, (ii) revealing that their impacts cannot be explained by silhouette segmentation, (iii) demonstrating that gait shape is more important than gait dynamics in recognition, and (iv) proposing a novel gait algorithm that outperforms other gait algorithms to date.
590
Adviser: Sarkar, Sudeep.
653
LDA.
Eigen-Stance.
Population HMM.
baseline.
gait biometrics.
690
Dissertations, Academic
z USF
x Computer Science and Engineering
Doctoral.
090
TK7885 (ONLINE)
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.529