Face Recognition Using Embedded Hidden Markov Model.

Slides:



Advertisements
Similar presentations
Lecture 16 Hidden Markov Models. HMM Until now we only considered IID data. Some data are of sequential nature, i.e. have correlations have time. Example:
Advertisements

Angelo Dalli Department of Intelligent Computing Systems
HMM II: Parameter Estimation. Reminder: Hidden Markov Model Markov Chain transition probabilities: p(S i+1 = t|S i = s) = a st Emission probabilities:
Hidden Markov Model 主講人:虞台文 大同大學資工所 智慧型多媒體研究室. Contents Introduction – Markov Chain – Hidden Markov Model (HMM) Formal Definition of HMM & Problems Estimate.
Hidden Markov Models. A Hidden Markov Model consists of 1.A sequence of states {X t |t  T } = {X 1, X 2,..., X T }, and 2.A sequence of observations.
Lecture 8: Hidden Markov Models (HMMs) Michael Gutkin Shlomi Haba Prepared by Originally presented at Yaakov Stein’s DSPCSP Seminar, spring 2002 Modified.
Introduction to Hidden Markov Models
Tutorial on Hidden Markov Models.
Hidden Markov Models: Applications in Bioinformatics Gleb Haynatzki, Ph.D. Creighton University March 31, 2003.
Statistical NLP: Lecture 11
Hidden Markov Models Theory By Johan Walters (SR 2003)
Statistical NLP: Hidden Markov Models Updated 8/12/2005.
1 Hidden Markov Models (HMMs) Probabilistic Automata Ubiquitous in Speech/Speaker Recognition/Verification Suitable for modelling phenomena which are dynamic.
Hidden Markov Models Fundamentals and applications to bioinformatics.
Hidden Markov Models in NLP
Hidden Markov Model based 2D Shape Classification Ninad Thakoor 1 and Jean Gao 2 1 Electrical Engineering, University of Texas at Arlington, TX-76013,
Lecture 15 Hidden Markov Models Dr. Jianjun Hu mleg.cse.sc.edu/edu/csce833 CSCE833 Machine Learning University of South Carolina Department of Computer.
Hidden Markov Models 1 2 K … 1 2 K … 1 2 K … … … … 1 2 K … x1x1 x2x2 x3x3 xKxK 2 1 K 2.
Apaydin slides with a several modifications and additions by Christoph Eick.
Albert Gatt Corpora and Statistical Methods Lecture 8.
What is the temporal feature in video sequences?
HMM-BASED PATTERN DETECTION. Outline  Markov Process  Hidden Markov Models Elements Basic Problems Evaluation Optimization Training Implementation 2-D.
Hidden Markov Models Pairwise Alignments. Hidden Markov Models Finite state automata with multiple states as a convenient description of complex dynamic.
Timothy and RahulE6886 Project1 Statistically Recognize Faces Based on Hidden Markov Models Presented by Timothy Hsiao-Yi Chin Rahul Mody.
Hidden Markov Models K 1 … 2. Outline Hidden Markov Models – Formalism The Three Basic Problems of HMMs Solutions Applications of HMMs for Automatic Speech.
Forward-backward algorithm LING 572 Fei Xia 02/23/06.
1 Hidden Markov Model Instructor : Saeed Shiry  CHAPTER 13 ETHEM ALPAYDIN © The MIT Press, 2004.
Doug Downey, adapted from Bryan Pardo,Northwestern University
Scenario Generation for the Asset Allocation Problem Diana Roman Gautam Mitra EURO XXII Prague July 9, 2007.
Hidden Markov Models 戴玉書
Fall 2001 EE669: Natural Language Processing 1 Lecture 9: Hidden Markov Models (HMMs) (Chapter 9 of Manning and Schutze) Dr. Mary P. Harper ECE, Purdue.
Visual Recognition Tutorial1 Markov models Hidden Markov models Forward/Backward algorithm Viterbi algorithm Baum-Welch estimation algorithm Hidden.
. Class 5: Hidden Markov Models. Sequence Models u So far we examined several probabilistic model sequence models u These model, however, assumed that.
ETHEM ALPAYDIN © The MIT Press, Lecture Slides for.
Ch10 HMM Model 10.1 Discrete-Time Markov Process 10.2 Hidden Markov Models 10.3 The three Basic Problems for HMMS and the solutions 10.4 Types of HMMS.
Isolated-Word Speech Recognition Using Hidden Markov Models
THE HIDDEN MARKOV MODEL (HMM)
Fundamentals of Hidden Markov Model Mehmet Yunus Dönmez.
1 HMM - Part 2 Review of the last lecture The EM algorithm Continuous density HMM.
International Conference on Intelligent and Advanced Systems 2007 Chee-Ming Ting Sh-Hussain Salleh Tian-Swee Tan A. K. Ariff. Jain-De,Lee.
Clustering Spatial Data Using Random Walk David Harel and Yehuda Koren KDD 2001.
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: ML and Simple Regression Bias of the ML Estimate Variance of the ML Estimate.
ECE 8443 – Pattern Recognition ECE 8527 – Introduction to Machine Learning and Pattern Recognition Objectives: Reestimation Equations Continuous Distributions.
HMM - Part 2 The EM algorithm Continuous density HMM.
CS Statistical Machine learning Lecture 24
1 CONTEXT DEPENDENT CLASSIFICATION  Remember: Bayes rule  Here: The class to which a feature vector belongs depends on:  Its own value  The values.
1 CS 552/652 Speech Recognition with Hidden Markov Models Winter 2011 Oregon Health & Science University Center for Spoken Language Understanding John-Paul.
1 CSE 552/652 Hidden Markov Models for Speech Recognition Spring, 2006 Oregon Health & Science University OGI School of Science & Engineering John-Paul.
Pattern Recognition and Machine Learning-Chapter 13: Sequential Data
 Present by 陳群元.  Introduction  Previous work  Predicting motion patterns  Spatio-temporal transition distribution  Discerning pedestrians  Experimental.
ECE 8443 – Pattern Recognition ECE 8527 – Introduction to Machine Learning and Pattern Recognition Objectives: Elements of a Discrete Model Evaluation.
Hidden Markov Models (HMMs) –probabilistic models for learning patterns in sequences (e.g. DNA, speech, weather, cards...) (2 nd order model)
1 Hidden Markov Models Hsin-min Wang References: 1.L. R. Rabiner and B. H. Juang, (1993) Fundamentals of Speech Recognition, Chapter.
ECE 8443 – Pattern Recognition ECE 8527 – Introduction to Machine Learning and Pattern Recognition Objectives: Reestimation Equations Continuous Distributions.
Hidden Markov Model Parameter Estimation BMI/CS 576 Colin Dewey Fall 2015.
Hidden Markov Models. A Hidden Markov Model consists of 1.A sequence of states {X t |t  T } = {X 1, X 2,..., X T }, and 2.A sequence of observations.
Definition of the Hidden Markov Model A Seminar Speech Recognition presentation A Seminar Speech Recognition presentation October 24 th 2002 Pieter Bas.
Other Models for Time Series. The Hidden Markov Model (HMM)
Visual Recognition Tutorial1 Markov models Hidden Markov models Forward/Backward algorithm Viterbi algorithm Baum-Welch estimation algorithm Hidden.
From DeGroot & Schervish. Example Occupied Telephone Lines Suppose that a certain business office has five telephone lines and that any number of these.
Hidden Markov Models HMM Hassanin M. Al-Barhamtoshy
Hidden Markov Models BMI/CS 576
Chapter 3: Maximum-Likelihood Parameter Estimation
Hidden Markov Models.
Hidden Markov Models Part 2: Algorithms
Hidden Markov Autoregressive Models
CONTEXT DEPENDENT CLASSIFICATION
Handwritten Characters Recognition Based on an HMM Model
LECTURE 15: REESTIMATION, EM AND MIXTURES
Visual Recognition of American Sign Language Using Hidden Markov Models 문현구 문현구.
Presentation transcript:

Face Recognition Using Embedded Hidden Markov Model

Overview Introduction. Motivation behind this research project. Markov Chains -- how to estimate probabilities. What is Hidden Markov Model (HMM). Embedded HMM. Observation Vectors. Training of face models. Face Recognition. Conclusion.

Introduction We implemented a real-time face recognition scheme “Embedded HMM” and compared it with the Human Visual System. Embedded HMM approach uses an efficient set of observation vectors and states in the Markov chain.

Motivation None of the researchers have compared how an objective measurement algorithm results (face recognition algorithm) perform against a subjective measurement (Human visual system). Learning the inner details of face recognition algorithm, how it works and what it actually does to compare two faces, what are the complexities involved and what are the improvements possible in future.

Markov Chains -- How to estimate probabilities S is a set of states. Random process {X t |t=1,2….} is a Markov Chain if  t, the random variable X t satisfies the Markov property. p ij =P{X t+1 =j | X t =i, X t-1 =i t-1, X t-2 =i t-2, …….X 1 =i 1 } = P{X t+1 =j | X t =i } So, we have : P(X 1 X 2 …….X k-1 X k )=P(X k | X 1 X 2 ……. X k-1 ).P(X 1 X 2 …….X k-1 ) =P(X k | X k-1 ).P(X 1 X 2 …….X k-1 ) =P(X 1 ).  P(X i+1 | X i ) for i=1,k-1 p ij is called the transition matrix. Markov Model=Markov Chain + Transition matrix.

Hidden Markov Model HMM is a Markov chain with finite number of unobservable states. These states has a probability distribution associated with the set of observation vectors. Things necessary to characterize HMM are: -State transition probability matrix. -Initial state probability distribution. -Probability density function associated with observations for each of state.

Example Illustration N=number of states in model M=number of distinct observation symbols. T=length of a observation sequence (no of symbols) Observation symbols=(v1,v2,...vM).  i =probability of being in state i at time 1(start). A={a ij }=probability of state j at time t+1, given state i at time t. B={b j k }=prob of observing symbol vk in state j. O t =observation symbol at time t. =(A,B,  ) denotes the HMM model.

Cont… O=O 1 …..O T is called the observation seq. How did we get this ??? How do we get the probability of occurrence of this sequence in the state model?? P(O| ) Hint: sum ( P(O| I, ).P(I| )), where I=state sequence. Find I such that we get Max P(O,I| ). (viterbi algorithm). How do we find this I ??? Hint: Converts into minimization of path weight problem in graph.

HMM Model Face recognition using HMM

Cont… How do we train this HMM?? Hint: Encode the given observation sequences in such a way that if a observation sequence having many characteristic similar to the given one is encountered later, it should identify it. (use k-means clustering algorithm)

K-means clustering explained… Form N clusters initially. Calculate initial probabilities and transition probabilities. ( ,A) Find mean and covariance matrix for each state. symbol probability distribution for each training vector in each state. (Gaussian Mixer). (B) So =(A,B,  ) as calculated above. Find optimal I for each training sequence using this. Re-assign symbols to different clusters if needed. Re-calculate (HMM). Repeat until no re-assignments are possible.

Embedded HMM Making each state in a 1-D HMM an HMM, makes an embedded HMM model with super states along with embedded states. Super states model the data in one direction(top-bottom). Embedded states model the data in another direction(left-right). Transition from one super state to another is not possible. Hence named “Embedded HMM”.

Embedded HMM Model

Elements of Embedded HMM: Number of super state: N 0, set of super states: S 0 ={S 0,i } Initial super state distribution,  0 =(  0,i ), where  0,i are the probabilities of being in super state i at time 0. super state transition probability matrix: A 0 ={a 0,ij }, where a 0,ij is the probability of transitioning from super state j to super state i. Parameters of embedded HMM: -Number of embedded states in the kth super state, N 1 (k), and the set of embedded states S 1 (k) ={S 1,i (k) }. -  1 (k) =(  1,I (k) ), where  1,I (k) are probabilities of being in state I of super state k at time 0. -State transition probability matrix A 1 (k) ={a 1,ij (k) }

Cont.. The state probability matrix B (k) ={b i (k) (O t 0,t 1 )}, for set of observation vector at row t 0, column t 1. Let  (k) ={  1 (k), A 1 (k), B (k) } is the set of parameter for k th super state. So, Embedded HMM could be defined as: ={  0, A 0,  }, where  ={  (1),  (2),…  (N 0 ) }

Observation Vectors P x L window scans the image from left-right and top-bottom, with overlap between adj windows is M lines vertically and Q columns horizontally. Size of observation vectors = P x L. Pixel value don’t represent robust features due to noise and changes in illuminations. 2D-DCT coefficients in each image block.(low freq components, often only 6 coefficients). This helps reduce size of obs_vector drastically.

Training of Images

Face recognition Get the observation sequence of test image. (obs_test) Given ( 1,…… 40) Find likely hood of obs_test with each i. The best likely hood identifies the person. Likely Hood = P(obs_test| i) Hint:use viterbi algorithm again to get the sequence state for this obs_test sequence.

Conclusion Small observation vector set. Reduced number of transitions among states. Lesser computing. 98% real-time recognition rate. Little overhead of complexity of algorithm.