Methods of Pattern Recognition with Application to Voice

(3-0-0-3)

CMPE Degree: This course is Not Applicable for the CMPE degree.

EE Degree: This course is Not Applicable for the EE degree.

Lab Hours: 0 supervised lab hours and 0 unsupervised lab hours.

Technical Interest Group(s) / Course Type(s): Digital Signal Processing

Course Coordinator:

Prerequisites: ECE 4270

Catalog Description

Theory and application of pattern recognition with a special
application section for automatic speech recognition and related
signal processing.

Textbook(s)

Pattern Recognition

Course Outcomes

Not Applicable

Strategic Performance Indicators (SPIs)

Outcome 1 (Students will demonstrate expertise in a subfield of study chosen from the fields of electrical engineering or computer engineering):
1. Using theory learned in class the students should be able to display speech waveforms and images for visually inspecting key information and cues needed for pattern recognition.

Outcome 2 (Students will demonstrate the ability to identify and formulate advanced problems and apply knowledge of mathematics and science to solve those problems):
1. Using tools learned in class the students should be able to design feature extraction and pattern classification modules needed for putting together a pattern recognition system.

Outcome 3 (Students will demonstrate the ability to utilize current knowledge, technology, or techniques within their chosen subfield):
1. Given a project at work, the students should be able to carry out literature survey, problem formulation, and experimental design by relating real-world pattern recognition problems to what they learn in class.

Topical Outline

Review of probablilty with an emphasis on random vectors
Linear transformations, diagonalizations, rotations, projections
Distance measures
Clustering (unsupervised pattern recognition).
Centroids
Interset distances
Sums of distances
Intraset distances (distortion measures)
Algorithms
Performance measures
VQ
Hierarchical clustering
Parametric Modeling
MAP classification, Bayesian analysis
Minimum risk criteria, Neyman-Pearson criteria
Gaussian assumptions
Gaussian mixture densities, EM algorithm
Non-Gaussian: training of densities using basis functions
Linear discriminant functions
Single layer perceptron
Gradient descent algorithms
Widrow-Hoff algorithm
Nonlinear transformations prior to LDFs (potential functions).
Neural Networks
Feedforward (MLPs)
Back Propagation.
Radial Basis function NNs (RBFs)
Self-organizing feature maps
Data (Dimensionality) Reduction
Intro to sequence comparisons: time warps and stochastic grammars.
Intro to acoustic phonetics
Front ends (feature acquisition) for speech
Filter banks and LPC
Auditory models
Development of Mel-Cepstra from both a PR and
DSP point of view (Karhunen-Loeve transformation)
Dynamic Time Warping (Deterministic and Probabilistic)
Clustering for VQ-DTW, Training, Template Adaptation
Discriminative Methods
Robust methods
Markov Processes, hidden and observed
Discrete HMM's, Recognition and Training
Continuous Observation HMM's
Semi-Markov Models
Model Adaptation
Connected Words: Level Building
Large Vocabulary Systems
Word Spotting
Speaker ID