AME Patterns library alpha

Stjepan Rajko

Arts, Media and Engineering Program

Arizona State University

AMELiA is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

Table of Contents

Preface

Downloading and using the library

Examples

Partial List
Offline Gesture Recognition from Video
Real-time mouse gesture recognition

Tasks

Introduction
Training
Classification
Synthesis

Experiments

Datasets

Introduction
List of Datasets
Dataset Concept
ClassificationDataset Concept

Observations, Distributions, Trainers and Generators

Introduction
normal

Concepts

Introduction
Observation
ObservationDistribution
ObservationDistributionTrainer

Reference

Header </Development/amelibraries/trunk/include/ame/observations/distribution/normal.hpp>
Header </Development/amelibraries/trunk/include/ame/observations/training/normal.hpp>
Header </Development/amelibraries/trunk/include/ame/observations/generation/normal.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/classification.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/classification_and_synthesis.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/filtered_classification.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/get_distance.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/mean_number_of_states.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/min_number_of_states.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/recognition.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/semantic_recognition.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/synthesis.hpp>
Header </Development/amelibraries/trunk/include/ame/patterns/task/training.hpp>

License

Preface

“Library documentation often begins with a quote.”
--The AME Patterns library

Description

The AME Patterns library is a library for recognition and generation of patterns, specifically patterns of sequential data such as time series data. Examples of such patterns are:

a head-nod gesture
the word "Hello"
a stock market crash

Some of the tasks you can accomplish using the library are:

classification of patterns (e.g., gesture classification - given a recording of a gesture, decide whether the gesture was a head nod or a head shake)
on-line detection of patterns in a continuous stream of observations (e.g., on-line word recognition - continously listen to speech and detect all utterances of the word "Hello").
generation of patterns (e.g., speech synthesis - generate an audio sample of the word "Hello")

The library is generic in the sense that it can be applied to patterns in different domains / modalities. To use the library in a particular domain, the library user will only need to provide some basic domain-specific building blocks. For example, the library does not come with any support for the domain specifics necessary to accomplish the example tasks above related to head gestures and speech.

This library is under active development. More details can be seen in the development plan.

Demo

To see the library in action, see the YouTube video showing recognition of mouse gestures. The application shown is the mouse gesture example.

Acknowledgements

This material is based upon work supported by the National Science Foundation CISE Infrastructure grant under Grant No.0403428 and National Science Foundation IGERT grant under Grant No. 0504647