The Use of PRAGMA on Distributed Virtual Instrumentation for Signal Analysis (DiVISA)

Domingo RodriguezWilson Rivera

ECE Department

University of Puerto Rico at MayaguezSeptember 24, 2007

Our Vision

Developing the concept of  distributed virtual instrumentation for signal analysis (DiVISA) as a means of fostering interdisciplinary collaboration in signal-based  information processing (SbIP) through the PRAGMA grid service community resource framework (GSCRF).

An Infrastructure for Human Collaboration

Applications Layer

Network Layer

Physical Layer

Physical WorldDistributed Sensor NetworksDSN

Medium Access ControlMAC

Service Oriented ArchitectureSOA

EnvironmentalObservatory

User’s TargetApplication

Observables

Signals

Data

Sensors Effectors

SignalProcessing

Information Processing

InformationKnowledgeProcessing

Knowledge

Decision System

Intelligence

PRAGMA: A Grid Service Community Resource Framework (GSCRF) for

Information Flow

“PRAGMA”

It deals with the gathering and processing of appropriate environmental information to aid in the process of effective decision making!

http://www.walsaip.uprm.edu

Environmental Surveillance* Monitoring (ESM)

*From French: sur- 'over' + veiller- 'watch'

ESM: WALSAIP’s Main Research Objective

WALSAIP: Wide Area Large Scale Automated Information Processing

Photo: Gail S Ross

Searching for the endangered Bufo

[Peltophryne] lemur through environmental surveillance monitoring

Aroma’s basin

Atolladora’s basin

Tamarindo’s basin

Picture: DRNA

Tamarindo’s basin

Master Sensor: gumstix embedded PC, Power supply for basic sensors, and remote internet access.

Basic sensors: gumstix embedded PC based acoustic recorders (frogloggers)

Ethernet and power cables

Environmental Surveillance Monitoring Region

WALSAIP Sensor Grid (WSG)

NS0

NS1

NS2

NSN-1

Basic Interface

Module (BIM)

Linear Sensor Array(LSA)

NSk: kth Sensor Node

EmbeddedComputer

System (ECSa)

Storage Device ~2TB

USA

China

Global users

ECS-G interface

Grid-S interface

Memory~2GB Grid Environment

Japan

Others

rth Master Sensor Node (MSN)

JBNERR, PR

The Concept of the Acoustical Map (A-MAP) Type I

Microphone Array

x

y

A-MAP processor

A-MAP Output Type I: Direction of Arrival (DoA)

x

y seagull_01

coqui_01

The Concept of the Acoustical Map (A-MAP) Type II

Sensor Array

x

y

A-MAP Processor

A-MAP Output Type II: Time-Frequency Distribution (TFD)

0.1 0.2 0.3 0.4 0.5 0.6 0.70

1000

2000

3000

4000

5000

6000

7000

Time

Frequency

0.05 0.1 0.15 0.2 0.25 0.3 0.350

1000

2000

3000

4000

5000

6000

7000Coqui Seagull

Analyzed sound

Full length sound

Analyzed sound

Full length sound

Signal Analysis Tools for Information Flow

Cohen-Class Type Time-frequency Distribution (TFD), C (t,f )

An example of distance measure between C1(t,f)=p1 and C2(t,f)=p2

Another example of distance measure: Kullback-Leibler Divergence

Rényi Divergence: Generalized Formulation of Kullback-Leibler Divergence

System Information Flow Characterization

Shannon entropy when applied to TFDs

The αth order Rényi entropy

Energy Flow Characterization: Power–Estimation in “energy change/unit time”

Information Flow Characterization: –Estimation in “entropy change/unit scale”

Raw Data Generation Requirements

Analyzing acoustic data to extract relevant information from a single site sensor array (M nodes) may be a “24/7/365” activity. At a 48K samples/sec rate, 16 bits A/D, single node raw data acquisition may generate about 5 Terabytes of data yearly. If a “single laptop” approach is taken for single node data analysis using existing software packages, it would take about four (4) person-year for a one (1) year raw data.

Advanced Computational Requirements

Large scale signal analysis techniques such as multivariate analysis and multispectral analysis of time-frequency distributions (TFD) bring orders of magnitude to initial raw data.

This work seeks to introduce automation techniques to large scale signal analysis by efficiently using distributed computing resources and

data on a grid infrastructure!

On Going Works

Developing a framework for automating large scale signal analysisIntegrating large scale signal analysis tools with a graphical user interface.Formulating a real time signal analysis framework for connecting to WSG testbeds.

Cyclic Short Time Fourier Transform (CSTFT)

CSTFT:

Virtual Sensor Grid Resource Infrastructure

iGIAB (INTEGRIDS Grid-in-a-Box)

Network-CentricSystem

iGIAB

More InteractionLess Interaction

iGIAB iGIAB iGIAB iGIAB

USGSServer

NWSServer

EPAServer

DRNAServer

(NOAA-JBNERRS)

Jobos NERRS Sensors(YSI 6600EDS,

WeatherStation, etc.)

DRNAServer

(Guanica Dry Forest Reserve)

UPRM-AIP Sensors(Xbow, Tmote,

Gumstix, Acoustics, etc.)

WALSAIPServer

Portal Host

Real-World Physical Signals

Operator Algebras Framework for Signal Analysis

2D Discrete Signal Spaces

One-Dimensional Discrete Finite Signals

)(2NZlx

Two-Dimensional Discrete Finite Signals

)(2, NNhx ZZla

One-Dimensional Signal Algebra

Operators

yxOx

ZlZlO

k

NNk

1

221 :

Time-Frequency Tools

hx

NNNN

ahxhx

ZZlZlZl

,

222

,,

:

Physical Signals

)(RLx

Sampling and Windowing

xyygg

ZlZlRL N

00

200

Two-Dimensional Signal Algebra

Operators baOa

ZZlZZlO

hxmhx

NNNNm

,2

,

222 :

1D and 2D Discrete Signal Spaces

Implementation on PRAGMA

Hardware Level

ProgrammingLevel

PRAGMACS 1

PRAGMACS 2

PRAGMACS N…

CS: Compute Site

C

MPI FFTW

C

MPI FFTW

C

MPI FFTW

NINF-G

G-FARM ApplicationLevel

LOCALCPUs

LOCALCPUs

LOCALCPUs

Application Development Tools

C

MPI

MPI

“Fastest Fourier Transform in the West.”

Ninf: A programming middleware which enables users to access resources on the Grid with an easy-to-use interface.

Gfarm File System: A next-generation network shared file system used as an infrastructure software.

PROGRAMMING TOOLS

SYSTEMRESOURCES

Conclusion and Future Works

Conclusion:The Concept of DiVISATime-Frequency Signal Analysis for Acoustical Environmental ApplicationsReal/Virtual Sensor Grid Resources PRAGMA as Community Resource

Future Works:Development of MPI-based Signal Analysis ApplicationsStudy Dynamic Behavior of PRAGMA Infrastructure for Signal-based Information Processing (SbIP).

QUESTIONS?