A Multiple Signal Classification Method for Directional Gravitational-wave Burst Search

1
A Multiple Signal Classification Method for
Directional Gravitational-wave Burst Search

• Junwei Cao
• LIGO Scientific Collaboration Research Group
• Tsinghua University, Beijing, China
• 3rd Galileo - Xu Guangqi meeting
• October 12, 2011

2
Outline

• Introduction
• Real-time / low-latency GW burst search
• Motivation running before data
• Our method
• Multiple signal classification (MUSIC)
• Extension for GW DOA
• Performance metrics
• Performance evaluation
• Performance Comparison
• Conclusions

3
Our Group

• The LSC member group in China, including 3
  faculty members and 3 students
• GW burst data analysis and computing
  infrastructure
• Also involved in LCGT, AIGO and ASTROD
• With close collaboration with MIT, Caltech and
  UWA
• This talk provides an introduction to one of our
  existing efforts on real-time / low latency GW
  burst search

4
LSC Burst Group

• Mission Detection of unmodeled bursts of
  gravitational radiation
• Three dedicated pipelines
• Coherent Wave Burst (CWB) Pipeline
• S. Klimenko et al, Class.Quant.Grav.25114029,2008
  
• Kleine Welle for online detector characterization
• LIGO Document, LIGO-G050158-00-Z, 2005
• Omega Pipeline
• https//geco.phys.columbia.edu/omega
• One group-crossed pipeline
• X-Pipeline for directional search
• https//geco.phys.columbia.edu/xpipeline

5
Real-time Search

• Real-time between online and offline mode for
  large-scale data analysis

Online Monitoring
Data Streams
On-site
Real-time Search
Data Streams Data Production
On-site Off-site
Offline Analysis
Data Production
Off-site
6
Motivation

• Prompt E/M follow-up by LIGOs external
  collaborators
• Detect astronomy events earlier than traditional
  observation methods
• Increase the confidence of the GW candidate
  event
• Obtain more information about GW candidate event
  and its source more accurate sky position,
  distance,
• Rapid detector characterization
• gt New algorithms, methods and computing
  technology to enable faster real-time search, in
  particular, directional search

7
Challenges in AdvLIGO

• More potential IFOs LCGT, AIGO,
• More data streams flood into central location
• Larger Data Volume

Cite from LIGO-G0900008
8
MUSIC

• The multiple signal classification (MUSIC)
  algorithm is one of the most popular
  subspace-based techniques for estimating the
  directions-of-arrival (DOAs) from linearly
  arrayed signal detectors.
• Dividing eigenspace to noise and signal
  subspaces, which are perpendicular to each other
• Giving arbitrary locations and arbitrary
  directional characteristics in a noisy
  environment of arbitrary covariance matrix, MUSIC
  is capable of giving asymptotically unbiased
  estimates of
• Number of signals
• DOA
• Strengths and cross correslations among the
  directional waveforms
• Polarizations
• Strength of noise or interference

9
MUSIC Extensions

• MUSIC is widely used in periodic sine radio wave
  detection by antenna arrays in the plane
  condition. Several aspects are extended before
  applying MUSIC on GW burst search
• Using Spherical coordinates to extend from 2D to
  3D
• Using the concept of equal-phase to extend
  linearly arrayed detectors to generally placed
  detectors
• Using linear transformation in time domain to
  extend the method to non-periodic signals

10
MUSIC Steps
Collect data and form the covariance matrix S
Calculate the Eigen structure of S in the matrix
S0
Assuming that there is one signal in a relatively
long period of time, get the eigenvectors of the
noise subspace with the number of M-1 (M is the
number of detectors)
Calculate the Pmu(?) and put it in a figure
Find the peak of the signal
Get DOA and other information of interest
11
Performance Evaluation
12
Experiment Design

• Self-generated Gaussian-moderated sinusoidal GW
  is injected into simulated LIGO data background.
• IIR filtering MUSIC vs. Omega Bayesian

13
Comparison Results

• The comparison result of MUSIC acting as the
  signal trigger versus Omega (Q transform).
• (Define A as the relative signal strength, which
  comes from the parameter of Factor of LogFile of
  injection part. A typical GW has a strength A1)

Parameters Low Limit of A Time Resolution Time Consuming
Omega 2 0.015s 14s
MUSIC 200 0.03s 3200s
14
Comparison Results

• The comparison result of MUSIC acting as DOA
  evaluator versus Bayesian.

Parameters Low Limit of A Angel Resolution Time Consuming
Bayesian 4 0.019rad 30s
MUSIC 1000 Complicated 4.2s
15
Bayesian Results

• Bayesian skymap A100

16
MUSIC Results
17
Conclusion

• Current burst real-time low latency search is
  successful, but not perfect
• Advanced computing technology and new signal
  processing methods can significantly boost
  real-time multi-messenger astronomy
• Multiple signal classification have potential to
  provide faster direction estimation, though
  current SNR ratio resolution and time resolution
  are not satisfactory