Acoustic Target Tracking Using Tiny Wireless Sensor Devices

1
Acoustic Target Tracking Using Tiny Wireless
Sensor Devices

• Qixin Wang, Wei-Peng Chen, Rong Zheng, Kihwal
  Lee, and Lui Sha
• Dept. of CS, UIUC

2
Introduction

• Context
• Delay based sound source locating algorithm,
  requires large number of redundant sensors for
  accuracy.
• -Tiny wireless sensors to real-world acoustic
  tracking applications.
• Tracking only impulsive acoustic signals, such as
  foot steps, sniper shots etc. No concept of
  tracking motion.

3
Introduction

• Challenges
• Partial info at one sensor site
• Inaccuracy and unreliability of sensors
• Effective use of scarce wireless bandwidth

• Solutions
• Sensor clustering and coordination
• Redundancy for robustness
• Quality-driven (QDR) networking. Info. flow
  oriented v.s. raw data flow oriented.

4
Introduction
Scenario
Sensor
Router
Cluster Head
Sink/Pursuer
Cluster Head
Sink/ Pursuer
5
System Overview

• System Architecture
• Acoustic target tracking subsystem

Sensor (mica motes)
Sensors belong to clusters with singular cluster
head. Cluster head knows the locations of its
slave sensors. Raw data gathered from sensors are
processed in cluster head to generate
localization results
Cluster Head (mono-board computer)
6
System Overview

• Communication Subsystem route back the reports
  generated by cluster heads to sink

Sink
cluster covered area
cluster head
router (mica motes)
7
Acoustic Target Tracking Subsystem

• Use RBS Time Synch (error ? 30?s).
• Onset Detection (on sensors)
• Small sliding window to compute moving average of
  acoustic signal magnitude.
• Use threshold to detect onset time t0.
• Record one buffer load of data, then
  post-process.

8
Acoustic Target Tracking Subsystem

• Cross Correlation (to find out delays)

Locate sound src loc.
Detected intersted sound
Broadcast sound signature
ClusterHead
Cross-correlation to detect local arrival time
Report local arrival time
SlaveSensor
9
Acoustic Target Tracking Subsystem

• Sound Source Locating Evaluation of Quality
  Rank (main idea)
• Throw away apparently erroneous sensor readings.
• Let A clusters monitored area, sound src
  location arg?p?Amind(p) - ds,where d(p)
  is the hypothetical sensors sound arrival time
  vector, while ds is the actual one. is an
  error measurement function.

10
Acoustic Target Tracking Subsystem

• In practice, we cannot check every location in A,
  instead, we apply a grid with 3?3inch2
  granularity onto A, and only check those grid
  points.
• Quality Rank percentage of d(p)s elements that
  falls outside ? boundary of ds .

11
Communication Subsystem

• Quality-driven(QDR) Redundancy Suppression and
  Contention Resolution
• Redundant clusters may report same events
  location. Good for reliability reasons.
• Quality Rank is used to suppress inferior reports
  and only report high quality rank localization
  reports to data sink

12
Acoustic Target Tracking Subsystem

• Quality Rank is also used for contention
  resolution along the routes (with CSMA as MAC) to
  let higher quality reports get to data sink
  earlierTbackoff QualityRank ? interval
  random

13
Experiment

• Locations of sensors and sound sources in a
  single cluster

14
Experiment

• Examples of localization results for different
  sound source locations

15
Experiment

• Average error vs. sound source locations. Note
  sound source is a 4inch speaker

16
Experiment
17
Experiment

• of reports within 3-inch error range higher
  quality rank, higher creditabi-lity

18
Experiment

• Quality-driven (QDR) Effect of various interval
  on the percen-tage of suppressed reports

19
Experiment

• Effect of Quality-driven(QDR)

Suppose info/bit is fixed the smaller Quality
Rank, the better the quality.
20
Conclusion

• Acoustic target tracking using tiny wireless
  devices with satisfying accuracy is possible.
• Quality Rank can be used to decide the quality of
  tracking result
• Quality-driven redundancy suppression and
  contention resolution is effective in improving
  the information throughput.