OVERVIEW

Objectives – Methods – Accomplishments

In recent years, sensor network systems have been proposed for various monitoring applications. In the past, most reported systems involved custom-made hardware. In this work, we considered three types of COTS platform for acoustic beamforming, namely, the iPAQ-based platform, Stargate-based platform, and the MOTE-based platform.

The first type of platform is consist of Compaq iPAQ 3760s with their built-in StrongARM processors, ROM and RAM memories, microphones and codecs for acoustic acquisition and processing. They also include external wireless Ethernet cards for radio communication to form a distributed sensor network, which we use to perform acoustical beamforming. Time synchronization among the microphones is achieved by the Reference-Broadcast Synchronization method. The built-in microphones of iPAQ 3760 have a low sensitivity and thus good for short-range acoustic monitoring.

To facilitate the study of the wireless sensor network for demanding acoustic Monitoring of long distance sources, we recently have also started working on the development of a new generation of wireless acoustic sensor network platform using the Stargate nodes. The 400 MHz PXA-255 XScale processor and the 64 MB SDRAM provide the Stargate platform a better processing capability. The VX Pocket 440 sound card with four external microphones are attached to each Stargate node through the PCMCIA slot for acoustic data acquisition. We use highly-sensitive external microphones with low self-noise. The VXPocket 440 sound card supports data sampling up to 48 KHz in 24 bits. Each Stargate node also has a flash interface to connect a 802.11 card for wireless communication. The Stargate-based platform is for the acoustic monitoring of woodpeckers in the James Reserve and the like.

Two beamforming algorithms, based on the time difference of arrivals (TDOAs) among the microphones and least-squares estimation of the TDOAs method, and the maximum-likelihood (ML) parameter estimation method, are used to perform source detection, enhancement, localization, delay-steered beamforming, and direction-of-arrival estimation. Measurements in free-space and most recently preliminary work in reverberant scenarios have demonstrated the effectiveness of the array processing algorithms and sensor platform for real-time beamforming applications. Most work in this area involve either proposing some new algorithms and verifying them either by analysis or simulation or building some hardware systems using well established algorithms. The fact we were able to implement these two state-of-the-art array processing algorithms onto a hardware system in limited time and resources and able to perform real-time operation is quite remarkable and significant. Several journal and conference publications have reported these works.

SYSTEMS / EXPERIMENTS

From 2004 and 2005:

Fig. 1 shows a circular array with 8 iPAQs wireless time-synchronized COTS sensor platforms facing one cement wall (Wall 1) and one cement floor (Wall 2).  Fig. 2 shows the locations of the circular subarray (at 1.5 m from Wall 2 and 3.5 m from Wall 1) and the true source location (at 1.7m from Wall 2 and 6 m from Wall 1), as well as that of the estimated location of the source. As can be seen, the estimated location is still very close to the true location even with a considerable amount of reverberations from the two cement walls.

From 2003:

Two beamforming algorithms, based on the time difference of arrivals (TDOAs) among the microphones and least-squares estimation of the TDOAs method, and the maximum-likelihood (ML) parameter estimation method, are used to perform source detection, enhancement, localization, delay-steered beamforming, and direction-of-arrival estimation.

ACCOMPLISHMENTS

From 2004 and 2005:

From our earlier work, we have shown that the Approximate-Maximum-Likelihood (AML) criterion algorithm was near optimum for estimating the direction-of-arrival (DOA) and/or the location(s) of one or more source(s).  This approach is applicable to acoustic as well as seismic array processing.  In the past, almost all known practical acoustic and seismic DOA and localization techniques assumed none or small amount of reverberation.  In 2004, we formulated a novel virtual array formulation in which if the location(s) of the subarray(s) are known relative to the nearby reflecting walls/floors, then under this controlled reverberation scenario, reliable estimation of the DOA(s) and localization(s) of the source(s) can still be achieved.  Field measurements were conducted, and a journal paper reporting on the successful use of this approach appeared as an invited paper in the IEEE Transactions on Mobile Computing.

From 2003:

Advances in microelectronics, array processing, and wireless networking, have motivated the analysis and design of low-cost integrated sensing, computing, and communicating nodes capable of performing various demanding collaborative space-time processing tasks. In this work, we consider the problem of coherent acoustic sensor array processing and localization on distributed wireless sensor networks. We first introduce some basic concepts of beamforming and localization for wideband acoustic sources. A review of various known localization algorithms based on time-delay followed by LS estimations as well as maximum likelihood method is given.

Issues related to practical implementation of coherent array processing including the need for fine-grain time synchronization were addressed. Then we describe the implementation of a Linux-based wireless networked acoustic sensor array testbed, utilizing commercially available iPAQs with built in microphones, codecs, and microprocessors, plus wireless Ethernet cards, to perform acoustic source localization. Various field-measured results using two localization algorithms show the effectiveness of the proposed testbed. Solution of the synchronization problem has allowed algorithms that were under development for many years to finally be demonstrated in hardware.

FUTURE DIRECTIONS

From 2004 and 2005

We are continuing with our collection of acoustical and seismic array data both indoor as well as out at the open fields. Recently, we are building three subarrays using 12 high quality LinearX microphones. We are also expecting to do some measurements with high quality 3-axis accelerometers. We plan to investigate both at the theoretical level as well as with the fusion of acoustic and seimic array field measured data. We are also planning to conduct measurements with 3-D AML-based acoustical subarrays at James Reserve to perform detection, location, and identification of various birds.

From 2003

In area (A), our main effort will be to implement real-time processing for the wireless time-synchronized COTS sensor platform for beamforming applications in reverberant conditions. Most realistic environments have various amount of reverberations due to walls/floors/ceilings inside a room or nearby reflecting surfaces/trees, etc. in outdoor scenarios. Our proposed novel "virtual array model" for reverberant beamforming conditions has been verified under simulation and limited outdoor conditions. We plan to continue work in this direction under more realistic reverberant conditions. Another effort is to initiate the implementation of a low-cost real-time CCD camera system capable of being steered toward a source located by the beamforming sensor platform. For various applications such as after determining the location of a woodpecker and enhancing its acoustical signature (by beamforming methods) for its identification, a visual record of the bird can complement its identification by its acoustical signature. Various sophisticated sensor fusion concepts have been proposed for many years. Here we propose a simple but realistic fusing of acoustical signature information with image information for the enhanced identification of the specie of the bird. The final effort in this topic is continue the development of the Stargate processor. The 400 MHz Xscale processor of the Stargate nodes is twice is powerful as the 200 MHz Strongarm processor of the iPAQ 3670 used as the current generation of wireless acoustic sensor network platform. Considerable future work need to be performed to obtain the acoustic data acquisition capability on these Stargate nodes. The goal of these new nodes is for higher quality and longer range acoustic source acquisition for James Reserve habitat applications.

PEOPLE

FACULTY

K. Yao

GRADUATE STUDENTS

A. Ali
S. Asgari
P. Bergamo
C.E. Chen
D. Maniezzo
H. Wang
L. Yip
J. Zhao

RESEARCH ENGINEER

Dr. R. E. Hudson