Page 8 - Summer 2008

operational frequency band and therefore the sensors used to measure the propagated data with wavelengths inversely pro- portional to the source frequency. Clearly, compared to RF transmissions, acoustic waves propagate at long wavelengths
1-4
Scattering of waves is determined by the object size at a wavelength or less. The point is that the advantage or disadvantage of acoustical waves over other modalities is determined by the acoustical properties of the source as well as the materials composing the propagation medium or environment. For instance, the audible acoustic range is the frequency band from 20 Hz to 20 kHz and is measured by microphone sensors, while seis- mic signals reside in the 0 to 10 Hz band measured by net- works of geophones. Inaudible ultrasonic signals are typical- ly in the 20 kHz to 20 MHz regime measured by piezoelectric
1,3,4
and are affected by materials much differently.
crystal sensors.
Acoustics can be used to perform the usual tasks of
detection, classification and localization much the same as RF with sonar replacing radar in the active problem. For instance, acoustic sources can be localized by triangulariza- tion techniques much like finding the epicenter of an earth- quake by a worldwide network of seismometers using their known location and the arrival times of the seismic event. Because of the large wavelengths in acoustic signals, arrays can be designed for coherent (phase) processing, that is, an array of acoustic sensors can be used to localize an acoustic source and passively scan an environment through beam steering while using beamforming techniques to search for sources. The human ear is a perfect example of passive (coherent) source localization in the audible range. In more complex environments, physics (model) based techniques can be used to enhance further the measured signal and per-
5-7
extract the desired information from the noisy measure- ments. For example, if a voice recording was available, then the microphone data could be processed to separate the voice signature from the environmental background and noise thereby decomposing the acoustic information of each. Kennedy assassination data was thoroughly analyzed from microphone data acquired from available audio and video/audio recordings. The number and location of shots
2
fired were extracted from the recordings. The frequency
characteristics of the sounds were analyzed even further to extract information about the environment (time-frequency estimation). Active acoustics (sonar) is typical in many appli- cations ranging from the underwater tracking of a moving target, to locating tumors in tissue for biomedical applica- tions, to imaging materials using an acoustic microscope. Vibrational acoustic signals provide critical information about structures and their condition in structural integrity studies. Thus, acoustical data much like RF or radar data uniquely provides information about the source, background environment, and noise that can be processed to extract use- ful information depending on the source characteristics and the supporting propagation medium as well as objects popu- lating that particular environment (e.g., urban environment with buildings).
form the localization.
The processing of acoustic data is necessary to
Signal processing approach
Signal processing relies on any a priori knowledge of the phenomenology generating the underlying measurements. Characterizing this phenomenology and propagation physics along with the accompanying measurement instrumentation and noise are the preliminaries that all acousticians must tackle to solve such a processing problem. In many cases this is much easier said than done. The first step is to determine what the desired information is and typically this is not the task of the signal processor, but that of the acoustician per- forming the study. In our case, we assume that the investiga- tion is to extract information stemming from acoustic signals either emanating from a source whether it be an autonomous unmanned vehicle (AUV) passively operating in the deep ocean or a vibrating structure responding to ground motion. Acoustic applications can be very complex especially in the case of ultrasound propagating through complex media such as tissue in biomedical applications or through heteroge- neous materials of critical parts in nondestructive evaluation
1,3,4
detection and localization in both biomedical and NDE.
Another view of the same problem is to decompose it
into a set of steps that capture the strategic essence of the pro-
cessing scheme. Inherently, we believe that the more a priori
knowledge about the measurement and its underlying phe-
nomenology we can incorporate into the processor, the bet-
ter we can expect it to perform—as long as the information
that is included is correct. One strategy, called the “model-
based approach,” provides the essence of model-based signal
6,7
In any case the processing usually involves manipulating the measured data to extract the desired information, such as, location and tracking of the AUV, to failure detection for the structure, or tumor/flaw
6,7
(NDE) investigations.
Some believe that all of the signal processing schemes can be cast into this generic framework. Simply, the model-based approach is “incorporating mathematical mod- els of both physical phenomenology and the measurement process (including noise) into the processor to extract the desired information.” This approach provides a mechanism to incorporate knowledge of the underlying physics or dynamics in the form of mathematical propagation models along with measurement system models and accompanying uncertainties such as instrumentation noise or ambient noise as well as model uncertainties directly into the resulting processor. In this way the model-based processor enables the interpretation of results directly in terms of the problem physics. The model-based processor is really an acoustic modeler’s tool enabling the incorporation of any a priori information about the particular application problem to extract the desired information. As depicted in Fig. 1, the fidelity of the incorporated model determines the complexi- ty of the processor. These models can range from simple implied non-physical representations of the measurement data such as the Fourier or wavelet transforms to parametric black-box models used for data prediction, to lumped math- ematical physical representations characterized by ordinary differential equations, and to full physical partial differential equation models capturing the critical details of the acoustic wave propagation in a complex medium. The dominating
processing.
Signal Processing in Acoustics 7