Page 60 - Fall_DTF

PI-iyaiaarflasad Signal Processing
5,“ ' Swdvupau. whale measured on a hydrophone in an underwater channel
M e,,__, onrtima "°“';,'j"_je'“‘“' adjacent to the Hawaiian Islands (Ou et al., 2013). As seen in
' ‘ Figure 2, top, the whale “song” consists of roughly four fre-
' ‘ V quency components that change slowly over time, resulting in
a chirp-like signal. The ability to distinguish biological signals
e -'.5...—. from other noise (e.g., anthropogenic signals) depends on the
—e ’ 7 ' i- H‘ __ ability to design joint time-frequency filters that are matched
t: ' 31'.‘ to the signal ofinterest. In Ou et al. (2013), this filter was im-
‘ plemented an image-processing technique that was
' H then applied to the acoustic spectrogram.
j ' '—"—' The ima e in Fi e 2, bottom, is the si ature of a surface
L , . , H 8 5'“ 3“
i.m¢ Hum O t‘ t _le_'e V vessel (ship). It was measured with a hydrophone deployed
T ‘ ‘:.....§r..r .3" "‘ "5...-r.r'-‘i'r" W in the Willamette River in the state of Oregon after propa-
Figure 1. Left: time-varying amplitude of :1 recorded sound. Right: gation and clear time-frequency structure can be observed.
f’5‘1"”“}’ 5'’"”'‘t '’f the Wm? 5""'”1 0l7’“i"9d b)’ '1 tflkiflg '1 Sh”? This structure is a result of both the radiation structure of
tirne Fourier transforrn with a sliding (and overlapping) window the source (vessel) and the propagation medh1etteh_ The
time-frequency filter is discussed later in this article.
1 V ' Spatial Eaamtforming
: ’ _ ..'. Beamforming is a signal-proces-sing technique that can be
1 e - .v-‘ __e.V , used to identify the direction of incoming (acoustic) energy.
3 .»/ : This is accomplished by using knowledge of the spatial struc-
_, {-3 f ~ ’ ture of the incoming wave front as measured across the acous-
-v ’ tic sensors deployed in an array. Although the processing can
' " ' " r—- i ' ’ ' be quite complex, as described in the article by Wage in this
mm“ ewe” eeme“ mm hm issue of Acoustics Today, it is based on the same principle used
in hearing to sense the direction of sound relative to the head.
. The delayin arrival of a sound from one ear to another is used
E no by the brain to determine the likely direction of the source
inn (i.e., “left” versus “right”). Deten-nination of the location or
E direction of a source is defined as localization.
uu Consider, for example, the simple case of a line array where
2 m an inn Ion’ mm hunt: I «on am am the elements are uniformly spaced (a uniform line array
"mm [ULA]). These types of arrays can be deployed horizontally
Figure 2. Speetrogmms of sound recorded underwater showing the as a horizontal line array (HLA), often towed from a surface
'i’“9:f’9'1"9"5)’5t’"”'4”3 Dfthe ’959iV9'15iX"“l- T01’-‘ 5P9‘"0X""" Of“ platform (such as a ship), or as a vertical line array (VLA)
humpback whale sound recorded in the Hawaiian islands (ou et al., tethered to the eeeeh hettem er ehethet ehtteee heme
2013). Bottorn: time-frequency signature ofa surface Vessel recorded . . . .
in the Willamette River, OR. Note the structure introduced by the f°"'“"gf':hthe P'°°ess'“g that w°"ld be used t° M311" ‘he
propagation physics (the "bathtul7'). A time-frequency filter was de- ‘°“““ ° 9 °“°‘3Y-
Signed b)’ "“5ki"Z the ll!/77771071!"-55 (1: 75110"). th€P10PflgI1ti'3" m0d'4' To implement a beamformer and compute the likely arrival
lation (2, red), and the Doppler shifts (3, blue) (Ogden et al., 2011). direction the eeeuetie held measured on th e may hydm_
phones is correlated with a vector of complex weights (the
Both time domain and frequency domain representations are “weight vector”) that represent the anticipated amplitude
useful, depending on the nature of the acoustic signal. Often, and phase structure of an incoming wave. Multiple weight
the ability to characterize the source of a signal is dependent vectors are computed, each one corresponding to a different
on a joint time-frequency signature. As an example, Figure possible arrival direction or steering direction (Van Trees
2, top, shows the time-frequency signature of a humpback and Bell, 1968). Under ideal circumstances, the peak of the
sa 1 AI:uuIl:l:I Tbday l i=oll2o1s