Page 66 - 2018Fall

Signal Processing
timation. Although human voiced acoustic communication is not a central focus of the TCSP, acoustic transmission of modulated data through spatially complex and temporally dynamic media has been an enduring interest within the SP community.
More broadly, the TCSP provides methodologies of analy- sis for inference problems in an open environment where diverse disciplines can come together to sharpen their ap- proaches and permit the cross-fertilization of methods. The TCSP is a place where the individual researcher can find an approach to analysis that best fits his/her observational assets and his/her inference needs. From the most funda- mental epistemological questions regarding probabilistic reasoning to the seemingly mundane and yet essential ones regarding computational efficiency, the TCSP community is an open- and generous-minded group. For these reasons, there is a very strong appreciation of the multidisciplinary nature of acoustics within the TCSP, and this reverence holds the TCSP in close proximity to other TCs. We have members with strong ties to TCs in underwater acoustics, noise, acoustical oceanography, architectural acoustics, and animal bioacoustics as well as others with ties to the number of jointly sponsored special sessions reflecting this deep and enduring synergy. Within the TCSP, it is not uncommon to see lively and productive interactions on fundamental issues between architectural acousticians and underwater acous- ticians. The TCSP takes great satisfaction in the broad dis- semination of methodologies that enhance and improve the accuracy of acoustic inferences across the diverse disciplines within acoustics. Because of this, it is a natural organic aim of the TCSP that our members enrich and enliven the vari- ous TCs with which they interact.
The TCSP is devoted to the development of a full range of statistical inference methods as necessary in acoustic infer- ence. The SP community continues to be an environment where acoustic modeling issues are addressed from the per- spective of model efficacy as supported by observations. This aspect of the TCSP takes many instances, from practi- cal risk-minimizing parameter point estimation (Gendron, 2016; Michaloupolou and Pole, 2016) to full posterior infer- ence (Dosso, 2002; Michaloupolou and Gerstoft, 2016) to the development of a framework for a fair comparison of disparate models from acoustic measurements (Dettmer et al., 2010; Xiang and Fackler, 2015).
Providing Practical Solutions
We continue to see the TCSP contribute solutions to impor- tant and useful problems in acoustics. For instance, the vital need for determining the location of the discharge of a fire- arm with a ballistic model-based method (Lo and Ferguson, 2015) directly addresses the problem of ranging small-arms fire to save lives. The approach operates with no more than a single acoustic sensor node collocated with the vital target and is accomplished without a priori knowledge of the muz- zle speed and ballistic constant of the bullet. The method re- quires the extraction of the differential time of arrival and angle of arrival of the muzzle blast and ballistic shock wave at the sensor node.
Another practical example of the reach of the TCSP is the nondestructive testing of critical containers by acoustically exploring for cracks (Anderson et al., 2017). The method fo- cuses high-frequency elastic energy to a point to probe for an anomalous fissure. The approach has been demonstrated to save time and resources in the necessary task of helping determine the strength and endurance of vital stainless steel canisters whose failure and breach could pose a significant risk to human life.
Recently, there has been much interest in the optimal use of extremely sparse acoustic arrays, with the methods em- ployed being termed “compressive sensing” (CS). Acoustic CS is the processing of sparse configurations of acoustic sen- sors for the purpose of drawing inferences regarding some acoustically relevant feature. CS asserts that such underde- termined problems can be solved provided the feature is likewise sparse. An example is the estimation of an ocean acoustic sound-speed profile (SSP) from the inversion of acoustic fields measured on sparse arrays (Bianco and Ger- stoft, 2016). CS has been demonstrated to estimate SSPs in a range-independent shallow ocean by inverting a nonlinear acoustic propagation model. The demerits of sparse configu- rations are of practical interest because sensor elements can and do fail and are not often able to be quickly replaced. CS methods have been compared with conventional beamform- ing using at-sea horizontal towed arrays for the discerning of an acoustic target bearing with comparisons in terms of a signal-to-interference ratio by Edelmann and Gaumond (2011). Sparse configurations of sensors can also take on purposeful and highly structured forms for reducing the computational burden of beamforming with coprime arrays (e.g., Adhikari et al., 2014; Xiang et al., 2015).
64 | Acoustics Today | Fall 2017