N221-036 TITLE: DIGITAL ENGINEERING - Exploitation of Ephemeral Features in Sonar Classification Algorithms

OUSD (R&E) MODERNIZATION PRIORITY: General Warfighting Requirements (GWR)

TECHNOLOGY AREA(S): Sensors

The technology within this topic is restricted under the International Traffic in Arms Regulation (ITAR), 22 CFR Parts 120-130, which controls the export and import of defense-related material and services, including export of sensitive technical data, or the Export Administration Regulation (EAR), 15 CFR Parts 730-774, which controls dual use items. Offerors must disclose any proposed use of foreign nationals (FNs), their country(ies) of origin, the type of visa or work permit possessed, and the statement of work (SOW) tasks intended for accomplishment by the FN(s) in accordance with the Announcement. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws.

OBJECTIVE: Develop automated classification techniques that improve performance with ephemeral features in active and passive sonar systems.

DESCRIPTION: In active and passive sonar systems, automated processing can include classification algorithms to reject false alarms (that is, clutter) while retaining true target detections.

State of the art classification algorithms are, in general, multiple hypothesis tests that can be implemented by extracting features from the acoustic measurement associated with each detected "contact". The features distill different characteristics of the measured sound such as aural features or descriptive features, similar to how arches, loops, and whorls help classify fingerprints. In automated sonar systems, the features extracted from a sonar contact are typically combined using non-linear algorithms to identify the class to which the contact belongs. In most cases, these algorithms have parameters that must be learned (that is, the classifier is trained) through analysis of existing data that has already been labeled as to its class.

The availability of off-the-shelf classifiers such as multiple hypothesis testing and machine learning tools, has enabled the development and testing of numerous features. A limitation of most off-the-shelf algorithms is that they typically assume every feature is available all the time. However, not all features are viable in every contact. Some are missing only occasionally and some only occur rarely (that is, ephemeral features). An example of such an ephemeral feature in facial recognition would be the shape of the nose, mouth, and chin during time when some are wearing masks. Sonar similarly has such recognition features that may be missing, either because the environment masks the feature or because submarines, trying to be stealthy by design, try to hide such "features." The standard approaches for handling this missing-data problem deal with it indirectly (such as, by replacing the missing feature with its mean over the training data). They may also incorrectly assume missing features occur in a uniformly random manner throughout the data. As such, the standard approaches to missing data do not fully exploit the information contained in ephemeral features when they exist. Expanding sonar classification to include ephemeral features (features that are not always present) will give Navy systems increased capacity 1) to detect stealthy submarines or torpedoes that may not otherwise be detected or 2) to accelerate detection of targets in cases where time to react is limited and the consequences of delayed detection are potentially fatal.

One system where ephemeral features exist is the AN/SQQ-89A(V)15, which contains functions for pulsed active sonar, continuous active sonar, towed array passive sonar, and hull array passive sonar. Technology developed under this topic should be extensible to each of these functions, with initial adoption most likely to occur within the pulsed active sonar function.

The ideal solution will exploit off-the-shelf classifier technology, have practical implementation and training procedures, and handle features that occur rarely or frequently. As real-world data sets associated with the AN/SQQ-89A(V)15 are classified, companies are encouraged to plan to obtain or generate unclassified data sets that demonstrate their solution.

While proposers are encouraged to demonstrate the power of their approach on unclassified data they have obtained, created, or simulated, the Phase II effort will involve tests of the technology developed under this topic with recorded data provided by the Navy both to assess stand-alone performance, as well as provide for the technology to be assessed within the overall sonar processing string. Details of exactly how this is to occur will be dependent on the nature of the proposed technology. Once the technology is independently deemed to provide value, the Navy will commit to incorporating the technology into a future sonar system build, which will go through certification as an integrated system.

Work produced in Phase II may become classified. Note: The prospective contractor(s) must be U.S. Owned and Operated with no Foreign Influence as defined by DOD 5220.22-M, National Industrial Security Program Operating Manual, unless acceptable mitigating procedures can and have been implemented and approved by the Defense Counterintelligence Security Agency (DCSA), formerly the Defense Security Service (DSS). The selected contractor must be able to acquire and maintain a secret level facility and Personnel Security Clearances, in order to perform on advanced phases of this contract as set forth by DCSA and NAVSEA in order to gain access to classified information pertaining to the national defense of the United States and its allies; this will be an inherent requirement. The selected company will be required to safeguard classified material IAW DoD 5220.22-M during the advance phases of this contract.

PHASE I: Develop a concept for improved sonar classification algorithms with ephemeral features that meet the parameters of the Description. Demonstrate the feasibility of the concept through modelling, analysis, and simulation. The Phase I Option, if exercised, will include the initial design specifications and capabilities description to build a prototype solution in Phase II.

PHASE II: Develop and deliver a prototype improved sonar classification algorithm with ephemeral features based on the results of the research in Phase I. Demonstrate the prototype meets the required range of desired performance attributes given in the Description. System performance will be demonstrated through installation and prototype testing on a testbed with the lead system integrator.

It is probable that the work under this effort will be classified under Phase II (see Description section for details).

PHASE III DUAL USE APPLICATIONS: Assist the Navy in transitioning the technology for Navy use in an operationally relevant environment to allow for further experimentation and refinement. The prototype algorithm will be integrated into the PEO-IWS 5 surface ship ASW combat system Advanced Capability Build (ACB) program used to update the AN/SQQ-89 Program of Record.

Commercial applications that could benefit from ephemeral features of sonar classification algorithms include both active and passive remote-sensing systems where the data includes ephemeral features. Examples outside of sonar include most applications of radar, lidar, satellite remote sensing, ultrasound, and thermal imaging.

REFERENCES:

1. Young, V. W. and Hines, P. C., "Perception-based automatic classification of impulsive-source active sonar echoes," Journal of the Acoustical Society of America, 122:3, pp. 1502-1517, September 2007. For libraries local to you holding this article, see https://www.worldcat.org/title/perception-based-automatic-classification-of-impulsive-source-active-sonar-echoes/oclc/211513436&referer=brief_results, accessed 3/31/2021.
2. Murphy, S. M. and Hines, P. C., "Examining the robustness of automated aural classification of active sonar echoes," The Journal of the Acoustical Society of America, 135:2, pp. 626-636, February 2014. For libraries local to you holding this article, see https://www.worldcat.org/title/examining-the-robustness-of-automated-aural-classification-of-active-sonar-echoes/oclc/5537024626&referer=brief_results, accessed 3/31/2021.
3. Buss, M., Benen, S., Stiller, D., Kraus, D. , and Kummert, A. , "Feature selection and classification for false alarm reduction on active diver detection sonar data," in Proceedings of 4th Underwater Acoustics Conference and Exhibition (UACE2017), pp. 569-576, 2017. At https://www.uaconferences.org/docs/Past_proceedings/UACE2017_Proceedings.pdf, accessed 3/31/2021.
4. Hastie, T. , Tibshirami, R. , and Friedman, J. , The Elements of Statistical Learning, Springer, 2009. For libraries local to you holding this article, see https://www.worldcat.org/title/elements-of-statistical-learning-data-mining-inference-and-prediction/oclc/1080370824&referer=brief_results, accessed 3/31/2021.
5. Eibe Frank, Mark A. Hall, and Ian H. Witten (2016). The WEKA Workbench. Online Appendix for "Data Mining: Practical Machine Learning Tools and Techniques", Morgan Kaufmann, Fourth Edition, 2016. For libraries local to you holding this book, see https://www.worldcat.org/title/data-mining-practical-machine-learning-tools-and-techniques-ed-4/oclc/1242613909&referer=brief_results, accessed 6/3/2021 Weka 3: Machine Learning Software in Java is available at https://www.cs.waikato.ac.nz/ml/weka/, accessed 3/31/2021.

KEYWORDS: Classifier; sonar contact; ephemeral features of sonar; classification algorithms; machine learning, multiple hypothesis testing