Detection of Underwater Acoustic Signals – Theory and Techniques
Wednesday, October 14
Presenter: Professor Roee Diamant
University of Haifa (Israel)
The detection of underwater signal is a key enabling technique for any active and passive underwater acoustic sensor applications. Technologies such as SONAR imaging, acoustic communication, depth detectors, and signal identification, all use as a backbone detection capability. Underwater acoustic detection is different from the established radio frequency techniques. Due to the fast-changing sea environment, approaches based on noise estimation face mismatches between the assume noise model and its actual distribution. Further, the significant multipath that arrive closely in time produce negative and positive superposition, which makes it hard to identify the first arrival. And the time-varying-frequency-selective characteristics of the acoustic channel makes it challenging to lock onto the received signal.
A simple binary detection is needed to distinguish between noise and signal. Since the distance to the emitter is also of interest, we target not only the detection rate and the false alarm rate, but also the time of detection. In the considered applications, either a template of the signal is known, or neither the structure of the detected signal nor its statistics are known (e.g., duration, carrier frequency, bandwidth, etc.). The signal is of known or unknown length and may change in time. With no training data available, a learning mechanism for detection is not possible and detection is highly challenging. Specifically, since the distribution of the ambient noise is unknown and may include noise transients, many false alarms may arise. In the considered harsh conditions, simple blind detection techniques, such as variants of the energy detector, are expected to fail. This is because when the distribution of the ambient noise is mismatched with the assumed model and a threshold-based detection is used; any small noise transient may trigger a false alarm. Another simple option is spectrum sensing through cyclostationary analysis, where detection is based on estimating some cyclic features the signal is assumed to possess (e.g., carrier frequency). However, in case the structure and statistics of the detected signal are unknown and the signal can be wideband or pseudo-random noise, we cannot assume that the signal is comprised of some cyclic features.
In this tutorial, we will examine the theoretical foundation of underwater acoustic signal detection and offer some practical techniques. We will introduce the existing noise models, and discuss ways to estimate their parameters. Based on these, several theoretical bounds for detection performance will be presented. Traditional and state-of-the-art detection schemes will be studied, and analyzed by their suitability to the underwater domain. We will then focus on tailor-made solutions for both active and passive underwater signal detection, and show some test cases on practical applications. In the second part of the tutorial, the attendees will practice on detection of signals over both simulations and using real signals recorded at sea. Finally, the current research challenges will be reviewed.
The objectives of tutorial are:
- Introduce basic approaches for underwater acoustic signal detection.
- Study practical noise distribution model for signal detection.
- Discuss current schemes for underwater acoustic signal detection.
- Experience techniques for signal detection.
The classes will include both theoretical analysis, simulation results, and a first-hand experiencing of acoustic detection of the participants over experimental data. Since live experiments will take place, the tutorial can accommodate up to 30 attendees.
The tutorial is targeted to electrical and computer engineers who has background on random processes, telecommunication, signals and systems, and probability analysis. The attendees should bring laptops with Matlab installed and be sufficiently comfortable in using Matlab.
The topics covered are highly relevant to engineers working on a verity of acoustic applications including the broad topics of:
Basics of underwater acoustics [45min]
- Propagation models
- Noise models
- Sound speed profiles
- Doppler shift
- The SONAR equation
Passive acoustic detection [45min]
- Energy detector
- Belief propagation
Active acoustic detection [45min]
- The matched filter
Threshold-less detection [45min]
Roee Diamant received his PhD from the Department of Electrical and Computer Engineering, University of British Columbia, in 2013, and his B.Sc. and the M.Sc. degrees from the Technion, Israel Institute of Technology, in 2002 and 2007, respectively. From 2001 to 2009, he worked in Rafael Advanced Defense Systems, Israel, as a project manager and systems engineer, where he developed a commercial underwater modem with network capabilities. In 2015 and 2016, he was a visiting Prof. at the University of Padova, Italy. In 2009, he received the Israel Excellent Worker First Place Award from the Israeli Presidential Institute. In 2010, he received the NSERC Vanier Canada Graduate Scholarship. Dr. Diamant has received three Best Paper awards, and serves as an associate editor for the IEEE Ocean Engineering. Currently, he is the coordinator of the EU H2020 project SYMBIOSIS (BG-14 track), and leads the underwater Acoustic and Navigation Laboratory (ANL) as an assist. Prof. at the Dept. of Marine Technology, University of Haifa. His research interests include underwater acoustic communication, underwater navigation, object identification, and classification.