Underwater systems are always prone to complex environmental conditions and various interferences that is the reason why Sonar images are filled up with contamination, noise, and blurriness. Collection of real-world underwater images data is very costly with respect to human resources, time, equipment and logistics. Due to these challenges, there are very few publicly available datasets of underwater sonar images and only few datasets are available for underwater images which are scarce because of lack of resources. Precise and fast detection of objects in underwater sonar images is very less explored and still an open challenge for underwater systems. This paper evaluates performance of state-of-the-art object detection deep learning network YOLO (You Only Look Once) for underwater sonar images through transfer learning on publicly available benchmark Marine Debris dataset and custom-tailored dataset captured in real sea environment. Experimental results shows that YOLOv8n was able to achieve high mAP score of 95.3% on Marine Debris dataset with inference speed of 1.3ms.

This paper addresses the efficient implementation of signal processing algorithms tailored for multichannel sensor array systems, with a particular focus on an underwater acoustic surveillance and detection application employing both linear and circular sensor arrays. Real-time processing of data from these arrays necessitates an optimized implementation of signal processing onto digital hardware. The paper delineates the processing requirements of signal processing algorithms suitable for sensor array systems, providing mathematical formulations to compute their processing load. By evaluating linear and circular sensor array configurations, it quantifies the processing load of key algorithms such as beamforming and spectral analysis, aiming to guide the selection of an appropriate processing architecture leveraging Commercial Off-The-Shelf (COTS) processors. The study proposes an implementation approach emphasizing the average time cost of signal processing algorithms using contemporary COTS processors. Results demonstrate that the latest Intel-based multicore processors exhibit comparable efficiency to equivalent Digital Signal Processing (DSP) processors in effectively implementing signal processing algorithms, thereby facilitating real-time system operation.

The paper presents a comparison of two subarray based beamforming techniques namely the conventional frequency domain beamformer (CBF) and the split-beam cross-correlation (SBCC) beamformer, in terms of their ability to estimate the direction-of-arrival (DoA) of potential targets at sea. First the CBF technique is examined with three small sensor arrays called subarrays, separated by a fixed distance and placed in a line array configuration. Second is the SBCC technique that is widely used in array processing applications for DoA estimation of broadband signals using a linear array. In this paper, the SBCC technique is used to generate a fixed number of beams for each of the three subarrays, followed by cross-correlation of beams from first subarray with beams from second subarray or the cross-correlation of beams from third subarray with beams from second subarray. The second subarray in the middle is taken as a reference. Simulation results show that the SBCC technique using subarrays provides better DoA estimates for broadband signals radiated by targets at sea in comparison to the CBF technique. The results are presented in the form of bearing vs. amplitude and bearing vs. time plots for comparison.

This paper presents the design of a low-noise, high-quality pre-amplifier optimized for an acoustic application related to high frequency underwater imaging. The proposed low-noise pre-amplifier (LNA) design is focused on amplifying low-power signals while maintaining a minimal signal-to-noise ratio (SNR) degradation. Noise reduction is achieved through careful selection of components, configurations, and operating points within the operational bandwidth. Key design goals, including impedance matching, variable gain (20-80 dB using 0-5 VDC control signal), a central frequency of 400 kHz, and a 10 kHz bandwidth, are met. The proposed pre-amplifier features a two-stage, adjustable gain of 80 dB, followed by a four-stage, 8th order filter using a multiple feedback topology. This design demonstrates significant improvements, achieving an input-referred voltage noise of 0.488nVrms /√Hz, a bandwidth of 9.8 kHz, and a quality factor of 41, surpassing conventional LNA designs.

Underwater acoustic data classification has received significant attention from the research community in recent years because of its potential applications in underwater object detection and classification. Underwater acoustic data classification presents a challenging problem due to several factors, including the complicated sound waves propagation in marine environment, diminished illumination, frequency-dependent absorption, dispersion of light, and the presence of complex background noise. Underwater sound propagation is largely effected by underwater environment and geographical change that is sound of same vessel is observed with different characteristics. The artificial intelligence based system may not perform well to classify the same data in different sea environments because model trained in one environment data may not perform well in other environment data. There is a need to make AI based system capable to learn in different sea environmental conditions and geographical location. In underwater acoustic classification field, it is very important for artificial intelligence based models to be able to learn and adapt across different tasks. To fill this gap, continual learning techniques are applied on the underwater acoustic dataset in order to improve classification accuracy. The proposed model is evaluated on benchmark real world dataset and the results demonstrates the superiority of the proposed model. Two famous techniques of continual learning are implemented in this proposed model which are experience replay and elastic weight consolidation. Experience replay techniques shows perfect balance between stability (not forgetting old knowledge) and plasticity (ablility to learn new tasks) as compare to elastic weight consolidation. Elastic weight consolidation also shown significant improved classification accuracy.

Low-noise amplifiers (LNAs) are designed to amplify only the desired signal while minimizing noise, making them essential for applications requiring precise signal detection, processing in low SNR and high-sensitivity environments. This paper presents a novel LNA design aimed at enhancing signal reception in underwater imaging and communication applications. The proposed design achieves a wideband amplifier with specific parameters: a variable gain range of 40 dB to 80 dB, a broad 30 kHz bandwidth centered at 20 kHz, and a low input-referred voltage noise density of 0.1 μV/√ Hz. The proposed design employs a two-stage amplification approach, first stage with a fixed 40 dB gain and a variable gain in the second stage, reaching up to 80 dB. An initial filter stage reduces out-of-band noise to improve signal quality. This paper also provides a comprehensive analysis of the selection of low-noise op-amps, passive components, filter topologies, and preamplifier configurations. Extensive simulations confirm the effectiveness of the proposed design in significantly enhancing signal reception for underwater applications.

The direction-of-arrival (DoA) estimation algorithms have a fundamental role in target bearing estimation by sensor array systems. Recently, compressive sensing (CS)-based sparse reconstruction techniques have been investigated for DoA estimation due to their superior performance relative to the conventional DoA estimation methods, for a limited number of measurement snapshots. In many underwater deployment scenarios, the acoustic sensor arrays must perform DoA estimation in the presence of several practical problems such as unknown source number, faulty sensors, low values of the received signal-to-noise ratio (SNR), and access to a limited number of measurement snapshots. In the literature, CS-based DoA estimation has been investigated for the individual occurrence of some of these errors but the estimation under joint occurrence of these errors has not been studied. This work investigates the CS-based robust DoA estimation to account for the joint impact of faulty sensors and low SNR conditions experienced by a uniform linear array of underwater acoustic sensors. Most importantly, the proposed CS-based DoA estimation technique does not require a priori knowledge of the source order, which is replaced in the modified stopping criterion of the reconstruction algorithm by taking into account the faulty sensors and the received SNR. Using Monte Carlo techniques, the DoA estimation performance of the proposed method is comprehensively evaluated in relation to other techniques.

This paper presents simulation of an Ultra Short Baseline (USBL) acoustic positioning system to provide range and bearing estimates of an underwater vehicle. The proposed method estimates the position of an Underwater Vehicle relative to the USBL, using the time of arrival (TOA) of underwater acoustic signals. An acoustic signal emitted from an underwater vehicle, received on an array of hydrophones, and detected using a matched filter or replica cross-correlator. This helps to determine the TOA and to estimate the position of the emitter. The system performance depends on the accurate detection of the signals received on hydrophone array and accurate TOA estimation. In shallow water environment, these received signals become distorted due to additive noise and multipath phenomena. Keeping this in view, pure tone pulse is compared with hyperbolic frequency modulated (HFM) signals resulting in improved TOA measurements with stronger multi-path and noise rejection. The performance of the proposed method is validated based on simulated USBL data with SNR level of 0 dB and bearing accuracy up to 3.22 0 .

The paper focuses on Ethernet based data transfers in real-time processing systems. These systems require fast and reliable data transfers among all the involved resources. In underwater acoustic processing systems, large amount of data is digitized, processed and displayed in real-time. These systems involve hundreds of acoustic sensors producing analog signals that are digitized and generate a large set of data points to be transferred to other processing units in a real-time. This time constraint is important in underwater applications as the processed data displayed to the system operator is critical for undersea operations. Data losses are not acceptable because it may result in incorrect information for the system operator thereby producing a wrong interpretation of the incoming information. Hence an efficient and reliable data transfer mechanism to meet real-time constraint becomes imperative. In this paper we have proposed a Gigabit Ethernet based data transfer methodology using TCP/IP for data transfers among different processing units of an underwater acoustic processing system. Here we have attempted to address some important challenges such as large amount of data to be transferred efficiently in a computer network and the ability to meet processing time (on average) required for successful completion of Gigabit Ethernet based data transfers.

Underwater acoustic classification is a challenging problem because of presence of high background noise and complex sound propagation patterns in the sea environment. Various algorithms proposed in last few years used own privately collected datasets for design and validation. Such data is not publicly available. To conduct research in this field, there is a dire need of publicly available dataset. To bridge this gap, we construct and present an underwater acoustic dataset, named DeepShip, which consists of 47 h and 4 min of real world un-derwater recordings of 265 different ships belong to four classes. The proposed dataset includes recording from throughout the year with different sea states and noise levels. The presented dataset will not only help to evaluate the performance of existing algorithms but it shall also benefit the research community in future. Using the proposed dataset, we also conducted a comprehensive study of various machine learning and deep learning algorithms on six time-frequency based extracted features. In addition, we propose a novel separable convo-lution based autoencoder network for better classification accuracy. Experiments results, which are compared based on classification accuracy, precision, recall, f1-score, and analyzed by using paired sampled statistical t -test, show that the proposed network achieves classification accuracy of 77.53% using CQT feature, which is better than as achieved by other methods.

The paper provides a comparative analysis of two frequency-domain beamforming (FDBF) algorithms in terms of their detection capability along with their implementation aspects i.e. execution times on general purpose processors. The two algorithms namely 2D-FFT beamforming and FFT with array steering vector multiplication (FFT-ASVM) have been compared in the context of linear array processing. MATLAB simulations of these algorithms have been developed and applied on linear array data. Outcome of these simulations show that FFT-ASVM algorithm is better than the 2D-FFT algorithm for accurate detection of both strong and weak targets in the presence of noise. In addition, the paper shows that FDBF algorithms are easy to implement on general purpose processors as their execution times remain well within the real-time processing requirements.

This paper aims to apply adaptive filtering algorithms in sensor array processing for underwater surveillance and detection systems with linear arrays. Adaptive filters are iterative in nature and continuously update the filter taps to minimize the error signal based upon certain criterion. Adaptive filters are used in beamforming to place the main beam in the direction of desired signal and place nulls in the direction of unwanted signals. Adaptive filters are also used in spectral analysis of acoustic signals for background noise cancellation. This paper evaluates the performance of three adaptive filtering algorithms with respect to sensor array design parameters i.e. number of sensors, sensor spacing, number of incident signals and their angular separation. The three adaptive filtering algorithms namely least mean square (LMS), normalized LMS (NLMS), and sample matrix inversion (SMI) have been compared in the context of linear array beamforming. The paper provides detailed discussions on the simulation results of these adaptive algorithms in terms of convergence speed, beamwidth, null depths and maximum side lobe levels. In addition the paper illustrates a practical example of applying adaptive filters in underwater sensor array processing systems for detection of target's acoustic signatures.

Underwater surveillance systems use passive sensor arrays for detection, classification and tracking of both surface and subsurface targets. A passive sensor array in a linear or circular configuration is used to detect and track targets with acoustic radiations into water i.e. frequency lines, machinery and propulsion noises, commonly known as narrowband signatures. Narrowband processing is responsible for analysis and classification of target's acoustic signatures. Broadband processing provides information about target's bearing, amplitude and its movement with time. Target bearing history along with detailed analysis of its narrowband signatures are valuable information for any underwater surveillance system. This paper presents an automated approach for association of narrowband signatures with broadband tracking that relates the frequency lines to the corresponding bearings common to a single target. The proposed methodology requires the system operator to select a target's frequency lines and the association algorithm automatically starts tracking the corresponding bearings on the bearing-time history display. This automated association approach provides several advantages such as resolving crossing targets scenario, facilitating target motion analysis, and improving the response time of system operators.

In this modern era companies all over the world have moved towards COTS (Commercial-off-the-shelf) based systems as they are less expensive, use open architecture standards, and provide rapid deployment. This paper not only addresses the requirements for selecting suitable COTS based systems but also provides details of products and services offered by various COTS vendors in the worldwide marketplace. In addition the paper deals with the development of COTS based processing system for sensor arrays such as an underwater acoustic processor for multi-channel data acquisition and signal processing tasks.

Surveillance and detection displays in underwater acoustic processing systems provide information about acoustic frequency signatures radiated by the surface and submerged targets. Data compression is one of the most important tasks in surveillance and detection displays as there is always limited amount of space for data presentation due to limited sizes of displays and number of pixels. Hence there is always a need for an efficient and accurate algorithm that can reduce the number of data points without disturbing or losing the information present in the data. The data compression algorithm needs to be fast enough to cope up with the real time processing requirements required by the detection displays. In this paper we have proposed an improved data compression algorithm that is compared with the well-known “cubic spline interpolation”. The proposed algorithm has successfully reduced the data points and showed almost no loss in the frequency contents of data and its mapping on the surveillance and detection displays.

Mobile communication systems based on 3G and 4G technologies use adaptive antenna arrays to increase their coverage and channel capacity. Adaptive antenna arrays involve direction finding and beamforming algorithms to localize and track both signals i.e. users and interferers. This paper presents simulation and analysis of three high resolution direction finding algorithms namely MUSIC, Root-MUSIC and ESPRIT. These algorithms provide an estimate about the number of incoming signal sources and their angles of arrival on an antenna array. Simulation results have been used to evaluate the performance of these algorithms by varying the antenna array parameters such as number of mobile users, number of antenna elements, time samples acquired and signal-to-noise ratio.

This paper evaluates general purpose processor architecture for efficient implementation of signal processing algorithms to meet the demands of a real-time system. A multi-channel linear sensor array configuration has been considered for underwater surveillance and detection purpose. The paper defines mathematical formulations for calculating the signal processing requirements in terms of data rate and processor loading i.e. FLOPS (floating point operations) that help in selecting an appropriate processor having suitable capacity to carry out the computationally intensive processing algorithms. The paper presents execution times of signal processing algorithms like beamforming and spectral analysis on modern Intel processors that include Core2Quad and Core i7 processors. The results show a considerable reduction in the execution times on Intel Core i7 processor using multithreaded programming approach. An example of DSP architecture has also been discussed to establish the perception of implementing signal processing algorithms on latest Intel processors.

This paper aims to enhance the target detection capability of passive arrays by improving the broadband processing approach. These passive arrays are used for underwater surveillance and detection of both surface and subsurface targets. Three types of energy detection techniques for broadband processing have been discussed namely CED, PED using Maxima and ED using Euclidean norm. MATLAB simulations of these techniques have been developed and applied on beamformer output of a uniform circular array. The simulation results showed that ED using Euclidean norm is the most suitable energy detection technique for broadband processing in terms of detecting multiple targets accurately in low SNR conditions.

This paper aims to evaluate the detection capability of three different beamforming techniques namely Bartlett, MUSIC, and 2D-FFT. MATLAB simulations of these techniques have been developed and applied on real-time data acquired through a linear sensor array. The simulation results and data analysis showed that 2D-FFT beamforming technique provides the best possible detection as compared to other techniques in terms of resolving the incoming targets accurately.

This paper presents simulation and analysis of adaptive signal processing techniques for interference cancelling i.e. noise cancellation and beamforming using antenna arrays. Adaptive processing involves an iterative filtering algorithm that continuously computes the filter coefficients to minimize the error signal based upon certain criterion. First the paper presents analysis of a noise cancellation simulation for noise reduction in spectral analysis of audio signals. Second the paper presents analysis of an adaptive beamforming simulation for exploiting the spatial separation between the desired signal and other interfering signals. This has been done by computing the weights for the antenna elements iteratively to obtain maximum reception in a specified direction. Results of these simulations provide improved performance in both the cases and can be used for analysis, design and optimization of linear antenna arrays.