The mathematical modeling of the acoustic wave propagation in seawater is the basis for realizing goals such as, underwater communication, seabed mapping, advanced fishing, oil and gas exploration, marine meteorology, positioning and explore the unknown targets within the water. However, due to the existence of various physical phenomena in the water environment and the various conditions governing the sea environment, such as, noise, sea state, depth, substrate, temperature and salinity, various methods have been proposed for the propagation of acoustic wave, which results in a high computational complexity in the numerical solution of acoustic wave propagation. Since the high speed of computing is a key factor in most real-time applications, in this study, a parallel algorithm for the implementation of computerized underwater acoustic wave propagation is presented, using the Gaussian Beamtracing method. In order to compare the performance of the proposed parallel algorithm with respect to the serial mode, the results are reported in figure and table for several different scenarios. The simulation results indicate a much higher rate of parallel algorithm than the serial mode despite its very precise accuracy.

Packet classification is a fundamental process in network processors. In this process, input packets are classified into distinct set of flows via matching against a set of filters. Software implementation of packet classification algorithms, though having lower cost and more scalability as compared with hardware implementations, are slower. In this paper, we use parallel processing capabilities of the graphical processors to accelerate Hierarchical-Trie packet classification algorithm and propose different scenarios based on the architecture of their global and shared memories. Results of implementing these scenarios, conforming computed time and memory complexities, show that the performance of the scenarios that divide the filter set into sub-trees, equal to/ smaller than the shared memory and copy them to it, is lower than that of a scenario which keeps the total data structure in the global memory. The performance of these scenarios increases by decreasing the number of sub-trees and duplicated filters. Moreover, a scenario that can keep hierarchical tree and corresponding filters in shared memory, without any partitioning, is the best scenario. The experimental results show that, on a same GPU, this scenario attains a throughput of approximately 2.1 times compared to the existing methods.

In this paper, an alternative approach in operational modal analysis is presented, utilizing image processing technique and transmissibility functions. Imaging sensors do not impose additional mass on the structure due to their non-contact nature, while transmissibility functions, independent of excitation type, can directly extract mode shapes. The innovation of this research lies in combining these two techniques to record dynamic responses and identify modal properties. To capture the temporal response history from video signals, the block-matching method with sub-pixel accuracy was employed. Validation was conducted by recording the response of the tip of a cantilevered steel beam subjected to impact excitation, using a high-speed camera and a laser vibrometer, simultaneously. The RMSE plots in the time domain and the PSD in the frequency domain indicate high accuracy of this method. Using this approach, the displacement time histories of various points on the structure were extracted from the video signals, and the modal properties, including natural frequencies, damping ratios, and mode shapes, were identified using the transmissibility matrix method. The results obtained from the proposed method were compared with the stochastic subspace identification (SSI) method and analytical solutions. The findings reveal the accuracy of the modal identification approach introduced in this article. The highest relative error in estimating the natural frequencies of the first and second modes, compared to the values from the laser method, are 0.19% and 0.13%, respectively, and in comparison to the analytical values, they are 0.34% and 1.5%, respectively.

There are different variants of Particle Swarm Optimization (PSO) algorithm such as Adaptive Particle Swarm Optimization (APSO) and Particle Swarm Optimization with an Aging Leader and Challengers (ALC-PSO). These algorithms improve the performance of PSO in terms of finding the best solution and accelerating the convergence speed. However, these algorithms are computationally intensive. The goal of this paper is high performance implementations of Traditional PSO (TPSO), APSO and ALCPSO using CUDA technology. We have implemented these three algorithms on both central processing unit (CPU) and graphics processing unit (GPU) in order to analyze and improve their performance and reduce their computational times. We have achieved speedups up to 14. 5x, 31x, and 152x, for GPUTPSO, GPU-ALCPSO, and GPU-APSO, respectively. In addition, different number of threads has been chosen in order to find an appropriate number of threads per block for both APSO and ALC-PSO algorithms. Our experimental results show that the best choice for number of threads per block depends on the number of existing variables and constants in each algorithm and the number of registers per multiprocessor.

In the CMOS circuit power dissipation is a major concern for VLSI functional units. With shrinking feature size, increased frequency and power dissipation on the data bus have become the most important factor compared to other parts of the functional units. One of the most important functional units in any processor is the Multiply-Accumulator unit (MAC). The current work focuses on the development of MAC unit bus encoders as well as the identification of an improved architecture for image processing applications. To reduce the power consumption in these functional units, two bus encoding architectures were developed by encoding data before it was sent on the data buses. One is MSB reference encoding, and another is Fourth and Fifth bit ANDing (FFA) without the need for an extra bus line with fewer transitions by using gray codes. The comparison of the proposed encoding architectures with the existing encoding architectures from the literature revealed an 8% to 36% significant improvement in power dissipation. The simulation was done with Xilinx ISE, and the Cadence RTL Compiler tool was utilized for the synthesis, which was done with the 180nm technology library. And also, the image filtering is analyzed using MATLAB.

We present an algorithm for NVIDIA CUDA platform based on Smith-Waterman algorithm for sequence alignment problem. CUDA is a new programming language which is very similar to the standard C language, with some extensions. By using the Smith-Waterman algorithm which is used to find similarity between two sequences by making a scoring matrix; this application tries to find the similarity between a sequence which called query sequence and sequences in a database file. In this program, each thread is used for calculating one scoring matrix between the query sequence and one of the sequences in database. The algorithm utilizes small but fast shared memory which is inside the GPU for holding four intermediate columns in each cycle for each thread. If the database is large enough, it can keep the GPU in full working order and results in better speedup in comparison to CPU. To demonstrate the performance, we used a high-end CPU, Intel Core 2 Due E6600 and a mid-end GPU, GeForce 8600GT and saw the speedup in about twenty times more than CPU.

In traditional speech processing, feature extraction and classification were conducted as separate steps. The advent of deep neural networks has enabled methods that simultaneously model the relationship between acoustic and phonetic characteristics of speech while classifying it directly from the raw waveform. The first convolutional layer in these networks acts as a filter bank. To enhance interpretability and reduce the number of parameters, researchers have explored the use of parametric filters, with the SincNet architecture being a notable advancement. In SincNet's initial convolutional layer, rectangular bandpass filters are learned instead of fully trainable filters. This approach allows for modeling with fewer parameters, thereby improving the network's convergence speed and accuracy. Analyzing the learned filter bank also provides valuable insights into the model's performance. The reduction in parameters, along with increased accuracy and interpretability, has led to the adoption of various parametric filters and deep architectures across diverse speech processing applications. This paper introduces different types of parametric filters and discusses their integration into various deep architectures. Additionally, it examines the specific applications in speech processing where these filters have proven effective.

The current manuscript presents the validation of Smoothed Particle Hydrodynamics (SPH) techniques for wave generation by underwater explosion, utilizing the so-called Duals Physics numerical model. This numerical method is used to analyze generated waves which are initiated by man-made or natural explosions below free surface level of sea. In spite of the modeling limitations (e.g. absence of open boundary conditions), reasonable agreement is accomplished with predictions of the existing formula as well as experimental results. This proved that SPH techniques such as incorporated in Duals Physics are becoming a suitable alternative to existing classical approaches to this particular water waves problem. It is also provided an inherently more accurate computational for the prediction of wave characteristics generated by underwater explosions.