Nothing Special   »   [go: up one dir, main page]
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
5.3
2.5
Journals
Applied Sciences
Special Issues
Advances in Digital Signal Processing: New Applications and Efficient Implementations
applsci-logo
Submit to Applied Sciences Review for Applied Sciences Propose a Special Issue

Journal Menu

► Journal Menu

Journal Browser

► Journal Browser
share announcement

Advances in Digital Signal Processing: New Applications and Efficient Implementations

A special issue of Applied Sciences (ISSN 2076-3417). This special issue belongs to the section "Computing and Artificial Intelligence".

Deadline for manuscript submissions: closed (20 December 2024) | Viewed by 4187

Special Issue Editor

Prof. Dr. Doru Florin Chiper

E-Mail Website
Guest Editor
Faculty of Electronics, Telecommunications and Information Technology, Gheorghe Asachi Technical University of Iaşi, 700506 Iaşi, Romania
Interests: digital signal processing (DSP); adaptive signal processing; blind equalization/identification; fast computational algorithms; parallel and VLSI algorithms and architectures for communications and DSP; high-level DSP design
Special Issues, Collections and Topics in MDPI journals

Special Issue Information

Dear Colleagues, 

In the new era of digital revolution, new advanced DSP applications have appeared. The advances in modern DSP applications, such as multimedia, big data, IoT, etc., have increased the importance of the optimization and efficient implementation of DSP algorithms and architectures, both for a VLSI or a software  VLSI implementation. We can say that they represent an essential part of the research in such modern applications.

For real time implementations of such modern DSP applications, an efficient optimization of such algorithms and architectures for an efficient VLSI implementation are often a critical and challenging issue. For example, real-time multimedia applications have increasingly greater performance requirements due to data processing and transmission of huge data volumes at high speeds, with resource constraints specific to portable devices.

This Special Issue focuses on papers that demonstrate how these design challenges can be overcome using innovative solutions.

Topics of interest for this Special Issue include but are not limited to:

  • VLSI signal processing;
  • Signal processing methods for an efficient implementation;
  • Optimization of the VLSI implementation of multimedia blocks;
  • Low-power circuits and systems for DSP applications;
  • Efficient adaptive/learning algorithms (low complexity/fast versions, optimized parameters, etc.);
  • Tensor-based signal processing (efficient decomposition methods, low-rank approximations, etc.);
  • Sparsity-aware algorithms.

Prof. Dr. Doru Florin Chiper
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Applied Sciences is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • VLSI signal processing
  • signal processing methods
  • efficient implementation
  • multimedia blocks
  • low power circuits and systems
  • efficient adaptive algorithms
  • learning algorithms
  • efficient decomposition methods
  • low-rank approximations
  • sparsity exploitation

Benefits of Publishing in a Special Issue

  • Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
  • Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
  • Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
  • External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
  • e-Book format: Special Issues with more than 10 articles can be published as dedicated e-books, ensuring wide and rapid dissemination.

Further information on MDPI's Special Issue polices can be found here.

Published Papers (4 papers)

Download All Papers
Order results
Result details
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:

Research

22 pages, 3664 KiB  
Article
Tensor Network Methods for Hyperparameter Optimization and Compression of Convolutional Neural Networks
by A. Naumov, A. Melnikov, M. Perelshtein, Ar. Melnikov, V. Abronin and F. Oksanichenko
Appl. Sci. 2025, 15(4), 1852; https://doi.org/10.3390/app15041852 - 11 Feb 2025
Viewed by 423
Abstract
Neural networks have become a cornerstone of computer vision applications, with tasks ranging from image classification to object detection. However, challenges such as hyperparameter optimization (HPO) and model compression remain critical for improving performance and deploying models on resource-constrained devices. In this work, [...] Read more.
Neural networks have become a cornerstone of computer vision applications, with tasks ranging from image classification to object detection. However, challenges such as hyperparameter optimization (HPO) and model compression remain critical for improving performance and deploying models on resource-constrained devices. In this work, we address these challenges using Tensor Network-based methods. For HPO, we propose and evaluate the TetraOpt algorithm against various optimization algorithms. These evaluations were conducted on subsets of the NATS-Bench dataset, including CIFAR-10, CIFAR-100, and ImageNet subsets. TetraOpt consistently demonstrated superior performance, effectively exploring the global optimization space and identifying configurations with higher accuracies. For model compression, we introduce a novel iterative method that combines CP, SVD, and Tucker tensor decompositions. Applied to ResNet-18 and ResNet-152, we evaluated our method on the CIFAR-10 and Tiny ImageNet datasets. Our method achieved compression ratios of up to 14.5× for ResNet-18 and 2.5× for ResNet-152. Additionally, the inference time for processing an image on a CPU remained largely unaffected, demonstrating the practicality of the method. Full article
(This article belongs to the Special Issue Advances in Digital Signal Processing: New Applications and Efficient Implementations)
►▼ Show Figures

Figure 1

Figure 1
<p>General schema of neural network hyperparameter optimization problem.</p> Full article ">Figure 2
<p>Tensor train (TT) and grid search (GS): expected runtime in maximum objective function evaluations vs. growth of problem dimension <span class="html-italic">d</span>.</p> Full article ">Figure 3
<p>Pipeline of compressing pre-trained model using our proposed method.</p> Full article ">Figure 4
<p>Maximum accuracy found at each iteration for different optimization algorithms across six experiments using NATS-Bench subsets. (<b>a</b>–<b>f</b>) corresponds to a specific dataset and training duration.</p> Full article ">Figure 5
<p>Maximum accuracy against algorithm accumulated runtime for TetraOpt and other optimization algorithms across six NATS-Bench experiments. (<b>a</b>–<b>f</b>) corresponds to specific dataset and training duration.</p> Full article ">
16 pages, 708 KiB  
Article
Forward/Backward Decomposition for Dispersive Wave Propagation Measurements
by Nicholas A. Corbin and Pablo Tarazaga
Appl. Sci. 2025, 15(2), 858; https://doi.org/10.3390/app15020858 - 16 Jan 2025
Viewed by 595
Abstract
Two complicating features commonly found in wave propagation applications include dispersion, i.e., frequency-dependent propagation velocity, and reflections, which introduce coherent noise. In this work, we present a signal processing technique which can be applied in a variety of applications to decompose signals into [...] Read more.
Two complicating features commonly found in wave propagation applications include dispersion, i.e., frequency-dependent propagation velocity, and reflections, which introduce coherent noise. In this work, we present a signal processing technique which can be applied in a variety of applications to decompose signals into their forward- and backward-propagating components. The theory is presented, along with algorithmic implementation and experimental validation on a Timoshenko beam. The implications and potential utility of the method are briefly discussed. Full article
(This article belongs to the Special Issue Advances in Digital Signal Processing: New Applications and Efficient Implementations)
►▼ Show Figures

Figure 1

Figure 1
<p>Diagram showing sensors along waveguide. The orange rectangle represents an actuator, and the black circles represent the sensors.</p> Full article ">Figure 2
<p>Experimental setup showing the view of the test article from the laser Doppler vibrometer. The scan region is labeled, along with stars highlighting the first and last scan points. The close-up pictures in call-outs depict, from left to right, the MFC actuator, two of the four strain gages, and the tensioning mechanism, which is present at both ends of the beam.</p> Full article ">Figure 3
<p>Time and frequency domain representations of the excitation signal.</p> Full article ">Figure 4
<p>Beam responses to the two-cycle toneburst with measurements at locations 1 and 40. (<b>a</b>) 1500 Hz center frequency toneburst. (<b>b</b>) 500 Hz center frequency toneburst.</p> Full article ">Figure 5
<p>Decomposed high-frequency (1500 Hz) beam responses showing the forward- and backward-propagating wave components for sensor 1. The relative error between the reconstructed signal and the measured signal is <math display="inline"><semantics> <mrow> <mn>35.37</mn> <mo>%</mo> </mrow> </semantics></math>.</p> Full article ">Figure 6
<p>Decomposed high-frequency (1500 Hz) beam responses showing the forward- and backward-propagating wave components for sensor 40. The relative error between the reconstructed signal and the measured signal is <math display="inline"><semantics> <mrow> <mn>21.27</mn> <mo>%</mo> </mrow> </semantics></math>.</p> Full article ">Figure 7
<p>Decomposed low-frequency (500 Hz) beam responses showing the forward- and backward-propagating wave components for sensor 1. The relative error between the reconstructed signal and the measured signal is <math display="inline"><semantics> <mrow> <mn>34.68</mn> <mo>%</mo> </mrow> </semantics></math>.</p> Full article ">Figure 8
<p>Decomposed low-frequency (500 Hz) beam responses showing the forward- and backward-propagating wave components for sensor 40. The relative error between the reconstructed signal and the measured signal is <math display="inline"><semantics> <mrow> <mn>115.41</mn> <mo>%</mo> </mrow> </semantics></math>.</p> Full article ">Figure 9
<p>Relative errors for the signal reconstructions at all 40 scan points. The average relative error for the high-frequency case (1500 Hz) is <math display="inline"><semantics> <mrow> <mn>17.37</mn> <mo>%</mo> </mrow> </semantics></math>, and the average relative error for the low-frequency case (500 Hz) is <math display="inline"><semantics> <mrow> <mn>17.71</mn> <mo>%</mo> </mrow> </semantics></math>.</p> Full article ">Figure 10
<p>Decomposed low-frequency (500 Hz) beam responses showing the forward- and backward-propagating wave components for sensor 39. The relative error between the reconstructed signal and the measured signal is <math display="inline"><semantics> <mrow> <mn>16.75</mn> <mo>%</mo> </mrow> </semantics></math>.</p> Full article ">
26 pages, 1137 KiB  
Article
A Novel Low-Complexity and Parallel Algorithm for DCT IV Transform and Its GPU Implementation
by Doru Florin Chiper and Dan Marius Dobrea
Appl. Sci. 2024, 14(17), 7491; https://doi.org/10.3390/app14177491 - 24 Aug 2024
Cited by 1 | Viewed by 1273
Abstract
This study proposes a novel factorization method for the DCT IV algorithm that allows for breaking it into four or eight sections that can be run in parallel. Moreover, the arithmetic complexity has been significantly reduced. Based on the proposed new algorithm for [...] Read more.
This study proposes a novel factorization method for the DCT IV algorithm that allows for breaking it into four or eight sections that can be run in parallel. Moreover, the arithmetic complexity has been significantly reduced. Based on the proposed new algorithm for DCT IV, the speed performance has been improved substantially. The performance of this algorithm was verified using two different GPU systems produced by the NVIDIA company. The experimental results show that the novel proposed DCT algorithm achieves an impressive reduction in the total processing time. The proposed method is very efficient, improving the algorithm speed by more than 4-times—that was expected by segmenting the DCT algorithm into four sections running in parallel. The speed improvements are about five-times higher—at least 5.41 on Jetson AGX Xavier, and 10.11 on Jetson Orin Nano—if we compare with the classical implementation (based on a sequential approach) of DCT IV. Using a parallel formulation with eight sections running in parallel, the improvement in speed performance is even higher, at least 8.08-times on Jetson AGX Xavier and 11.81-times on Jetson Orin Nano. Full article
(This article belongs to the Special Issue Advances in Digital Signal Processing: New Applications and Efficient Implementations)
►▼ Show Figures

Figure 1

Figure 1
<p>A flowchart of the parallel execution of the proposed algorithm.</p> Full article ">Figure 2
<p>A detailed memory workload analysis of the GPU through NVIDIA Nsight Compute tool.</p> Full article ">Figure 3
<p>The GPU throughput for a SM unit (magenta—classical implementation, green—parallel implementation with four sections, blue—parallel implementation with eight sections)—GPU at 114.75 MHz.</p> Full article ">Figure 4
<p>The GPU throughput for a SM unit (magenta—classical implementation, green—parallel implementation with four sections, blue—parallel implementation with eight sections)—GPU at 1.377 GHz.</p> Full article ">Figure 5
<p>Speedup factor variation on different GPU working frequencies.</p> Full article ">
17 pages, 1050 KiB  
Article
Kalman Filter Using a Third-Order Tensorial Decomposition of the Impulse Response
by Laura-Maria Dogariu, Constantin Paleologu, Jacob Benesty and Felix Albu
Appl. Sci. 2024, 14(11), 4507; https://doi.org/10.3390/app14114507 - 24 May 2024
Viewed by 942
Abstract
For system identification problems associated with long-length impulse responses, the recently developed decomposition-based technique that relies on a third-order tensor (TOT) framework represents a reliable choice. It is based on a combination of three shorter filters, which merge their estimates in tandem with [...] Read more.
For system identification problems associated with long-length impulse responses, the recently developed decomposition-based technique that relies on a third-order tensor (TOT) framework represents a reliable choice. It is based on a combination of three shorter filters, which merge their estimates in tandem with the Kronecker product. In this way, the global impulse response is modeled in a more efficient manner, with a significantly reduced parameter space (i.e., fewer coefficients). In this paper, we further develop a Kalman filter based on the TOT decomposition method. As compared to the recently designed recursive least-squares (RLS) counterpart, the proposed Kalman filter achieves superior performance in terms of the main criteria (e.g., tracking and accuracy). In addition, it significantly outperforms the conventional Kalman filter, while also having a lower computational complexity. Simulation results obtained in the context of echo cancellation support the theoretical framework and the related advantages. Full article
(This article belongs to the Special Issue Advances in Digital Signal Processing: New Applications and Efficient Implementations)
►▼ Show Figures

Figure 1

Figure 1
<p>Normalized misalignment of the KF-TOT using different values of <span class="html-italic">P</span> and null uncertainty parameters. The input signal is an AR(1) process and <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math> dB.</p> Full article ">Figure 2
<p>Normalized misalignment of the KF-TOT using <math display="inline"><semantics> <mrow> <mi>P</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math> and different constant values of the uncertainty parameters. The input signal is an AR(1) process, <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math> dB, and the echo path changes after 15 s.</p> Full article ">Figure 3
<p>Normalized misalignment of the conventional KF and KF-TOT (with <math display="inline"><semantics> <mrow> <mi>P</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math>). The input signal is an AR(1) process, <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math> dB, and the echo path changes after 3.75 s.</p> Full article ">Figure 4
<p>Normalized misalignment of the conventional KF and KF-TOT (with <math display="inline"><semantics> <mrow> <mi>P</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math>). The input signal is an AR(1) process, the echo path changes after 3.75 s: (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>10</mn> </mrow> </semantics></math> dB and (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>0</mn> </mrow> </semantics></math> dB.</p> Full article ">Figure 5
<p>Normalized misalignment of the KF-NKP [<a href="#B27-applsci-14-04507" class="html-bibr">27</a>] and KF-TOT, using different values of <math display="inline"><semantics> <msup> <mi>P</mi> <mo>*</mo> </msup> </semantics></math> and <span class="html-italic">P</span>, respectively. The input signal is a speech sequence and <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math> dB.</p> Full article ">Figure 6
<p>Normalized misalignment of the KF-NKP [<a href="#B27-applsci-14-04507" class="html-bibr">27</a>] (with <math display="inline"><semantics> <mrow> <msup> <mi>P</mi> <mo>*</mo> </msup> <mo>=</mo> <mn>3</mn> </mrow> </semantics></math>) and KF-TOT (using different values of <span class="html-italic">P</span>). The input signal is a speech sequence, the echo path changes after 3.75 s: (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>10</mn> </mrow> </semantics></math> dB, <math display="inline"><semantics> <mrow> <mi>P</mi> <mo>=</mo> <mn>3</mn> </mrow> </semantics></math>; and (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>0</mn> </mrow> </semantics></math> dB, <math display="inline"><semantics> <mrow> <mi>P</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math>.</p> Full article ">Figure 7
<p>Normalized misalignment of the RLS-TOT [<a href="#B37-applsci-14-04507" class="html-bibr">37</a>] (using different forgetting factors) and KF-TOT, with <math display="inline"><semantics> <mrow> <mi>P</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math>. The input signal is a speech sequence, <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math> dB, and the echo path changes after 3.75 s.</p> Full article ">Figure 8
<p>Normalized misalignment of the RLS-TOT [<a href="#B37-applsci-14-04507" class="html-bibr">37</a>] (using different forgetting factors) and KF-TOT, with <math display="inline"><semantics> <mrow> <mi>P</mi> <mo>=</mo> <mn>2</mn> </mrow> </semantics></math>. The input signal is a speech sequence, while (<b>a</b>) <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>10</mn> </mrow> </semantics></math> dB and (<b>b</b>) <math display="inline"><semantics> <mrow> <mi>SNR</mi> <mo>=</mo> <mn>0</mn> </mrow> </semantics></math> dB.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-4
Back to TopTop