FFT Network Research Articles

An Algorithm for Crack Detection, Segmentation, and Fractal Dimension Estimation in Low-Light Environments by Fusing FFT and Convolutional Neural Network

The segmentation of crack detection and severity assessment in low-light environments presents a formidable challenge. To address this, we propose a novel dual encoder structure, denoted as DSD-Net, which integrates fast Fourier transform with a convolutional neural network. In this framework, we incorporate an information extraction module and an attention feature fusion module to effectively capture contextual global information and extract pertinent local features. Furthermore, we introduce a fractal dimension estimation method into the network, seamlessly integrated as an end-to-end task, augmenting the proficiency of professionals in detecting crack pathology within low-light settings. Subsequently, we curate a specialized dataset comprising instances of crack pathology in low-light conditions to facilitate the training and evaluation of the DSD-Net algorithm. Comparative experimentation attests to the commendable performance of DSD-Net in low-light environments, exhibiting superlative precision (88.5%), recall (85.3%), and F1 score (86.9%) in the detection task. Notably, DSD-Net exhibits a diminutive Model Size (35.3 MB) and elevated Frame Per Second (80.4 f/s), thereby endowing it with the potential to be seamlessly integrated into edge detection devices, thus amplifying its practical utility.

A Measurement Method for Wide-frequency Harmonic Signal in Power Grid Based on Window Interpolation FFT and RLS-Adaline Neural Network

With the increase of non-linear loads in power grid and the development of power electronics, the growing serious harmonic pollution has put forward higher requirements for the accuracy and real-time performance of the measurement method. For this problem, this paper proposes a wide-frequency harmonic signal measurement method based on window interpolation FFT and adaptive neural network. After the window function is selected, the frequency of harmonic is estimated by double-spectrum-line interpolation algorithm based on three-term third derivative Nuttall window, and then the smoothed frequencies are input into RLS-Adaline neural network to estimate the amplitude and phase of each harmonic. The simulation results show that the proposed method has higher measurement accuracy and stability under few data conditions and low signal-to-noise ratios than conventional window interpolation algorithms. High accuracy measurement of harmonics with short time window in the range of 0∼2.5kHz is effectively realized.

Detection of EEG Signal Post-Stroke Using FFT and Convolutional Neural Network

Detection of EEG Signal Post-Stroke Using FFT and Convolutional Neural Network

Detection of EEG Signal Post-Stroke Using FFT and Convolutional Neural Network

Detection of EEG Signal Post-Stroke Using FFT and Convolutional Neural Network

Brain tumour detection based on FFT, curve analysis, k-space and neural network classifier

Magnetic Resonance Imaging (MRI) has become an efficient instrument for clinical diagnoses in recent years. In this paper, an efficient MRI image segmentation for tumour detection is proposed using FFT, curve analysis and k-space. Input MRI image is pre-processed and segmentation is carried out using EM. Subsequently, features are extracted by using FFT, curve analysis and k-space. Finally, neural network classifier is employed to diagnose brain tumour. The MRI image dataset used to evaluate the proposed image technique is taken from the publicly available sources. The evaluation metrics used to evaluate the proposed technique consists of sensitivity, specificity and accuracy. Overall, the proposed technique could achieve sensitivity, specificity and accuracy values of 0.80, 0.81 and 0.805 respectively. The comparative analysis is also made comparing with other existing techniques. From the results, it can be seen that our proposed technique performed well and obtained better evaluation metrics than the existing methods.

Brain tumour detection based on FFT, curve analysis, k-space and neural network classifier

Magnetic Resonance Imaging (MRI) has become an efficient instrument for clinical diagnoses in recent years. In this paper, an efficient MRI image segmentation for tumour detection is proposed using FFT, curve analysis and k-space. Input MRI image is pre-processed and segmentation is carried out using EM. Subsequently, features are extracted by using FFT, curve analysis and k-space. Finally, neural network classifier is employed to diagnose brain tumour. The MRI image dataset used to evaluate the proposed image technique is taken from the publicly available sources. The evaluation metrics used to evaluate the proposed technique consists of sensitivity, specificity and accuracy. Overall, the proposed technique could achieve sensitivity, specificity and accuracy values of 0.80, 0.81 and 0.805 respectively. The comparative analysis is also made comparing with other existing techniques. From the results, it can be seen that our proposed technique performed well and obtained better evaluation metrics than the existing methods.

Easily testable and fault-tolerant FFT butterfly networks

With the advent of deep submicron very large scale integration technology, the integration of a large fast-Fourier-transform (FFT) network into a single chip is becoming possible. However, a practical FFT chip is normally very big, so effective testing and fault-tolerance techniques usually are required. In this paper, we first propose a C-testable FFT network design. Only 20 test patterns are required to cover all combinational single-cell faults and interconnect stuck-at and break faults for the FFT network, regardless of its size. A spare-row based fault-tolerant FFT network design is subsequently proposed. Compared with previous works, our approach shows higher reliability and lower hardware overhead, and only three bit-level cell types are needed for repairing a faulty row in the multiply-subtract-add module. Also, special cell design is not required to implement the reconfiguration scheme. The hardware overhead for the testable design is low-about 4% for 16-bit numbers, regardless of the FFT network size.

The Black-Box Model for Cryptographic Primitives

We introduce the black-box model for cryptographic primitives. In this model cryptographic primitives are given by a computation graph, where the computation boxes sitting on the vertices of the graph act as random oracles. We formalize and study a family of generic attacks which generalize exhaustive search and the birthday paradox. We establish complexity lower bounds for these attacks and we apply it to compression functions based on the FFT network.

Optimal book embeddings of the FFT, benes, and barrel shifter networks

This paper gives 3-page book embeddings of three important interconnection networks: the FFT network, the Benes rearrangeable permutation network, and the barrel shifter network. Since all three networks are eventually nonplanar, they require three pages and the present embeddings are optimal. Also, the embeddings have pages of comparable widths.

Optimal three-dimensional VLSI layouts

In this note we propose VLSI layouts for the cube-connectedcycles (CCC) in a three-dimensional medium. The first type of layout is for the all-active-volume mode, where computing modules are placed anywhere within the solid chip; the second type is for the one-active-layer mode, where computing modules are placed on an external face of the solid chip. In both cases we present volume × time optimal realizations. In addition, by opening the cycles of the CCC, one obtains the well-known FFT network. Both realizations use minimum volume; an all-active-volume realization with the same volume performance was earlier proposed by Rosenberg, while the present one-active-layer realization is the only optimal one that is known.