https://doi.org/10.5573/IEIESPC.2023.12.4.283

(Liyan Shi)

To improve the efficiency of public security image inspection and monitoring during epidemics and sudden events, an image recognition algorithm based on LabVIEW is proposed. By using fusion detection for human appearance parameters and dynamic gaits, a block feature-matching model for suspicious dynamic information in security images is constructed. Considering the environmental factors, security images with suspicious background information taken during epidemics and sudden events are decomposed, and edge contour detection is constructed. Based on suspicious edge detection results, the spatial structure of the images is extracted, risk difference characteristics of human body shapes in the images are captured, and a variety of local and global feature analysis models for public-place security images based on epidemics and sudden events are constructed. Through cross-regional block fusion and by using gray edge information decomposition, the suspicious features extracted from security images captured during epidemics and sudden events are realized, and visual simulations are carried out based on LabVIEW. The results show that the proposed method has high output recognition, good fusion performance on safety factors, an improved recognition ability with security images, and good performance in feature detection from security images.

https://doi.org/10.5573/IEIESPC.2023.12.4.290

(Yaping Zhao)

The study of fault identification of vibration signals from rotating machinery is essential for enhancing industrial production safety. A method combining a capsule network and frequency-slicing wavelet transform is proposed to improve the fault identification accuracy, considering the problem that the original vibration signal of rotating machinery carries multiple noises. The capsule network learning model was also optimized using a dynamic weighting method based on the channel attention mechanism, considering the variable operating conditions of rotating machinery. The dynamic weighting algorithm based on the channel attention mechanism used in the study achieved the highest fault recognition rates, with 99.65%, 99.25%, and 99.90% on sensor 1, sensor 2, and feature fusion data, respectively. Hence, the proposed model for fault identification in rotating machinery vibration signals is superior to other models.

https://doi.org/10.5573/IEIESPC.2023.12.4.300

(Shucheng Wei) ; (Hui Wang)

The image information captured by a sensor in a network environment shows diversity and uncertainty, and it is difficult to achieve good data information processing and fusion because of the difference in characteristics of multiple images collected without time and space, which has caused considerable interference to the authenticity of the image. A multi-sensor-based information fusion analysis algorithm for moving images is proposed to improve the visual effects of image fusion and the signal-to-noise ratio and information entropy. The convolutional neural network (CNN) is used to extract the features of moving images. The mixed function control curve method generates the time series of moving images. According to the time series of the moving image obtained, the moving image is decomposed by a wavelet. A color space model (CSM) is established, and image fusion and optimization are realized using the multi-sensor fusion and multi-objective particle swarm optimization (PSO) algorithm. The proposed method significantly improved the SNR value and information entropy and reduced the standard mean square error. In addition, it had a remarkable image fusion visual effect.

https://doi.org/10.5573/IEIESPC.2023.12.4.312

(Zhaosheng Chen) ; (Na Chen)

In order to improve teaching children how to play football, combining the Vector of Locally Aggregated Descriptors (VLAD) model and a deep belief network (DBN) into a V-DBN is proposed based on 3D bone recognition that recognizes football actions. We use the contrast method to reduce the dimensionality of action features, and we complete the action recognition through analysis of key parameters. After experimental testing with the MSRAction3D data set, Grassmann manifold and graph-based action classification and recognition reach accuracies of 79.2% and 93.4%, respectively, after 100 iterations of training, but the V-DBN reaches 98.6%. In the UTKinect-Action database test, the average recognition rates of Grassmann manifold and graph-based action classification and recognition are 88.38% and 91.31% accurate, respectively, while the VLAD is 93.96% accurate, showing the best overall performance. However, the effect in single-action recognition is only average. Using the LSTM optimization model on results from infant football action recognition, the average accuracy rate of LSTM+V-DBN is 0.981 compared to the V-DBN at 0.892. Clearly, the optimized LSTM+V-DBN model performs better in toddler action recognition. This research provides important reference value to the application of human action recognition technology in children’s football education.

https://doi.org/10.5573/IEIESPC.2023.12.4.323

(Xuran Ni)

In machine translation, the processing of tenses is a crucial element. This paper briefly introduces neural machine translation (NMT) and analyzes the NMT model based on long short-term memory (LSTM) and bidirectional-LSTM (Bi-LSTM). A neural network method based on LSTM was designed to recognize verb tenses in Chinese by collating bilingual corpora. The method was combined with NMT to recognize and translate different tenses. Experiments were conducted on the dataset. The results suggested that Bi-LSTM had a higher accuracy in tense recognition than LSTM, with an average accuracy of 89.89%, 7.8% higher than LSTM. A comparison of the baseline showed that the NMT model based on Bi-LSTM had the highest bilingual evaluation understudy (BLEU) score. The BLEU score of Bi-LSTM combined with tense recognition was improved by 6.9. The experimental results demonstrated the effectiveness of Bi-LSTM combined with tense recognition in recognizing and translating different tenses in Chinese–English translations. This method can be applied in practice.

https://doi.org/10.5573/IEIESPC.2023.12.4.329

(Dasaem Jeong)

The Ai Music Generation Challenge is a competition that evaluates music generation systems. The participants submit a model capable of generating a specific genre of traditional music, which human experts then evaluate. The 2022 competition aimed to generate the most plausible reel, a type of Irish traditional dance music. This paper presents our submitted model, VirtuosoTune, which utilizes musically-structured encodings and the hierarchical structure of gated recurrent units. One of the tunes generated by the model was awarded the first prize, achieving perfect scores from all four judges.

https://doi.org/10.5573/IEIESPC.2023.12.4.334

(Sungmin Woo) ; (Rayun Boo)

In this study, we present a novel deep-fusion architecture aimed at enhancing color prediction for a light source by leveraging both spectral sensor and image. Recently, the advancements in sensor technology have led to the emergence of spectral sensors capable of capturing multi-wavelength information. While images obtained from camera sensors provide abundant pixel-level details, the information pertaining to wavelength is limited to the conventional RGB channels. Therefore, our approach introduces a deep learning framework that simultaneously integrates images, spectral information, and their combined representation. By comparing the performance of various architectures when predicting the light source color, we identify the most suitable structure and verify the helpfulness of multi-wavelength information without spatial information. To achieve this, we construct a dataset that captures both spectrum and image data concurrently, which are then utilized for conducting experiments.

https://doi.org/10.5573/IEIESPC.2023.12.4.341

(Haiyang Cao)

Automatic scoring of English interpretations is a technology that evaluates content accuracy and pronunciation. The technology is of great practical significance for English interpreting tests. This paper uses DNN-HMM speech recognition to evaluate the interpretations of English translators. Post-probabilistic features of phonemes to the speech vector are decoded in the sub-recognition framework. The correlation between posterior probability and manual scoring is analyzed. The calculation of a posterior probability feature depends highly on recognition performance. To improve recognition performance, a Recurrent Neural Network (RNN) is used to construct speech recognition models. Through comparison experiments, results show that the correlation between the RNN model’s scoring method and manual scoring is 0.781. The recognition accuracy of the RNN model is 0.923 on the training set and 0.915 on the test set. The posterior probability error of the RNN model is 0.028 for the training set and 0.033 for the test set. Therefore, the RNN structure has more reasonable and scientific scoring performance in automatic English interpretation.

https://doi.org/10.5573/IEIESPC.2023.12.4.350

(Liton Chandra Paul) ; (Akash Majumder) ; (Tithi Rani) ; (Md. Najmul Hossain) ; (Md. Abdur Rahim) ; (Jungpil Shin) ; (Keun Soo Yun)

An ultra-wideband (UWB) slotted compact Vivaldi antenna with a microstrip line feed was evaluated for microwave imaging (MI) applications. The recommended FR4 substrate-based Vivaldi antenna is 50×50×1.5 mm3 in size. The proposed compact Vivaldi antenna showed good radiation characteristics and spanned an ultra-wide bandwidth of 10 GHz, ranging from 3.6 to 13.6 GHz, using a microstrip line feed with a radial stub, tapered slots, and resonant cavities. After analyzing the simulation findings, the proposed compact slotted UWB Vivaldi antenna was compared with the basic Vivaldi antenna. The maximum gain and efficiency were 7.35 dB and 82%, respectively. The antenna was simulated in the time domain using the computer simulation technology (CST) microwave studio. The frequency domain solver was also used for cross-validation, and the results were equally good. In conclusion, the proposed compact Vivaldi antenna is a good option for microwave imaging applications.

https://doi.org/10.5573/IEIESPC.2023.12.4.358

(Byungjoo Park) ; (Jaehwan Kim) ; (Janise McNair)

Security risks were raised during the handover of Mobile Nodes (MNs) in the Mobile Internet Protocol version 6 (MIPv6), which were demonstrated in the instigation of route optimization between the MN and its Correspondent Node (CN). The return routability procedure provides a basic security measure for the communication between the MN and CN, but this security scheme still suffers from various weaknesses and loopholes. Thus, various methodologies in securing the handovers for MIPv6 were incorporated in line with the security agreement with authentication, authorization, and accounting (AAA) infrastructure. This paper proposes an establishment of the Integration Security Authentication System based on the AAA infrastructure to maintain the security level of the fast handover mobility management to provide an improved Quality of Service.