Accuracy Of Recognition System Research Articles

Experimental Study on Using Synthetic Images as a Portion of Training Dataset for Object Recognition in Construction Site

The application of Artificial Intelligence (AI) across various industries necessitates the acquisition of relevant environmental data and the implementation of AI recognition learning based on this data. However, the data available in real-world environments are limited and difficult to obtain. Construction sites represent dynamic and hazardous environments with a significant workforce, making data acquisition challenging and labor-intensive. To address these issues, this experimental study explored the potential of generating synthetic data to overcome the challenges of obtaining data from hazardous construction sites. Additionally, this research investigated the feasibility of hybrid dataset in securing construction-site data by creating synthetic data for scaffolding, which has a high incidence of falls but low object recognition rates due to its linear object characteristics. We generated a dataset by superimposing scaffolding objects, from which the backgrounds were removed, onto various construction site background images. Using this dataset, we produced a hybrid dataset to assess the feasibility of synthetic data for construction sites and to evaluate improvements in object recognition performance. By finding the optimal composition ratio with real data and conducting model training, the highest accuracy was achieved at an 8:2 ratio, with a construction object recognition accuracy of 0.886. Therefore, this study aims to reduce the risk and labor associated with direct data collection at construction sites through a hybrid dataset, achieving data generation at a low cost and high efficiency. By generating synthetic data to find the optimal ratio and constructing a hybrid dataset, this research demonstrates the potential to address the problems of data scarcity and data quality on construction sites. The improvement in recognition accuracy of the construction safety management system is anticipated, suggesting that the creation of synthetic data for constructing a hybrid dataset can reduce construction safety-accident issues.

Defect Detection of Tomato Fruit Using Image Dataset

Abstract: The tomato crop is more susceptible to disease than any other vegetable, and it can be infected with over 200 diseases caused by different pathogens worldwide. Tomato plant diseases have become a challenge to food security globally. Currently, diagnosing and preventing tomato plant diseases is a challenge due to the lack of essential methods or tools.The traditional techniques of detecting plant disease are arduous and error-prone. Utilizing precise or automatic detection methods in spotting early plant disease can improve the quality of food production and reduce adverse effects. Deep learning has significantly increased the recognition accuracy of image classification and object detection systems in recent years. In this study, a 15-layer convolutional neural network is proposed as the backbone for single shot detector (SSD) to improve the detection of healthy, and three classes of tomato fruit diseases. The proposed model performance is compared with ResNet-50, AlexNet, VGG 16, and VGG19 as the backbone for Single shot detector. The findings of the experiment showed that the proposed CNN-SDD achieved 98.87% higher detection accuracy, which outperformed state-of-the-art models.

Robot control system based on deep learning and RPA

Robots can help people complete repetitive and high-risk tasks, such as industrial production, medical care, environmental monitoring, etc. The control system of robots is the key to their ability to complete tasks, and studying robot control systems is of great significance. This article used Convolutional Neural Network (CNN) and Robotic Process Automation (RPA) technologies to optimize and train the robot control system and constructed a robot control system. This article conducts perception and decision-making experiments and execution experiments in the experimental section. According to the experimental results, it can be concluded that the average image recognition accuracy of the robot control system in perception and decision-making experiments was 94.62%. The average decision accuracy was 87.5%, and the average time efficiency was 176 seconds. During the execution of the experiment, the deviation of the motion trajectory shall not exceed 5 cm, and the oscillation amplitude shall not exceed 6°; the distance from the obstacle shall not exceed 20 cm, and the movement speed shall not exceed 0.6 m/s; the operating time shall not exceed 25 hours, and the number of faults shall not exceed 0.2 times per hour, all within the normal range. The robot control system based on Deep Learning (DL) and RPA has broad application prospects and research value, which would bring new opportunities and challenges to the development and application of robot technology.

Multi-layer CNN-LSTM network with self-attention mechanism for robust estimation of nonlinear uncertain systems.

With the help of robot technology, intelligent rehabilitation of patients with lower limb motor dysfunction caused by stroke can be realized. A key factor constraining the clinical application of rehabilitation robots is how to realize pattern recognition of human movement intentions by using the surface electromyography (sEMG) sensors to ensure unhindered human-robot interaction. A multilayer CNN-LSTM prediction network incorporating the self-attention mechanism (SAM) is proposed, in this paper, which can extract and learn the periodic and trend characteristics of the sEMG signals, and realize the accurate autoregressive prediction of the human motion information. Firstly, the multilayer CNN-LSTM network utilizes the CNN layer for initial feature extraction of data, and the LSTM network is used to improve the enhancement of the historical time-series features. Then, the SAM is used to improve the global feature extraction performance and parallel computation speed of the network. In comparison with existing test is carried out using actual data from five healthy subjects as well as a clinical hemiplegic patient to verify the superiority and practicality of the proposed algorithm. The results show that most of the model's prediction R > 0.9 for different motion states of healthy subjects; in the experiments oriented to the motion characteristics of patient subjects, the angle prediction results of R > 0.99 for the untrained data on the affected side, which proves that our proposed model also has a better effect on the angle prediction of the affected side. The main contribution of this paper is to realize continuous motion estimation of ankle joint for healthy and hemiplegic individuals under non-ideal conditions (weak sEMG signals, muscle fatigue, high muscle tension, etc.), which improves the pattern recognition accuracy and robustness of the sEMG sensor-based system.

Better than Random: Reliable NLG Human Evaluation with Constrained Active Sampling

Human evaluation is viewed as a reliable evaluation method for NLG which is expensive and time-consuming. To save labor and costs, researchers usually perform human evaluation on a small subset of data sampled from the whole dataset in practice. However, different selection subsets will lead to different rankings of the systems. To give a more correct inter-system ranking and make the gold standard human evaluation more reliable, we propose a Constrained Active Sampling Framework (CASF) for reliable human judgment. CASF operates through a Learner, a Systematic Sampler and a Constrained Controller to select representative samples for getting a more correct inter-system ranking. Experiment results on 137 real NLG evaluation setups with 44 human evaluation metrics across 16 datasets and 5 NLG tasks demonstrate CASF receives 93.18\% top-ranked system recognition accuracy and ranks first or ranks second on 90.91\% of the human metrics with 0.83 overall inter-system ranking Kendall correlation. Code and data are publicly available online.

Using transfer learning-based plant disease classification and detection for sustainable agriculture

Subsistence farmers and global food security depend on sufficient food production, which aligns with the UN's “Zero Hunger,” “Climate Action,” and “Responsible Consumption and Production” sustainable development goals. In addition to already available methods for early disease detection and classification facing overfitting and fine feature extraction complexities during the training process, how early signs of green attacks can be identified or classified remains uncertain. Most pests and disease symptoms are seen in plant leaves and fruits, yet their diagnosis by experts in the laboratory is expensive, tedious, labor-intensive, and time-consuming. Notably, how plant pests and diseases can be appropriately detected and timely prevented is a hotspot paradigm in smart, sustainable agriculture remains unknown. In recent years, deep transfer learning has demonstrated tremendous advances in the recognition accuracy of object detection and image classification systems since these frameworks utilize previously acquired knowledge to solve similar problems more effectively and quickly. Therefore, in this research, we introduce two plant disease detection (PDDNet) models of early fusion (AE) and the lead voting ensemble (LVE) integrated with nine pre-trained convolutional neural networks (CNNs) and fine-tuned by deep feature extraction for efficient plant disease identification and classification. The experiments were carried out on 15 classes of the popular PlantVillage dataset, which has 54,305 image samples of different plant disease species in 38 categories. Hyperparameter fine-tuning was done with popular pre-trained models, including DenseNet201, ResNet101, ResNet50, GoogleNet, AlexNet, ResNet18, EfficientNetB7, NASNetMobile, and ConvNeXtSmall. We test these CNNs on the stated plant disease detection and classification problem, both independently and as part of an ensemble. In the final phase, a logistic regression (LR) classifier is utilized to determine the performance of various CNN model combinations. A comparative analysis was also performed on classifiers, deep learning, the proposed model, and similar state-of-the-art studies. The experiments demonstrated that PDDNet-AE and PDDNet-LVE achieved 96.74% and 97.79%, respectively, compared to current CNNs when tested on several plant diseases, depicting its exceptional robustness and generalization capabilities and mitigating current concerns in plant disease detection and classification.

Tri-stage offline Telugu character recognition system based on fusion of HOG and ULBP

The domain of pattern recognition has experienced a notable upsurge in the prevalence of handwritten character recognition, which is a research area that has been widely recognized for decades.However, this task presents significant challenges, especially for Indian languages, due to the vast array of character shapes and curved linear structures they possess. Despite some research having been conducted on Indian languages such as Kannada, Hindi, Tamil Devanagari, and Telugu, among others, the field is still in its nascent stages of development. The nascent phase of handwritten Telugu script presents a promising opportunity for further investigation and exploration. We introduce a Tri-stage Telugu handwritten recognition system that utilizes hyper parameters of k- NN, SVM, and Random Forest based on the fusion of HOG and ULBP features. By implementing a Tri-stage classifier in a sequential manner, the system's recognition accuracy is significantly enhanced.The proposed Tri-stage classifier achieves exceptionally high accuracies on Telugu handwritten character HP datasets.

An Iris Image Super-Resolution Model Based on Swin Transformer and Generative Adversarial Network

The process of iris recognition can result in a decline in recognition performance when the resolution of the iris images is insufficient. In this study, a super-resolution model for iris images, namely SwinGIris, which combines the Swin Transformer and the Generative Adversarial Network (GAN), is introduced. SwinGIris performs quadruple super-resolution reconstruction for low-resolution iris images, aiming to improve the resolution of iris images and thereby improving the recognition accuracy of iris recognition systems. The model utilizes residual Swin Transformer blocks to extract depth global features, and the progressive upsampling method along with sub-pixel convolution is conducive to focusing on the high-frequency iris information in the presence of more non-iris information. In order to preserve high-frequency details, the discriminator employs a VGG-style relative classifier to guide the generator in generating super-resolution images. In experimental section, we enhance low-resolution (56 × 56) iris images to high-resolution (224 × 224) iris images. Experimental results indicate that the SwinGIris model achieves satisfactory outcomes in restoring low-resolution iris image textures while preserving identity information.

LD-Recognition: Classroom Action Recognition Based on Passive RFID

Classroom learning is one of the main ways for students to acquire knowledge because the class usually has a large number of students and it is difficult for teachers to pay attention to the learning status of each student at the same time. Therefore, mastering the learning status of each student and dealing with it is the key factor to determine the quality of course teaching, and classroom actions can accurately reflect the learning state of students. Based on this, in this article, a passive radio frequency identification (RFID)-based classroom action recognition system LD-recognition is proposed. The system pastes the label on the right side of the desk, and the learning state of the students was judged by recognizing the four movements of raising the left hand, raising the right hand, nodding off, and holding the book. The system uses a multichannel attentional graph convolutional neural network (ATGCN) to deeply learn the phase and signal strength of actions and conduct action recognition. LD-recognition verifies the accuracy of actions from different distances, different experimenters, and different network models. The experimental results show that the recognition accuracy of LD-recognition system is high, reaching 96.9% on average.

Research on the Path of English Translation Teaching and External Communication Based on the Multimodal Analysis Method

Abstract With the development of the times and the continuous innovation of modern education technology, the multimodal theory is developing deeply in the field of education. The first step in this study is to use the optimization principle to select modes that meet the needs of English translation teaching and then develop a multimodal analysis framework for external communication that incorporates multimedia communication methods. Then, we constructed an emotion recognition model for English translation teaching by using speech recognition and expression recognition combined with convolutional neural networks for the students’ classroom situation and attitude and carried out the empirical analysis. The recognition accuracy of the emotion recognition system constructed in this paper is 86.4%. In the multimodal analysis of English translation teaching, the teacher who uses the linguistic modality type for the most number of times 387 times, and the teacher who uses the mixed modality for the most number of times 271 times, the students have the best classroom status. It has been stable at about 5 points, which is a “more positive” state. The students in the class had the highest level of motivation. Through the analysis, it is found that multimodal teaching can effectively improve the students’ status in class and keep their attention, which verifies the appropriateness of the external communication path combined with multimedia communication in multimodal English translation teaching. It provides a practical path for the development of English translation teaching.

A cost-effective over-temperature alarm system for cold chain delivery

Managing over-temperature alarms in cold chain logistics (CCL) is crucial for temperature monitoring systems. However, existing research in this area primarily focuses on improving identification accuracy while overlooking the associated cost implications. This study addresses this gap by proposing a cost-effective over-temperature alarm system using an artificial neural network (ANN) model that integrates multi-source data (MSD). This article demonstrates that utilizing ANN and MSD leads to significantly improved recognition accuracy compared to using single-source data while keeping the cost increase to a minimum. The paper found recognition accuracy to be 97.4% with three feature values and 98.6% with six feature values. Sensitivity analyses reveal that factors such as the initial food temperature, cumulative open time, data acquisition interval, and algorithm significantly impact the recognition accuracy of the over-temperature alarm system. The location of sensors, however, has a slight effect. Furthermore, the choice of the number of feature values and the granularity of the data influences the cost of temperature management and recognition accuracy. This study provides valuable insights into the factors influencing recognition accuracy, aiding in designing cost-effective over-temperature alarm systems for food supply chain management.

Deep neural network based acoustic pattern recognition system for fault localization application

Abstract This paper explores the process of traditional voiceprint recognition, analyzes the traditional GMM recognition algorithm, and proposes a GE2E-based voiceprint recognition algorithm by combining it with the deep neural network. It firstly uses the Bi-GRU network to replace the LSTM network to prevent the lack of semantic information, then adds the SGD algorithm to optimize the speech features, and finally improves the stability and accuracy of recognition by the GE2E loss function. On this basis, a voiceprint recognition system based on GE2E is designed, and the overall performance of the system is tested. Additionally, a voiceprint recognition system is being explored for fault localization. The results show that the recognition accuracy of male voiceprints in the test is at [0.89,0.95], and the recognition accuracy of female voiceprints is at [0.88,0.96], and there is not much difference in the voiceprint recognition accuracy of the voiceprint recognition system for both male and female students, and the overall recognition accuracy is greater than 0.9. When applied in fault location, the error between the measured distance and the actual fault distance is within 0.1 meters, enabling fault location.

Smart Farming Using Deep Learning

The agricultural sector plays important role in supplying quality food and makes the greatest contribution to growing economies and populations. Agriculture is extremely important in human life. Almost 60% of the population is engaged in some kind of agriculture either directly or indirectly. Plant disease may cause significant losses in food production and eradicate diversity in species. Early of plant disease using accurate of automatic detection techniques can enhance the quality of food production and minimize economic losses. In recent years, deep learning has brought tremendous improvements in the recognition accuracy of image classification and object detection systems. Hence, in this paper, we utilized convolutional neural network (CNN)-based pre-trained models for efficient plant disease detection. The proposed method uses a convolutional neural network and deep neural network to identify and recognize crop disease symptoms effectively and accurately. This Research paper offers a through description of the DL models that are used to visualize crop diseases. In this experiment we use plant dataset, which has real time image samples of different plant, fruit, flower in different. The proposed methodology aims to develop a convolution neural network-Based strategy for detecting plant leaf, fruit, flower disease.

The Design of Intelligent Building Lighting Control System Based on CNN in Embedded Microprocessor

A convolutional neural network (CNN) was designed and built on an embedded building lighting control system to determine whether the application of CNN could increase the accuracy of image recognition and reduce energy consumption. Currently, lighting control systems rely mainly on information technology, with sensors to detect people’s existence or absence in an environment. However, due to the deviation of this perception, the accuracy of image detection is not high. In order to validate the effectiveness of the new system based on CNN, an experiment was designed and operated. The importance of the research lies in the fact that high image detection would bring in less energy consumption. The result of the experiment indicated that, when comparing the actual position with the positioning position, the difference was between 0.01 to 0.20 m, indicating that the image recognition accuracy of the CNN-based embedded control system was very high. Moreover, comparing the luminous flux of the designed system with natural light and the designed system without natural light with the system without intelligent control, the energy savings is about 40%.

A real-time smart sensing system for automatic localization and recognition of vegetable plants for weed control.

Tomato is a globally grown vegetable crop with high economic and nutritional values. Tomato production is being threatened by weeds. This effect is more pronounced in the early stages of tomato plant growth. Thus weed management in the early stages of tomato plant growth is very critical. The increasing labor cost of manual weeding and the negative impact on human health and the environment caused by the overuse of herbicides are driving the development of smart weeders. The core task that needs to be addressed in developing a smart weeder is to accurately distinguish vegetable crops from weeds in real time. In this study, a new approach is proposed to locate tomato and pakchoi plants in real time based on an integrated sensing system consisting of camera and color mark sensors. The selection scheme of reference, color, area, and category of plant labels for sensor identification was examined. The impact of the number of sensors and the size of the signal tolerance region on the system recognition accuracy was also evaluated. The experimental results demonstrated that the color mark sensor using the main stem of tomato as the reference exhibited higher performance than that of pakchoi in identifying the plant labels. The scheme of applying white topical markers on the lower main stem of the tomato plant is optimal. The effectiveness of the six sensors used by the system to detect plant labels was demonstrated. The computer vision algorithm proposed in this study was specially developed for the sensing system, yielding the highest overall accuracy of 95.19% for tomato and pakchoi localization. The proposed sensor-based system is highly accurate and reliable for automatic localization of vegetable plants for weed control in real time.

Estimate the noise effect on automatic speech recognition accuracy for mandarin by an approach associating articulation index

The automatic speech recognition (ASR) technology has become one of the fast-growing engineering technologies. The ASR accuracy is often reduced by environmental interference noise. Many studies attempted to improve the ASR word accuracy in noisy backgrounds, but few were concerned about estimating the influence of noise on the word accuracy of ASR systems. This study investigated the effect of stationary noise on the word accuracy for mandarin of two ASR systems (i.e., the Baidu Cloud and Microsoft Azure). We selected the speech material from an open Mandarin speech database. The interference noise contained pink noise, white noise, random noise with equivalent power spectrum to the average power spectrum of mandarin, band-stop filtered noise and several typical household appliances noise (i.e., hairdryer, range hood, blender and water purifier). The signal-to-noise ratios (SNRs) were set to the range −12–3 dB. The Azure ASR system was less affected by noise and tended to have a higher word accuracy under the same noise condition than the Baidu ASR system. We proposed an approach to associate the articulation index (AI) of interference noise with the speech recognition accuracy of ASR systems. A three-parameter logistic model could be established for a specific ASR system, and each parameter is linearly dependent on the AI of interference noise. The performance of ASR systems under different stationary noise conditions could be quantitatively assessed and compared through the approach. Different models should be established for different versions of the ASR systems through this AI-associating approach since the parameters of one model are only applicable to one particular version of an ASR system.

Biometric Template Protection for Neural-Network-Based Face Recognition Systems: A Survey of Methods and Evaluation Techniques

As automated face recognition applications tend towards ubiquity, there is a growing need to secure the sensitive face data used within these systems. This paper presents a survey of biometric template protection (BTP) methods proposed for securing face “templates” (images/features) in neural-network-based face recognition systems. The BTP methods are categorised into two types: Non-NN and NN-learned. Non-NN methods use a neural network (NN) as a feature extractor, but the BTP part is based on a non-NN algorithm applied at either image-level or feature-level. In contrast, NN-learned methods specifically employ a NN to learn a protected template from the unprotected face image/features. We present examples of Non-NN and NN-learned face BTP methods from the literature, along with a discussion of the two categories’ comparative strengths and weaknesses. We also investigate the techniques used to evaluate these BTP methods, in terms of the three most common BTP criteria: “recognition accuracy”, “irreversibility”, and “renewability/unlinkability”. As expected, the recognition accuracy of protected face recognition systems is generally evaluated using the same (empirical) techniques employed for evaluating standard (unprotected) biometric systems. On the contrary, most irreversibility and renewability/unlinkability evaluations are found to be based on theoretical assumptions/estimates or verbal implications, with a lack of empirical validation in a practical face recognition context. We recommend, therefore, a greater focus on empirical evaluation strategies, to provide more concrete insights into the irreversibility and renewability/unlinkability of face BTP methods in practice. Additionally, an exploration of the reproducibility of the studied BTP works, in terms of the public availability of their implementation code and evaluation datasets/procedures, suggests that it would currently be difficult for the BTP community to faithfully replicate (and thus validate) most of the reported findings. So, we advocate for a push towards reproducibility, in the hope of furthering our understanding of the face BTP research field.

SSEC: Semantic Segmentation and Ensemble Classification Framework for Static Hand Gesture Recognition using RGB-D Data

Hand Gesture Recognition (HGR) refers to identifying various hand postures used in Sign Language Recognition (SLR) and Human Computer Interaction (HCI) applications. Complex background in uncontrolled environmental condition is the major challenging issue which impacts the recognition accuracy of HGR system. This can be effectively addressed by discarding the background using suitable semantic segmentation method, where it predicts the hand region pixels into foreground and rest of the pixels into background. In this paper, we have analyzed and evaluated well known semantic segmentation architectures for hand region segmentation using both RGB and depth data. Further, ensemble of segmented RGB and depth stream is used for hand gesture classification through probability score fusion. Experimental results shows that the proposed novel framework of Semantic Segmentation and Ensemble Classification (SSEC) is suitable for static hand gesture recognition and achieved F1-score of 88.91% on OUHANDS test dataset.

Inertial Sensing for Lateral Walking Gait Detection and Application in Lateral Resistance Exoskeleton

Lateral walking gait detection is necessary for the development of wearable devices applied to side stepping. To our knowledge, rarely work has been conducted to identify lateral walking gait by wearable sensors. Based on a hip exoskeleton, we presented a method for lateral walking gait phase detection only by two IMUs mounted on the shank. Experiments were conducted to detect narrow double support, swing of the leading leg, wide double support and swing of trailing leg phases of 12 healthy subjects walking at various speeds. The performance of four different algorithms including thresholding (THR), modified k-nearest neighbor named urban buildings indexing (UBI), random forest (RF) and neural networks (NN) were evaluated. The occupation of space resources of NN is the smallest besides THR. The total recognition accuracy (mean and SEM) of the RF-based, UBI-based, and NN-based systems was 97.07 ± 0.07% (off-line), 96.64 ± 0.16% (off-line) and 95.22 ± 0.60% (real-time), respectively. The recognition time of the models based on RF, UBI, NN, and THR was 13.3 ± 1.1 ms, 5.7 ± 0.5 ms, 2.6 ± 0.2 ms and 0.8 ± 0.2 ms, respectively. The recognition accuracy (cross-subjects) of RF-based, UBI-based and NN-based systems was 91.72 ± 0.42%, 89.63 ± 0.48% and 89.60 ± 0.43%, respectively. The results demonstrated that the proposed method can be applied to lateral walking gait detection.

Optimized LightGBM Power Fingerprint Identification Based on Entropy Features.

The huge amount of power fingerprint data often has the problem of unbalanced categories and is difficult to upload by the limited data transmission rate for IoT communications. An optimized LightGBM power fingerprint extraction and identification method based on entropy features is proposed. First, the voltage and current signals were extracted on the basis of the time-domain features and V-I trajectory features, and a 56-dimensional original feature set containing six entropy features was constructed. Then, the Boruta algorithm with a light gradient boosting machine (LightGBM) as the base learner was used for feature selection of the original feature set, and a 23-dimensional optimal feature subset containing five entropy features was determined. Finally, the Optuna algorithm was used to optimize the hyperparameters of the LightGBM classifier. The classification performance of the power fingerprint identification model on imbalanced datasets was further improved by improving the loss function of the LightGBM model. The experimental results prove that the method can effectively reduce the computational complexity of feature extraction and reduce the amount of power fingerprint data transmission. It meets the recognition accuracy and efficiency requirements of a massive power fingerprint identification system.