International Journal of Pattern Recognition and Artificial Intelligence

Heart rate variability (HRV) has been used to determine whether patients have latent illnesses and as an assessment indicator of the autonomic nervous system. Studies have also employed HRV for lie detection, which is commonly performed in the law system. However, several studies have concluded that due to environmental and procedure-related problems, lie detectors may yield inconsistent results. This study used smart recognition technology and proposed a systemized method of lie detection. To verify the reliability of the noncontact lie detection approach, empirical lie detection studies were conducted and the relevant literature was searched for item selection techniques. This study employed a noncontact HRV lie detection system. By experimenting on 15 participants split into four groups, this study verified the reliability of using the physiological characteristic of HRV for lie detection.

To improve the accuracy of cross-domain object detection, the existing unsupervised domain adaptation (UDA) object detection methods mostly use Feature Pyramid Network (FPN), multiple Region Proposal Network (RPN), and multiple domain classifier, but these methods lead to complex network structures, slow model convergence, and low detection efficiency. To solve the above problems, this paper proposes an Efficient and Accurate Cross-domain Object Detection Method Using One-level Feature and Domain Adaptation (OFDA). This method realizes one-level feature object detection through feature fusion and divide-and-conquer technology; realizes overfitting feature suppression and unsupervised domain adaptation through domain-specific suppression and domain feature alignment technology; realizes background feature suppression through the Objectness branch, which replaces the time-consuming Region Proposal Network (RPN) structure and improves the efficiency of unsupervised adaptive detection. The paper verifies the feasibility and superiority of the proposed method by the comparative experiments and ablation experiments of multiple datasets. The proposed OFDA method not only improves the efficiency of object detection, but also ensures the accuracy of cross-domain detection.

Counting the number of people in a specific area is crucial in maintaining proper crowd safety and management especially in highly-congested indoor scenarios. Recent convolutional neural network (CNN) approaches integrate auxiliary sub-networks to increase the accuracy of the model in estimating crowd size. However, these models require large computational costs due to additional calculations, resulting in an impractically slow inference speed for real-world applications. In this paper, we propose a fast, efficient, and robust crowd counting model called Condensed Network or ConNet. We utilize a composite technique composed of multiple compression methods to reduce the number of parameters of our proposed model. ConNet attains counting accuracy on par with state-of-the-art crowd counting methods on benchmark datasets featuring indoor scenes, while significantly reducing parameter count and increasing inference speed. Moreover, ConNet still performs accurately even with extreme changes in lighting conditions, image resolutions, and camera orientations. Our smallest model ConNet-04 has 61.0× less parameters and is at most 9.0× faster than the baseline approach. Our code and trained models are publicly available at https://github.com/mikatej/ConNet.

In this study, an algorithm, which had combined the match degree and entropy weighting method, was proposed to predict the efficiency of the photovoltaic output power. First, the key characteristic quantities were selected by the correlation analyses of output power history data and meteorological data of the photovoltaic power generation system. Second, according to the Euclidean distances between the predicted points and the historical data points, a method for the selections of similar sample points was proposed based on point-to-point theory. Finally, the match degrees between each characteristic quantity of the prediction points and the similar sample points were determined using the Mamdani reasoning method in fuzzy mathematics. The fitting weighting was then solved using the entropy weighting method, which was applied to fit the match degrees of the same characteristic quantities between the similar sample points. Next, the correlation coefficients were utilized as the weighting to fit the match degrees of the different characteristic quantities, resulting in a total match degree. This total match degree was then used to fit the photovoltaic output power of the similar sample points to the prediction points. We had proven that this method could have effective predictions and good adaptabilities in various weather statuses with high accuracy and real-time performance, especially, in the case of sudden weather change.

Computer vision and optical character recognition techniques have rapidly advanced in order to accurately capture text and other features from paper documents. While state-of-the-art tools in these fields now yield high accuracy, analyzing their outputs requires more research. Since tables are common in such documents, a new pipeline, based on multimodal learning, is proposed to better extract key–value pairs from tables. Its performance is evaluated with a synthetically generated dataset with randomly generated tables and a dataset of mechanical part documents provided by SiliconExpert Technologies. Its performance is also compared with another state-of-the-art model built for similar tasks, LayoutLM. The proposed pipeline provides a fully automated, end-to-end scalable solution, beginning with image processing and computer vision components to a machine learning model that uses data from optical character recognition and natural language processing to make the final decisions. In the best configuration, the pipeline achieved a 96.26% accuracy on a large, synthetically generated training and test set. When comparing the proposed pipeline with LayoutLM, the proposed pipeline performed similarly on the synthetic dataset and better on the real dataset. These results show the potential of the multimodal approach in extracting key–value pairs from tables from real paper documents.

With the changing lifestyle, a large population suffers from a bone disease known as an osteoarthritis affecting the knee, spine, and hip. Therefore, timely detection and classification of the disease are necessary to minimize the loss, however, it is a time-consuming task and requires various tests and physicians’ in-depth analysis. Thus, an accurate automated technique, timely detection and classification are needed to cope with the aforementioned challenges. This study proposes a technique based on an efficient DenseNet that uses the knee image’ features to identify the Knee Osteoarthritis (KOA) and determine its severity level according to the KL grading system such as Grade-I, Grade-II, Grade-III, and Grade-IV. We introduced the reweighted cross-entropy loss function which makes our proposed algorithm more robust as the training data is imbalanced. The dense connections of efficient DenseNet with regularization power help to reduce the overfitting during the training of small knee sample training sets. The proposed algorithm is an efficient approach that can identify the early symptoms of KOA and classify the severity level of the disease for better decision making by orthopedics. The algorithm is a pre-trained network that does not require a huge training set, therefore, the existing dataset i.e. Mendeley VI has been utilized for the training and testing. Additionally, cross-validation has been employed using the OAI dataset to assess the performance of the proposed model. The algorithm achieved 98.22% accuracy over the testing set and 98.08% accuracy over cross-validation. Various experiments have been performed to confirm that our proposed algorithm is more consistent and capable of detecting and classifying the KOA disease than existing state of the art.

The number of jellyfish outbreaks is on the rise around the world, and they have been considered a serious ecological disaster. As part of the emergency response plan for jellyfish disasters, in-situ detection research that can distinguish jellyfish species and quantities is urgently required to support accurate data collection. As a typical fully supervised regression task, counting is usually regarded as requiring a large number of labeled datasets in conventional counting methods. To treat counting as a few-shot regression task that is semi-supervised, a novel adaptation strategy based on deep learning is presented in this paper. The method combines the test image with several example objects from the test image and takes advantage of the strong similarities present in the test image and the example objects contained in the image. Effective counting can be achieved without training the target object. Prediction of the density map of the test image’s objects of interest is the objective of the test. This method has been shown to be more robust than the method of detection first and counting later, and its accuracy can exceed 95%.

Investigating realistic visual exploration is quite challenging in sport climbing, but it promises a deeper understanding of how performers adjust their perception-action couplings during task completion. However, the samples of participants and the number of trials analyzed in such experiments are often reduced to a minimum because of the time-consuming treatments of the eye-tracking data. Notably, mapping successive points of gaze from local views to the global scene is generally performed manually by watching eye-tracking video data frame by frame. This manual procedure is not suitable for processing a large number of datasets. Consequently, this study developed an automatic method for solving this global point of gaze localization in indoor sport climbing. Particularly, an eye-tracking device was used for acquiring local image frames and points of gaze from a climber’s local views. Artificial landmarks, designed as four-color-disk groups, were distributed on the wall to facilitate localization. Global points of gaze were computed based on planar homography transforms between the local and global positions of the detected landmarks. Thirty climbing trials were recorded and processed by the proposed methods. The success rates (Mean±SD) were up to 85.72%±13.90%, and the errors (Mean±SD) were up to $0.1302\pm 0.2051$m. The proposed method will be employed for computing global points of gaze in our current climbing dataset for understanding the dynamics intertwining of gaze and motor behaviors during the climbs.

Introduction: The echo states that networks in the Internet of Things (IoT) are currently being implemented in the widest sense. Echo state networks are fast and efficient recurrent neural networks. This consists of an input layer, a reservoir with many sparsely connected neurons, and an output layer. Issues: In the existing wireless sensor networks, strong mobility may disrupt an existing link between two communicating nodes. There is an inconvenience in data communication, and then it searches for a new node to build a better connection. Methods: To overcome these issues, the recently introduced echo state network (ESN) model opened the way to an extremely efficient approach for designing neural networks for temporal data. The study focuses on the ESN-enabled Intelligent Smart Sensor Design (IS2D) for creating the robotic nervous system with a smart healthcare Digital Nervous System (DNS) using the techniques of IoT, DNS, and Smart Sensor Design and Strain Sensor Fabrication (SSF). Results: Experimental results demonstrate the training set testing against the IS2D, the confusion matrix for ESN outcome, the real-time healthcare monitoring for the DNS, the IS2D sensor accuracy, and the DNS intensity calculation. Discussion: The performance analysis of the proposed model in realistic environments attests to the benefits of energy-centric metrics such as energy consumption, network lifetime, delay, and throughput. Finally, we discuss the challenges and opportunities by summarizing the study and proposing possible future works. The training set testing against the IS2D is based on time count, and the voltage result is estimated. The first portion of the data set should be 11.46% at the initial level. Further, this will increase from 1% to 5%, from 6% to 10%, and from 16% to 28% at the consecutive data set. The confusion matrix for ESN outcome is based on accuracy 28.45% higher than the existing strategies. In this part, the initial accuracy is 8.45% while accessing the initial stage. This value should increase with consecutive data sets from 18.45% to 28.45%.

Aspect-based sentiment classification is currently an important research direction to identify the sentiment expressed by sentences in different aspects. The primary approach for performing aspect-level sentiment analysis involves extracting both grammatical and semantic information. However, analyzing the grammatical connection between aspect words and other words within a review sentence using morphological features like part of speech can be exceedingly complex. This paper proposes the concept of sentiment-supporting words, dividing sentences into aspectual words, sentiment-supporting words and non-sentiment-supporting words, which simplifies the core task of sentiment analysis. Three rules are designed for determining the “sentiment-support words” of the text in different aspects. Subsequently, the application of sentiment-support words in sentiment analysis models is given, and five classical sentiment analysis models are improved accordingly. According to the experimental outcomes on two publicly available datasets, the “sentiment-support words” and corresponding sentiment support rules proposed in this paper are capable of significantly enhancing aspect-based sentiment analysis.

To solve the difficult problem of low-frequency noise processing in pipeline infrasonic leakage detection signals, a pipeline infrasonic leakage signal denoising method based on improved Prony algorithm and differential energy model was proposed to reduce the low-frequency interference noise in the signal and to improve the signal denoising effect. First, the frequency window of the effective signal is obtained according to the spectrogram of the infrasound signal. Aiming at the problem that the Prony algorithm is affected by noise, a difference energy model is proposed. The difference energy model is used to filter out the part of the frequency domain signal with large energy fluctuation, and a relatively stable preprocessing signal is obtained. In view of the instability of the traditional Prony algorithm, a Hankel matrix is established in the operation process. The stability is improved by extracting the extremum and residue of the signal instead of directly solving the sampling data points, and the extremum and residue of the effective signal are selected by combining the frequency window of the active ingredient. Finally, the effective signal is reconstructed to obtain a relatively stable infrasound leakage noise reduction signal. Experimental results show that the noise reduction technology based on the improved Prony algorithm and differential energy model can effectively reduce the noise of pipeline leakage signals. Compared with the traditional Prony algorithm, the noise reduction effect of the proposed method is up to 38.01% higher. Compared with the empirical mode decomposition method, the noise reduction effect of this method is improved by 9.25% at least, which opens up a new idea for pipeline leakage signal noise reduction.

In response to the aging trend in society and to Human Augmentation beings for home-based activities, this paper proposes an Abnormal Movement Detection system, using the common at-home movements of standing up and hand tremors while picking up items for abnormal movement verification. This can be easily applied in ordinary homes or long-term care institutions; for those living alone with limited resources, there is no longer any need to purchase expensive monitoring equipment to achieve improved quality of life. Therefore, this research collected and built the own dataset as the first important step of the study. The proposed Abnormal Movement Detection system is implemented by designing a deep learning network. Several issues, including the network architecture, the novel method of data augmentation and the scoring method of expanding the intervals between abnormality levels, are studied. For achieving the home-based real-time detection, there are four main contributions of this paper. The first is that a training dataset was collected and established: From this, the pathognomonic movement categories are easy to observe in home activities and geometric data augmentation can be used to improve the related home activity video collection. The second is the abnormal behavior detection architecture: This architecture has several important function blocks including detecting object, detecting action, inspecting abnormal movement and reminding event, using Convolutional Neural Network combined with Long Short-Term Memory ($\mathrm{\text{CNN}}+\mathrm{\text{LSTM}}$) as the core network for abnormal motion detection. With movement abnormality evaluation based on different levels, it can judge abnormal behaviors and conduct model training, performance evaluation and architecture optimization with both public domain datasets and the movement dataset collected in this research project. The third is the proliferation of new attributes in the videos: New attributes are added to the original videos through a Generative Adversarial Network (GAN), producing new training videos; the effectiveness of two different generation methods is evaluated. Finally, the algorithms developed in this paper are deployed on resource-constrained On-device Artificial Intelligence (AI). Activity videos from a total of 20 people were collected; in all, 53 videos of StandUp and 60 videos of PickUpItems were obtained to establish the training dataset. When CNN and LSTM network were added to Batch Normalization (BN), and Global Average Pooling (GAP) replaced Fully Connected (FC) layers, the accuracy rate reached 98.4%. In terms of data augmentation, geometric transformations and GAN were used to estimate the performance. The experimental results showed that the geometric transformation using brightness adjustment had the highest accuracy rate of 98.6%. Finally, the Softmax layer using Phi-Softmax–tan(⋅) function was shown to be the best method to expand the intervals between abnormality levels.

Many current multivariate filter feature selection approaches consider redundancy and relevance between features and class vectors simultaneously. However, these multivariate filter algorithms calculate the suitability of features by only the intrinsic characteristics of the data. In this paper, we suggest a new distributed framework to offset the multivariate feature selection problem. We propose the interaction with classifiers in multivariate filter feature selection. Our proposed framework calculates the relevance of each feature to class labels by embedded algorithms. Then, this technique examines redundancy among features through multivariate filter algorithms. In addition, in the proposed framework, we use horizontal distribution of data instead of using all them once. This approach reduces the runtime of the process in datasets with many samples and environments without centralized data. The results of the evaluation show that the proposed framework can improve classification accuracy compared with the methods just based on multivariate filters. In addition, the experimental results demonstrate that our algorithm outperforms compared approaches in precision, recall, and runtime.

In this paper, the operation process of cleaning of intelligent rice–wheat combine harvester is divided into two key steps: initial setting of cleaning device operation parameters and dynamic control of cleaning device operation parameters. Combined with the operation experience of cleaning control of agricultural machinery operators, the dynamic control knowledge-based system of cleaning device operation parameters was built based on production rule reasoning. The cleaning device of the rice–wheat combine harvester is intelligently controlled based on the dynamic monitoring and control system of the cleaning device operation quality and operation parameters, so as to achieve the purpose of controlling the cleaning operation quality of the rice–wheat combine harvester in the normal range. Through the field experiment results and analysis, it is proved that the intelligent control system of the cleaning device operation parameters based on the dynamic control knowledge-based system of cleaning device operation parameters can effectively keep the cleaning impurity content and loss rate of intelligent rice–wheat combine harvester in the normal range, so as to verify the effectiveness of the intelligent control knowledge-based system of the cleaning device operation parameters.