Search

Recent advancements in low-rank adaptation (LoRA) have shown its effectiveness in fine-tuning diffusion models for generating images tailored to new downstream tasks. Research on integrating multiple LoRA modules to accommodate new tasks has also gained traction. One emerging approach constructs several LoRA modules, but more than three typically decrease the generation performance of pre-trained models. The mixture-of-experts model solves the performance issue, but LoRA modules are not combined using text prompts; hence, generating images by combining LoRA modules does not dynamically reflect the user’s desired requirements. This paper proposes a LoRA fusion method that applies an attention mechanism to effectively capture the user’s text-prompting intent. This method computes the cosine similarity between predefined keys and queries and uses the weighted sum of the corresponding values to generate task-specific LoRA modules without the need for retraining. This method ensures stability when merging multiple LoRA modules and performs comparably to fully retrained LoRA models. The technique offers a more efficient and scalable solution for domain adaptation in large language models, effectively maintaining stability and performance as it adapts to new tasks. In the experiments, the proposed method outperformed existing methods in text–image alignment and image similarity. Specifically, the proposed method achieved a text–image alignment score of 0.744, surpassing an SVDiff score of 0.724, and a normalized linear arithmetic composition score of 0.698. Moreover, the proposed method generates superior semantically accurate and visually coherent images.

Audio music genre classification is performed to categorize audio music into various genres. Traditional approaches based on convolutional recurrent neural networks do not consider long temporal information, and their sequential structures result in longer training times and convergence difficulties. To overcome these problems, a traditional transformer-based approach was introduced. However, this approach employs pre-training based on momentum contrast (MoCo), a technique that increases computational costs owing to its reliance on extracting many negative samples and its use of highly sensitive hyperparameters. Consequently, this complicates the training process and increases the risk of learning imbalances between positive and negative sample sets. In this paper, a method for audio music genre classification called Deformer is proposed. The Deformer learns deep representations of audio music data through a denoising process, eliminating the need for MoCo and additional hyperparameters, thus reducing computational costs. In the denoising process, it employs a prior decoder to reconstruct the audio patches, thereby enhancing the interpretability of the representations. By calculating the mean squared error loss between the reconstructed and real patches, Deformer can learn a more refined representation of the audio data. The performance of the proposed method was experimentally compared with that of two distinct baseline models: one based on S3T and one employing a residual neural network-bidirectional gated recurrent unit (ResNet-BiGRU). The Deformer achieved an 84.5% accuracy, surpassing both the ResNet-BiGRU-based (81%) and S3T-based (81.1%) models, highlighting its superior performance in audio classification.

In unmanned systems, remote sensing is an approach that collects and analyzes data such as visual images, infrared thermal images, and LiDAR sensor data from a distance using a system that operates without human intervention. Recent advancements in deep learning enable the direct mapping of input images in remote sensing to desired outputs, making it possible to learn through imitation learning and for unmanned systems to learn by collecting and analyzing those images. In the case of autonomous cars, raw high-dimensional data are collected using sensors, which are mapped to the values of steering and throttle through a deep learning network to train imitation learning. Therefore, by imitation learning, the unmanned systems observe expert demonstrations and learn expert policies, even in complex environments. However, in imitation learning, collecting and analyzing a large number of images from the game environment incurs time and costs. Training with a limited dataset leads to a lack of understanding of the environment. There are some augmentation approaches that have the limitation of increasing the dataset because of considering only the locations of objects visited and estimated. Therefore, it is required to consider the diverse kinds of the location of objects not visited to solve the limitation. This paper proposes an enhanced model to augment the number of training images comprising a Preprocessor, an enhanced Swin Transformer model, and an Action model. Using the original network structure of the Swin Transformer model for image augmentation in imitation learning is challenging. Therefore, the internal structure of the Swin Transformer model is enhanced, and the Preprocessor and Action model are combined to augment training images. The proposed method was verified through an experimental process by learning from expert demonstrations and augmented images, which reduced the total loss from 1.24068 to 0.41616. Compared to expert demonstrations, the accuracy was approximately 86.4%, and the proposed method achieved 920 points and 1200 points more than the comparison model to verify generalization.

In reinforcement learning, the epsilon (ε)-greedy strategy is commonly employed as an exploration technique This method, however, leads to extensive initial exploration and prolonged learning periods. Existing approaches to mitigate this issue involve constraining the exploration range using expert data or utilizing pretrained models. Nevertheless, these methods do not effectively reduce the initial exploration range, as the exploration by the agent is limited to states adjacent to those included in the expert data. This paper proposes a method to reduce the initial exploration range in reinforcement learning through a pretrained transformer decoder on expert data. The proposed method involves pretraining a transformer decoder with massive expert data to guide the agent’s actions during the early learning stages. After achieving a certain learning threshold, the actions are determined using the epsilon-greedy strategy. An experiment was conducted in the basketball game FreeStyle1 to compare the proposed method with the traditional Deep Q-Network (DQN) using the epsilon-greedy strategy. The results indicated that the proposed method yielded approximately 2.5 times the average reward and a 26% higher win rate, proving its enhanced performance in reducing exploration range and optimizing learning times. This innovative method presents a significant improvement over traditional exploration techniques in reinforcement learning.

Artificial intelligence plays a significant role in traffic-accident detection. Traffic accidents involve a cascade of inadvertent events, making traditional detection approaches challenging. For instance, Convolutional Neural Network (CNN)-based approaches cannot analyze temporal relationships among objects, and Recurrent Neural Network (RNN)-based approaches suffer from low processing speeds and cannot detect traffic accidents simultaneously across multiple frames. Furthermore, these networks dismiss background interference in input video frames. This paper proposes a framework that begins by subtracting the background based on You Only Look Once (YOLOv5), which adaptively reduces background interference when detecting objects. Subsequently, the CNN encoder and Transformer decoder are combined into an end-to-end model to extract the spatial and temporal features between different time points, allowing for a parallel analysis between input video frames. The proposed framework was evaluated on the Car Crash Dataset through a series of comparison and ablation experiments. Our framework was benchmarked against three accident-detection models to evaluate its effectiveness, and the proposed framework demonstrated a superior accuracy of approximately 96%. The results of the ablation experiments indicate that when background subtraction was not incorporated into the proposed framework, the values of all evaluation indicators decreased by approximately 3%.

With the development of artificial intelligence, techniques such as machine learning, object detection, and trajectory tracking have been applied to various traffic fields to detect accidents and analyze their causes. However, detecting traffic accidents using closed-circuit television (CCTV) as an emerging subject in machine learning remains challenging because of complex traffic environments and limited vision. Traditional research has limitations in deducing the trajectories of accident-related objects and extracting the spatiotemporal relationships among objects. This paper proposes a traffic accident detection method that helps to determine whether each frame shows accidents by generating and considering object trajectories using influence maps and a convolutional neural network (CNN). The influence maps with spatiotemporal relationships were enhanced to improve the detection of traffic accidents. A CNN is utilized to extract latent representations from the influence maps produced by object trajectories. Car Accident Detection and Prediction (CADP) was utilized in the experiments to train our model, which achieved a traffic accident detection accuracy of approximately 95%. Thus, the proposed method attained remarkable results in terms of performance improvement compared to methods that only rely on CNN-based detection.

Artificial intelligence, particularly machine learning, has begun to permeate various real-world applications and is continually being explored in automatic music generation. The approaches to music generation can be broadly divided into two categories: rule-based and data-driven methods. Rule-based approaches rely on substantial prior knowledge and may struggle to handle large datasets, whereas data-driven approaches can solve these problems and have become increasingly popular. However, data-driven approaches still face challenges such as the difficulty of considering long-distance dependencies when handling discrete-sequence data and convergence during model training. Although the diffusion model has been introduced as a generative model to solve the convergence problem in generative adversarial networks, it has not yet been applied to discrete-sequence data. This paper proposes a transformer-based diffusion model known as MelodyDiffusion to handle discrete musical data and realize chord-conditioned melody generation. MelodyDiffusion replaces the U-nets used in traditional diffusion models with transformers to consider the long-distance dependencies using attention and parallel mechanisms. Moreover, a transformer-based encoder is designed to extract contextual information from chords as a condition to guide melody generation. MelodyDiffusion can automatically generate diverse melodies based on the provided chords in practical applications. The evaluation experiments, in which Hits@k was used as a metric to evaluate the restored melodies, demonstrate that the large-scale version of MelodyDiffusion achieves an accuracy of 72.41% (k = 1).

Recently, the craze of K-POP contents is promoting the development of Korea’s cultural and artistic industries. In particular, with the development of various K-POP contents, including dance, as well as the popularity of K-POP online due to the non-face-to-face social phenomenon of the Coronavirus Disease 2019 (COVID-19) era, interest in Korean dance and song has increased. Research on dance Artificial Intelligent (AI), such as artificial intelligence in a virtual environment, deepfake AI that transforms dancers into other people, and creative choreography AI that creates new dances by combining dance and music, is being actively conducted. Recently, the dance creative craze that creates new choreography is in the spotlight. Creative choreography AI technology requires the motions of various dancers to prepare a dance cover. This process causes problems, such as expensive input source datasets and the cost of switching to the target source to be used in the model. There is a problem in that different motions between various dance genres must be considered when converting. To solve this problem, it is necessary to promote creative choreography systems in a new direction while saving costs by enabling creative choreography without the use of expensive motion capture devices and minimizing the manpower of dancers according to consideration of various genres. This paper proposes a system in a virtual environment for automatically generating continuous K-POP creative choreography by deriving postures and gestures based on bidirectional long-short term memory (Bi-LSTM). K-POP dance videos and dance videos are collected in advance as input. Considering a dance video for defining a posture, users who want a choreography, a 3D dance character in the source movie, a new choreography is performed with Bi-LSTM and applied. For learning, considering creativity and popularity at the same time, the next motion is evaluated and selected with probability. If the proposed method is used, the effort for dataset collection can be reduced, and it is possible to provide an intensive AI research environment that generates creative choreography from various existing online dance videos.

The performance of game AI can significantly impact the purchase decisions of users. User experience (UX) technology can evaluate user satisfaction with game AI by analyzing user interaction input through a user interface (UI). Although traditional UX-based game agent systems use a UX evaluation to identify the common interaction trends of multiple users, there is a limit to evaluating UX data, i.e., creating a UX evaluation and identifying the interaction trend for each individual user. The loss of UX data features for each user should be minimized and reflected to provide a personalized game agent system for each user. This paper proposes a UX framework for game agent systems in which a UX data reduction method is applied to improve the interaction for each user. The proposed UX framework maintains non-trend data features in the UX dataset where overfitting occurs to provide a personalized game agent system for each user, achieved by minimizing the loss of UX data features for each user. The proposed UX framework is applied to a game called “Freestyle” to verify its performance. By using the proposed UX framework, the imbalanced UX dataset of the Freestyle game minimizes overfitting and becomes a UX dataset that reflects the interaction trend of each user. The UX dataset generated from the proposed UX framework is used to provide customized game agents of each user to enhanced interaction. Furthermore, the proposed UX framework is expected to contribute to the research on UX-based personalized services.

Deep learning technology has been extensively studied for its potential in music, notably for creative music generation research. Traditional music generation approaches based on recurrent neural networks cannot provide satisfactory long-distance dependencies. These approaches are typically designed for specific tasks, such as melody and chord generation, and cannot generate diverse music simultaneously. Pre-training is used in natural language processing to accomplish various tasks and overcome the limitation of long-distance dependencies. However, pre-training is not yet widely used in automatic music generation. Because of the differences in the attributes of language and music, traditional pre-trained models utilized in language modeling cannot be directly applied to music fields. This paper proposes a pre-trained model, MRBERT, for multitask-based music generation to learn melody and rhythm representation. The pre-trained model can be applied to music generation applications such as web-based music composers that includes the functions of melody and rhythm generation, modification, completion, and chord matching after being fine-tuned. The results of ablation experiments performed on the proposed model revealed that under the evaluation metrics of HITS@k, the pre-trained MRBERT considerably improved the performance of the generation tasks by 0.09–13.10% and 0.02–7.37%, compared to the usage of RNNs and the original BERT, respectively.

Traffic closed-circuit television (CCTV) devices can be used to detect and track objects on roads by designing and applying artificial intelligence and deep learning models. However, extracting useful information from the detected objects and determining the occurrence of traffic accidents are usually difficult. This paper proposes a CCTV frame-based hybrid traffic accident classification model that enables the identification of whether a frame includes accidents by generating object trajectories. The proposed model utilizes a Vision Transformer (ViT) and a Convolutional Neural Network (CNN) to extract latent representations from each frame and corresponding trajectories. The fusion of frame and trajectory features was performed to improve the traffic accident classification ability of the proposed hybrid method. In the experiments, the Car Accident Detection and Prediction (CADP) dataset was used to train the hybrid model, and the accuracy of the model was approximately 97%. The experimental results indicate that the proposed hybrid method demonstrates an improved classification performance compared to traditional models.

Machine learning is widely used in various practical applications with deep learning models demonstrating advantages in handling huge data. Treating music as a special language and using deep learning models to accomplish melody recognition, music generation, and music analysis has proven feasible. In certain music-related deep learning research, recurrent neural networks have been replaced with transformers. This has achieved significant results. In traditional approaches with recurrent neural networks, input sequences are limited in length. This paper proposes a method to generate chord progressions for melodies using a transformer-based sequence-to-sequence model, which is divided into a pre-trained encoder and decoder. A pre-trained encoder extracts contextual information from melodies, whereas a decoder uses this information to produce chords asynchronously and finally outputs chord progressions. The proposed method addresses length limitation issues while considering the harmony between chord progressions and melodies. Chord progressions can be generated for melodies in practical music composition applications. Evaluation experiments are conducted using the proposed method and three baseline models. The baseline models included the bidirectional long short-term memory (BLSTM), bidirectional encoder representation from transformers (BERT), and generative pre-trained transformer (GPT2). The proposed method outperformed the baseline models in Hits@k (k = 1) by 25.89, 1.54, and 2.13 %, respectively.

Research on the application of artificial intelligence (AI) in games has recently gained momentum. Most commercial games still use AI based on a finite state machine (FSM) due to complexity and cost considerations. However, FSM-based AI decreases user satisfaction given that it performs the same patterns of consecutive actions in the same situations. This necessitates a new AI approach that applies domain-specific expertise to existing reinforcement learning algorithms. We propose a behavioral-cloning-based advantage actor-critic (A2C) that improves learning performance by applying a behavioral cloning algorithm to an A2C algorithm in basketball games. The state normalization, reward function, and episode classification approaches are used with the behavioral-cloning-based A2C. The results of the comparative experiments with the traditional A2C algorithms validated the proposed method. Our proposed method using existing approaches solved the difficulty of learning in basketball games.

With developments in artificial intelligence (AI), it is possible for novel applications to utilize deep learning to compose music by the format of musical instrument digital interface (MIDI) even without any knowledge of musical theory. The composed music is generally evaluated by human-based Turing test, which is a subjective approach and does not provide any quantitative criteria. Therefore, objective evaluation approaches with many general descriptive parameters are applied to the evaluation of MIDI data while considering MIDI features such as pitch distances, chord rates, tone spans, drum patterns, etc. However, setting several general descriptive parameters manually on large datasets is difficult and has considerable generalization limitations. In this paper, an enhanced evaluation method based on random masking and sequence-to-sequence (Seq2Seq) model is proposed to evaluate MIDI data. An experiment was conducted on real MIDI data, generated MIDI data, and random MIDI data. The bilingual evaluation understudy (BLEU) is a common MIDI data evaluation approach and is used here to evaluate the performance of the proposed method in a comparative study. In the proposed method, the ratio of the average evaluation score of the generated MIDI data to that of the real MIDI data was 31%, while that of BLEU was 79%. The lesser the ratio, the greater the difference between the real MIDI data and generated MIDI data. This implies that the proposed method quantified the gap while accurately identifying real and generated MIDI data.

Accurate semantic analysis of LiDAR point clouds enables the interaction between intelligent vehicles and the real environment. This paper proposes a hybrid 2D and 3D Hough Net by combining 3D global Hough features and 2D local Hough features with a classification deep learning network. Firstly, the 3D object point clouds are mapped into the 3D Hough space to extract the global Hough features. The generated global Hough features are input into the 3D convolutional neural network for training global features. Furthermore, a multi-scale critical point sampling method is designed to extract critical points in the 2D views projected from the point clouds to reduce the computation of redundant points. To extract local features, a grid-based dynamic nearest neighbors algorithm is designed by searching the neighbors of the critical points. Finally, the two networks are connected to the full connection layer, which is input into fully connected layers for object classification.

Three-dimensional (3D) point clouds are important for many applications, including object tracking and 3D scene reconstruction. Point clouds are usually obtained from laser scanners, but their high cost impedes the widespread adoption of this technology. We propose a method to generate the 3D point cloud corresponding to a single red–green–blue (RGB) image. The method retrieves high-quality 3D data from two-dimensional (2D) images captured by conventional cameras, which are generally less expensive. The proposed method comprises two stages. First, a generative adversarial network generates a depth image estimation from a single RGB image. Then, the 3D point cloud is calculated from the depth image. The estimation relies on the parameters of the depth camera employed to generate the training data. The experimental results verify that the proposed method provides high-quality 3D point clouds from single 2D images. Moreover, the method does not require a PC with outstanding computational resources, further reducing implementation costs, as only a moderate-capacity graphics processing unit can efficiently handle the calculations.

A Deep-Q-Network (DQN) controls a virtual agent as the level of a player using only screenshots as inputs. Replay memory selects a limited number of experience replays according to an arbitrary batch size and updates them using the associated Q-function. Hence, relatively fewer experience replays of different states are utilized when the number of states is fixed and the state of the randomly selected transitions becomes identical or similar. The DQN may not be applicable in some environments where it is necessary to perform the learning process using more experience replays than is required by the limited batch size. In addition, because it is unknown whether each action can be executed, a problem of an increasing amount of repetitive learning occurs as more non-executable actions are selected. In this study, an enhanced DQN framework is proposed to resolve the batch size problem and reduce the learning time of a DQN in an environment with numerous non-executable actions. In the proposed framework, non-executable actions are filtered to reduce the number of selectable actions to identify the optimal action for the current state. The proposed method was validated in Gomoku, a strategy board game, in which the application of a traditional DQN would be difficult.

This paper proposes an unmanned aerial vehicle (UAV) flight control method where a graphbased path is generated after the collected UAV flight data by a pilot are analyzed. UAV flights are planned by using hierarchical A* search algorithms based on graph-based generated flight paths to take images at multiple surveillance points. Generating a graph-based path makes it possible for UAVs to fly autonomously along paths shorter than that of the pilot collecting UAV flight data given that the shorter paths can be derived by connecting partially flied paths. A* search algorithms can be applied hierarchically to a graph-based path that contains circulation paths. The proposed method was experimentally verified through an analysis of the collected UAV flight data to generate graph-based and planned paths. The pilot flew the UAV six times and obtained 8115 UAV flight data points. The generated graph-based path included 17 monitoring points for taking surveillance images and 90 intermediate flight points. The length of the flight paths collected by six time flights was 1364.32 m, and the length of the flight paths by the proposed method was 764.27 m. Given that 8115 flight points were collected and 109 flight points were selected by the proposed method, the complexity of the generated graph-based path consisted of flight points was reduced to 1.34% by hierarchical A* search algorithms.

With unlabeled music data widely available, it is necessary to build an unsupervised latent music representation extractor to improve the performance of classification models. This paper proposes an unsupervised latent music representation learning method based on a deep 3D convolutional denoising autoencoder (3D-DCDAE) for music genre classification, which aims to learn common representations from a large amount of unlabeled data to improve the performance of music genre classification. Specifically, unlabeled MIDI files are applied to 3D-DCDAE to extract latent representations by denoising and reconstructing input data. Next, a decoder is utilized to assist the 3D-DCDAE in training. After 3D-DCDAE training, the decoder is replaced by a multilayer perceptron (MLP) classifier for music genre classification. Through the unsupervised latent representations learning method, unlabeled data can be applied to classification tasks so that the problem of limiting classification performance due to insufficient labeled data can be solved. In addition, the unsupervised 3D-DCDAE can consider the musicological structure to expand the understanding of the music field and improve performance in music genre classification. In the experiments, which utilized the Lakh MIDI dataset, a large amount of unlabeled data was utilized to train the 3D-DCDAE, obtaining a denoising and reconstruction accuracy of approximately 98%. A small amount of labeled data was utilized for training a classification model consisting of the trained 3D-DCDAE and the MLP classifier, which achieved a classification accuracy of approximately 88%. The experimental results show that the model achieves state-of-the-art performance and significantly outperforms other methods for music genre classification with only a small amount of labeled data.

Recently, with the development of computer technology, deep learning has expanded to the field of art, which requires creativity, which is a unique ability of humans, and an understanding of the human emotions expressed in art to process them as data. The field of art is integrating with various industrial fields, among which artificial intelligence (AI) is being used in stage art, to create visual images. As it is difficult for a computer to process emotions expressed in songs as data, existing stage background images for song performances are human designed. Recently, research has been conducted to enable AI to design stage background images on behalf of humans. However, there is no research on reflecting emotions contained in song lyrics to stage background images. This paper proposes a style transformation method to reflect emotions in stage background images. First, multiple verses and choruses are derived from song lyrics, one at a time, and emotion words included in each verse and chorus are extracted. Second, the probability distribution of the emotion words is calculated for each verse and chorus, and the image with the most similar probability distribution from an image dataset with emotion word tags in advance is selected for each verse and chorus. Finally, for each verse and chorus, the stage background images with the transferred style are outputted. Through an experiment, the similarity between the stage background and the image transferred to the style of the image with similar emotion words probability distribution was 38%, and the similarity between the stage background image and the image transferred to the style of the image with completely different emotion word probability distribution was 8%. The proposed method reduced the total variation loss of change from 1.0777 to 0.1597. The total variation loss is the sum of content loss and style loss based on weights. This shows that the style transferred image is close to edge information about the content of the input image, and the style is close to the target style image.

Internet of Things simulations play significant roles in the diverse kinds of activities in our daily lives and have been extensively researched. Creating and controlling virtual agents in three-dimensional Internet of Things simulations is a key technology for achieving realism in three-dimensional simulations. Given that traditional virtual agent-based approaches have limitations for realism, it is necessary to improve the realism of three-dimensional Internet of Things simulations. This article proposes a Q-Network-based motivation framework that applies a Q-Network to select motivations from desires and hierarchical task network planning to execute actions based on goals of the selected motivations. The desires are to be identified and calculated based on states. Selected motivations will be chosen to determine the goals that agents must achieve. In the experiments, the proposed framework achieved an average accuracy of up to 85.5% when the Q-Network-based motivation model was trained. To verify the Q-Network-based motivation framework, a traditional Q-learning is also applied in the three-dimensional virtual environment. Comparing the results of the two frameworks, the Q-Network-based motivation framework shows better results than those of traditional Q-learning, as the accuracy of the Q-Network-based motivation is higher by 15.58%. The proposed framework can be applied to the diverse kinds of Internet of Things systems such as a training autonomous vehicle. Moreover, the proposed framework can generate big data on animal behaviors for other training systems.

Unmanned aerial vehicles may collide with obstacles, such as trees or other unmanned aerial vehicles, while flying. A waypoint-based flight path is an approach to avoid such obstacles. To specify waypoints for the safe flight of unmanned aerial vehicles, it is necessary to define a flight path in advance by analyzing the flight records of unmanned aerial vehicles and thereby designate the waypoints automatically. However, there is a problem in that pilots tend to make errors in controlling unmanned aerial vehicles and collecting flight records. This article proposes a method to generate candidate waypoints for a flight path by removing such unintended flight records. In this method, images representing the positions in the collected flight records are generated. The candidate waypoints are generated as positions corresponding to the overlapping pixels of the images generated via image accumulation based on the flight records and the ones generated by accumulating the images reconstructed using an Autoencoder. The unmanned aerial vehicles can be set the waypoints for an autonomous flight using the candidate waypoints. An experiment was conducted in a university to generate candidate waypoints for road monitoring. The results obtained using the proposed method and K -means algorithm were compared. The candidate waypoints generated using the proposed method were reduced by 84.21% compared to those generated using the K -means algorithm.

Three-dimensional virtual environments can be configured as test environments of autonomous things, and remote sensing by 3D point clouds collected by light detection and range (LiDAR) can be used to detect virtual human objects by segmenting collected 3D point clouds in a virtual environment. The use of a traditional encoder-decoder model, such as DeepLabV3, improves the quality of the low-density 3D point clouds of human objects, where the quality is determined by the measurement gap of the LiDAR lasers. However, whenever a human object with a surrounding environment in a 3D point cloud is used by the traditional encoder-decoder model, it is difficult to increase the density fitting of the human object. This paper proposes a DeepLabV3-Refiner model, which is a model that refines the fit of human objects using human objects whose density has been increased through DeepLabV3. An RGB image that has a segmented human object is defined as a dense segmented image. DeepLabV3 is used to make predictions of dense segmented images and 3D point clouds for human objects in 3D point clouds. In the Refiner model, the results of DeepLabV3 are refined to fit human objects, and a dense segmented image fit to human objects is predicted. The dense 3D point cloud is calculated using the dense segmented image provided by the DeepLabV3-Refiner model. The 3D point clouds that were analyzed by the DeepLabV3-Refiner model had a 4-fold increase in density, which was verified experimentally. The proposed method had a 0.6% increase in density accuracy compared to that of DeepLabV3, and a 2.8-fold increase in the density corresponding to the human object. The proposed method was able to provide a 3D point cloud that increased the density to fit the human object. The proposed method can be used to provide an accurate 3D virtual environment by using the improved 3D point clouds.

Music is a type of time-series data. As the size of the data increases, it is a challenge to build robust music genre classification systems from massive amounts of music data. Robust systems require large amounts of labeled music data, which necessitates time- and labor-intensive data-labeling efforts and expert knowledge. This paper proposes a musical instrument digital interface (MIDI) preprocessing method, Pitch to Vector (Pitch2vec), and a deep bidirectional transformers-based masked predictive encoder (MPE) method for music genre classification. The MIDI files are considered as input. MIDI files are converted to the vector sequence by Pitch2vec before being input into the MPE. By unsupervised learning, the MPE based on deep bidirectional transformers is designed to extract bidirectional representations automatically, which are musicological insight. In contrast to other deep-learning models, such as recurrent neural network (RNN)-based models, the MPE method enables parallelization over time-steps, leading to faster training. To evaluate the performance of the proposed method, experiments were conducted on the Lakh MIDI music dataset. During MPE training, approximately 400,000 MIDI segments were utilized for the MPE, for which the recovery accuracy rate reached 97%. In the music genre classification task, the accuracy rate and other indicators of the proposed method were more than 94%. The experimental results indicate that the proposed method improves classification performance compared with state-of-the-art models.

Gomoku is a two-player board game that originated in ancient China. There are various cases of developing Gomoku using artificial intelligence, such as a genetic algorithm and a tree search algorithm. Alpha-Gomoku, Gomoku AI built with Alpha-Go’s algorithm, defines all possible situations in the Gomoku board using Monte-Carlo tree search (MCTS), and minimizes the probability of learning other correct answers in the duplicated Gomoku board situation. However, in the tree search algorithm, the accuracy drops, because the classification criteria are manually set. In this paper, we propose an improved reinforcement learning-based high-level decision approach using convolutional neural networks (CNN). The proposed algorithm expresses each state as One-Hot Encoding based vectors and determines the state of the Gomoku board by combining the similar state of One-Hot Encoding based vectors. Thus, in a case where a stone that is determined by CNN has already been placed or cannot be placed, we suggest a method for selecting an alternative. We verify the proposed method of Gomoku AI in GuPyEngine, a Python-based 3D simulation platform.

Deep learning has made significant progress in the field of automatic music generation. At present, the research on music generation via deep learning can be divided into two categories: predictive models and generative models. However, both categories have the same problems that need to be resolved. First, the length of the music must be determined artificially prior to generation. Second, although the convolutional neural network (CNN) is unexpectedly superior to the recurrent neural network (RNN), CNN still has several disadvantages. This paper proposes a conditional generative adversarial network approach using an inception model (INCO-GAN), which enables the generation of complete variable-length music automatically. By adding a time distribution layer that considers sequential data, CNN considers the time relationship in a manner similar to RNN. In addition, the inception model obtains richer features, which improves the quality of the generated music. In experiments conducted, the music generated by the proposed method and that by human composers were compared. High cosine similarity of up to 0.987 was achieved between the frequency vectors, indicating that the music generated by the proposed method is very similar to that created by a human composer.

3D (3-Dimensional) object recognition is a hot research topic that benefits environment perception, disease diagnosis, and the mobile robot industry. Point clouds collected by range sensors are a popular data structure to represent a 3D object model. This paper proposed a 3D object recognition method named Dynamic Graph Convolutional Broad Network (DGCB-Net) to realize feature extraction and 3D object recognition from the point cloud. DGCB-Net adopts edge convolutional layers constructed by weight-shared multiple-layer perceptrons (MLPs) to extract local features from the point cloud graph structure automatically. Features obtained from all edge convolutional layers are concatenated together to form a feature aggregation. Unlike stacking many layers in-depth, our DGCB-Net employs a broad architecture to extend point cloud feature aggregation flatly. The broad architecture is structured utilizing a flat combining architecture with multiple feature layers and enhancement layers. Both feature layers and enhancement layers concatenate together to further enrich the features’ information of the point cloud. All features work on the object recognition results thus that our DGCB-Net show better recognition performance than other 3D object recognition algorithms on ModelNet10/40 and our scanning point cloud dataset.

Malware detection and classification methods are being actively developed to protect personal information from hackers. Global images of malware (in a program that includes personal information) can be utilized to detect or classify it. This method is efficient, given that small changes in the program can be detected while maintaining the overall structure of the program. However, if any obfuscation approach that encrypts malware code is implemented, it becomes difficult to extract features such as opcodes and application programming interface functions. Given that malware detection and classification are performed differently depending on whether malware is obfuscated or not, methods that can simultaneously detect and classify general and obfuscated malware are required. This paper proposes a method that uses a generative adversarial network (GAN) and global image-based local image to classify unobfuscated and obfuscated malware. Global and local images of unobfuscated malware are generated using pixel and local feature visualizers. The GAN is utilized to visualize local features and generate local images of obfuscated malware by learning global and local images of unobfuscated malware. The local image of unobfuscated malware is merged with the global image generated via the pixel visualizer. To merge the global and local images of unobfuscated and obfuscated malware, the pixels extracted from global and local images are stored in a two-dimensional array, and then merged images are generated. Finally, unobfuscated and obfuscated malware are classified using a convolutional neural network (CNN). The results of experiments conducted on the Microsoft Malware Classification Challenge (BIG 2015) dataset indicate that the proposed method has a malware classification accuracy of 99.65%, which is 2.18% higher than that of the malware classification approach based on only global images and local features.

Point clouds have been widely used in three-dimensional (3D) object classification tasks, i.e., people recognition in unmanned ground vehicles. However, the irregular data format of point clouds and the large number of parameters in deep learning networks affect the performance of object classification. This paper develops a 3D object classification system using a broad learning system (BLS) with a feature extractor called VB-Net. First, raw point clouds are voxelized into voxels. Through this step, irregular point clouds are converted into regular voxels which are easily processed by the feature extractor. Then, a pre-trained VoxNet is employed as a feature extractor to extract features from voxels. Finally, those features are used for object classification by the applied BLS. The proposed system is tested on the ModelNet40 dataset and ModelNet10 dataset. The average recognition accuracy was 83.99% and 90.08%, respectively. Compared to deep learning networks, the time consumption of the proposed system is significantly decreased.

To provide a realistic environment for remote sensing applications, point clouds are used to realize a three-dimensional (3D) digital world for the user. Motion recognition of objects, e.g., humans, is required to provide realistic experiences in the 3D digital world. To recognize a user’s motions, 3D landmarks are provided by analyzing a 3D point cloud collected through a light detection and ranging (LiDAR) system or a red green blue (RGB) image collected visually. However, manual supervision is required to extract 3D landmarks as to whether they originate from the RGB image or the 3D point cloud. Thus, there is a need for a method for extracting 3D landmarks without manual supervision. Herein, an RGB image and a 3D point cloud are used to extract 3D landmarks. The 3D point cloud is utilized as the relative distance between a LiDAR and a user. Because it cannot contain all information the user’s entire body due to disparities, it cannot generate a dense depth image that provides the boundary of user’s body. Therefore, up-sampling is performed to increase the density of the depth image generated based on the 3D point cloud; the density depends on the 3D point cloud. This paper proposes a system for extracting 3D landmarks using 3D point clouds and RGB images without manual supervision. A depth image provides the boundary of a user’s motion and is generated by using 3D point cloud and RGB image collected by a LiDAR and an RGB camera, respectively. To extract 3D landmarks automatically, an encoder–decoder model is trained with the generated depth images, and the RGB images and 3D landmarks are extracted from these images with the trained encoder model. The method of extracting 3D landmarks using RGB depth (RGBD) images was verified experimentally, and 3D landmarks were extracted to evaluate the user’s motions with RGBD images. In this manner, landmarks could be extracted according to the user’s motions, rather than by extracting them using the RGB images. The depth images generated by the proposed method were 1.832 times denser than the up-sampling-based depth images generated with bilateral filtering.

Fast and accurate obstacle detection is essential for accurate perception of mobile vehicles’ environment. Because point clouds sensed by light detection and ranging (LiDAR) sensors are sparse and unstructured, traditional obstacle clustering on raw point clouds are inaccurate and time consuming. Thus, to achieve fast obstacle clustering in an unknown terrain, this paper proposes an elevation-reference connected component labeling (ER-CCL) algorithm using graphic processing unit (GPU) programing. LiDAR points are first projected onto a rasterized x–z plane so that sparse points are mapped into a series of regularly arranged small cells. Based on the height distribution of the LiDAR point, the ground cells are filtered out and a flag map is generated. Next, the ER-CCL algorithm is implemented on the label map generated from the flag map to mark individual clusters with unique labels. Finally, obstacle labeling results are inverse transformed from the x–z plane to 3D points to provide clustering results. For real-time 3D point cloud clustering, ER-CCL is accelerated by running it in parallel with the aid of GPU programming technology.

Plane extraction is regarded as a necessary function that supports judgment basis in many applications, including semantic digital map reconstruction and path planning for unmanned ground vehicles. Owing to the heterogeneous density and unstructured spatial distribution of three-dimensional (3D) point clouds collected by light detection and ranging (LiDAR), plane extraction from it is recently a significant challenge. This paper proposed a parallel 3D Hough transform algorithm to realize rapid and precise plane detection from 3D LiDAR point clouds. After transforming all the 3D points from a Cartesian coordinate system to a pre-defined 3D Hough space, the generated Hough space is rasterised into a series of arranged cells to store the resided point counts into individual cells. A 3D connected component labeling algorithm is developed to cluster the cells with high values in Hough space into several clusters. The peaks from these clusters are extracted so that the targeting planar surfaces are obtained in polar coordinates. Because the laser beams emitted by LiDAR sensor holds several fixed angles, the collected 3D point clouds distribute as several horizontal and parallel circles in plane surfaces. This kind of horizontal and parallel circles mislead plane detecting results from horizontal wall surfaces to parallel planes. For detecting accurate plane parameters, this paper adopts a fraction-to-fraction method to gradually transform raw point clouds into a series of sub Hough space buffers. In our proposed planar detection algorithm, a graphic processing unit (GPU) programming technology is applied to speed up the calculation of 3D Hough space updating and peaks searching.

The importance of cybersecurity has recently been increasing. A malware coder writes malware into normal executable files. A computer is more likely to be infected by malware when users have easy access to various executables. Malware is considered as the starting point for cyber-attacks; thus, the timely detection, classification and blocking of malware are important. Malware visualization is a method for detecting or classifying malware. A global image is visualized through binaries extracted from malware. The overall structure and behavior of malware are considered when global images are utilized. However, the visualization of obfuscated malware is tough, owing to the difficulties encountered when extracting local features. This paper proposes a merged image-based malware classification framework that includes local feature visualization, global image-based local feature visualization, and global and local image merging methods. This study introduces a fastText-based local feature visualization method: First, local features such as opcodes and API function names are extracted from the malware; second, important local features in each malware family are selected via the term frequency inverse document frequency algorithm; third, the fastText model embeds the selected local features; finally, the embedded local features are visualized through a normalization process. Malware classification based on the proposed method using the Microsoft Malware Classification Challenge dataset was experimentally verified. The accuracy of the proposed method was approximately 99.65%, which is 2.18% higher than that of another contemporary global image-based approach.

The importance of estimating human movement has increased in the field of human motion capture. HTC VIVE is a popular device that provides a convenient way of capturing human motions using several sensors. Recently, the motion of only users’ hands has been captured, thereby greatly reducing the range of motion captured. This paper proposes a framework to estimate single-arm orientations using soft sensors mainly by combining a Bi-long short-term memory (Bi-LSTM) and two-layer LSTM. Positions of the two hands are measured using an HTC VIVE set, and the orientations of a single arm, including its corresponding upper arm and forearm, are estimated using the proposed framework based on the estimated positions of the two hands. Given that the proposed framework is meant for a single arm, if orientations of two arms are required to be estimated, the estimations are performed twice. To obtain the ground truth of the orientations of single-arm movements, two Myo gesture-control sensory armbands are employed on the single arm: one for the upper arm and the other for the forearm. The proposed framework analyzed the contextual features of consecutive sensory arm movements, which provides an efficient way to improve the accuracy of arm movement estimation. In comparison with the ground truth, the proposed method estimated the arm movements using a dynamic time warping distance, which was the average of 73.90% less than that of a conventional Bayesian framework. The distinct feature of our proposed framework is that the number of sensors attached to end-users is reduced. Additionally, with the use of our framework, the arm orientations can be estimated with any soft sensor, and good accuracy of the estimations can be ensured. Another contribution is the suggestion of the combination of the Bi-LSTM and two-layer LSTM.

Clustering plays an important role in processing light detection and ranging points in the autonomous perception tasks of robots. Clustering usually occurs near the start of processing three-dimensional point clouds obtained from light detection and ranging for detection and classification. Therefore, errors caused by clustering will directly affect the detection and classification accuracy. In this article, a clustering method is presented that combines density-based spatial clustering of application with noise and two-dimensional range image composed by scan lines of light detection and ranging based on the order of generation time. The results show that the proposed method achieves state-of-the-art performance in aspect of time efficiency and clustering accuracy. A ground extraction method based on scan line is also presented in this article, which has strong ability to separate ground points and non-ground points.

Micro unmanned aircraft systems (micro UAS)-related technical research is important because micro UAS has the advantage of being able to perform missions remotely. When an omnidirectional camera is mounted, it captures all surrounding areas of the micro UAS. Normal field of view (NFoV) refers to a view presented as an image to a user in a 360-degree video. The 360-degree video is controlled using an end-to-end controls method to automatically provide the user with NFoVs without the user controlling the 360-degree video. When using the end-to-end controls method that controls 360-degree video, if there are various signals that control the 360-degree video, the training of the deep learning model requires a considerable amount of training data. Therefore, there is a need for a method of autonomously determining the signals to reduce the number of signals for controlling the 360-degree video. This paper proposes a method to autonomously determine the output to be used for end-to-end control-based deep learning model to control 360-degree video for micro UAS controllers. The output of the deep learning model to control 360-degree video is automatically determined using the K-means algorithm. Using a trained deep learning model, the user is presented with NFoVs in a 360-degree video. The proposed method was experimentally verified by providing NFoVs wherein the signals that control the 360-degree video were set by the proposed method and by user definition. The results of training the convolution neural network (CNN) model using the signals to provide NFoVs were compared, and the proposed method provided NFoVs similar to NFoVs of existing user with 24.4% more similarity compared to a user-defined approach.

Nowadays, deep learning methods based on a virtual environment are widely applied to research and technology development for autonomous vehicle’s smart sensors and devices. Learning various driving environments in advance is important to handle unexpected situations that can exist in the real world and to continue driving without accident. For training smart sensors and devices of an autonomous vehicle well, a virtual simulator should create scenarios of various possible real-world situations. To create reality-based scenarios, data on the real environment must be collected from a real driving vehicle or a scenario analysis process conducted by experts. However, these two approaches increase the period and the cost of scenario generation as more scenarios are created. This paper proposes a scenario generation method based on deep learning to create scenarios automatically for training autonomous vehicle smart sensors and devices. To generate various scenarios, the proposed method extracts multiple events from a video which is taken on a real road by using deep learning and generates the multiple event in a virtual simulator. First, Faster-region based convolution neural network (Faster-RCNN) extracts bounding boxes of each object in a driving video. Second, the high-level event bounding boxes are calculated. Third, long-term recurrent convolution networks (LRCN) classify each type of extracted event. Finally, all multiple event classification results are combined into one scenario. The generated scenarios can be used in an autonomous driving simulator to teach multiple events that occur during real-world driving. To verify the performance of the proposed scenario generation method, experiments using real driving video data and a virtual simulator were conducted. The results for deep learning model show an accuracy of 95.6%; furthermore, multiple high-level events were extracted, and various scenarios were generated in a virtual simulator for smart sensors and devices of an autonomous vehicle.

In traditional music composition, the composer has a special knowledge of music and combines emotion and creative experience to create music. As computer technology has evolved, various music-related technologies have been developed. To create new music, a considerable amount of time is required. Therefore, a system is required that can automatically compose music from input music. This study proposes a novel melody composition method that enhanced the original generative adversarial network (GAN) model based on individual bars. Two discriminators were used to form the enhanced GAN model: one was a long short-term memory (LSTM) model that was used to ensure correlation between the bars, and the other was a convolutional neural network (CNN) model that was used to ensure rationality of the bar structure. Experiments were conducted using bar encoding and the enhanced GAN model to compose a new melody and evaluate the quality of the composition melody. In the evaluation method, the TFIDF algorithm was also used to calculate the structural differences between four types of musical instrument digital interface (MIDI) file (i.e., randomly composed melody, melody composed by the original GAN, melody composed by the proposed method, and the real melody). Using the TFIDF algorithm, the structures of the melody composed were compared by the proposed method with the real melody and the structure of the traditional melody was compared with the structure of the real melody. The experimental results showed that the melody composed by the proposed method had more similarity with real melody structure with a difference of only 8% than that of the traditional melody structure.

With intelligent big data, a variety of gesture-based recognition systems have been developed to enable intuitive interaction by utilizing machine learning algorithms. Realizing a high gesture recognition accuracy is crucial, and current systems learn extensive gestures in advance to augment their recognition accuracies. However, the process of accurately recognizing gestures relies on identifying and editing numerous gestures collected from the actual end users of the system. This final end-user learning component remains troublesome for most existing gesture recognition systems. This paper proposes a method that facilitates end-user gesture learning and recognition by improving the editing process applied on intelligent big data, which is collected through end-user gestures. The proposed method realizes the recognition of more complex and precise gestures by merging gestures collected from multiple sensors and processing them as a single gesture. To evaluate the proposed method, it was used in a shadow puppet performance that could interact with on-screen animations. An average gesture recognition rate of 90% was achieved in the experimental evaluation, demonstrating the efficacy and intuitiveness of the proposed method for editing visualized learning gestures.

Clinical data analysis and forecasting have made substantial contributions to disease control, prevention and detection. However, such data usually suffer from highly imbalanced samples in class distributions. In this paper, we aim to formulate effective methods to rebalance binary imbalanced dataset, where the positive samples take up only the minority. We investigate two different meta-heuristic algorithms, particle swarm optimization and bat algorithm, and apply them to empower the effects of synthetic minority over-sampling technique (SMOTE) for pre-processing the datasets. One approach is to process the full dataset as a whole. The other is to split up the dataset and adaptively process it one segment at a time. The experimental results reported in this paper reveal that the performance improvements obtained by the former methods are not scalable to larger data scales. The latter methods, which we call Adaptive Swarm Balancing Algorithms, lead to significant efficiency and effectiveness improvements on large datasets while the first method is invalid. We also find it more consistent with the practice of the typical large imbalanced medical datasets. We further use the meta-heuristic algorithms to optimize two key parameters of SMOTE. The proposed methods lead to more credible performances of the classifier, and shortening the run time compared to brute-force method.

Games that use brainwaves via brain–computer interface (BCI) devices, to improve brain functions are known as BCI serious games. Due to the difficulty of developing BCI serious games, various BCI engines and authoring tools are required, and these reduce the development time and cost. However, it is desirable to reduce the amount of technical knowledge of brain functions and BCI devices needed by game developers. Moreover, a systematic BCI serious game development process is required. In this paper, we present a methodology for the development of BCI serious games. We describe an architecture, authoring tools, and development process of the proposed methodology, and apply it to a game development approach for patients with mild cognitive impairment as an example. This application demonstrates that BCI serious games can be developed on the basis of expert-verified theories.

Demand for interactive wearable devices is rapidly increasing with the development of smart devices. To accurately utilize wearable devices for remote robot controls, limited data should be analyzed and utilized efficiently. For example, the motions by a wearable device, called Myo device, can be estimated by measuring its orientation, and calculating a Bayesian probability based on these orientation data. Given that Myo device can measure various types of data, the accuracy of its motion estimation can be increased by utilizing these additional types of data. This paper proposes a motion estimation method based on weighted Bayesian probability and concurrently measured data, orientations and electromyograms (EMG). The most probable motion among estimated is treated as a final estimated motion. Thus, recognition accuracy can be improved when compared to the traditional methods that employ only a single type of data. In our experiments, seven subjects perform five predefined motions. When orientation is measured by the traditional methods, the sum of the motion estimation errors is 37.3%; likewise, when only EMG data are used, the error in motion estimation by the proposed method was also 37.3%. The proposed combined method has an error of 25%. Therefore, the proposed method reduces motion estimation errors by 12%.

Given that robots provide services in any locations after they move toward humans, the pervasive sensing environment can provide diverse kinds of services through the robots not depending on the locations of humans. For various services, robots need to learn accurate motor primitives such as walking and grabbing objects. However, learning motor primitives in a pervasive sensing environment are very time consuming. Several previous studies have considered robots learning motor primitives and interacting with humans in virtual environments. Given that a robot learns motor primitives based on observations, a disadvantage is that there is no way of defining motor primitives that cannot be observed by a robot. In this paper, we develop a novel interaction learning approach based on a virtual environment. The motor primitives are defined by manipulating a robot directly using demonstration-based learning. In addition, a robot can apply Q -learning to learn interactions with humans. In an experiment, using the proposed method, the motor primitives were generated intuitively and the amount of movement required by a virtual human in one of the experiments was reduced by about 25% after applying the generated motor primitives.

Unmanned aerial vehicles (UAVs) with auto-pilot capabilities are often used for surveillance and patrol. Pilots set the flight points on a map in order to navigate to the imaging point where surveillance or patrolling is required. However, there is the limit denoting the information such as absolute altitudes and angles. Therefore, it is required to set the information accurately. This paper hereby proposes a method to construct environmental symmetric big data using an unmanned aerial vehicle (UAV) during flight by designating the imaging and non-imaging points for surveillance and patrols. The K-Means-based algorithm proposed in this paper is then employed to divide the imaging points, which is set by the pilot, into K clusters, and K imaging points are determined using these clusters. Flight data are then used to set the points to which the UAV will fly. In our experiment, flight records were gathered through an UAV in order to monitor a stadium and the imaging and non-imaging points were set using the proposed method and compared with the points determined by a traditional K-Means algorithm. Through the proposed method, the cluster centroids and cumulative distance of its members were reduced by 87.57% more than with the traditional K-Means algorithm. With the traditional K-Means algorithm, imaging points were not created in the five points desired by the pilot, and two incorrect points were obtained. However, with the proposed method, two incorrect imaging points were obtained. Due to these two incorrect imaging points, the two points desired by the pilot were not generated.

The Internet of Things (IoT) is the intelligent technologies and services that mutually communicate information between humans and devices or between Internet-based devices. In IoT environments, various device information is collected from the user for intelligent technologies and services that control the devices. Recently, wireless sensor networks based on IoT environments are being used in sectors as diverse as medicine, the military, and commerce. Specifically, sensor techniques that collect relevant area data via mini-sensors after distributing smart dust in inaccessible areas like forests or military zones have been embraced as the future of information technology. IoT environments that utilize smart dust are composed of the sensor nodes that detect data using wireless sensors and transmit the detected data to middle nodes. Currently, since the sensors used in these environments are composed of mini-hardware, they have limited memory, processing power, and energy, and a variety of research that aims to make the best use of these limited resources is progressing. This paper proposes a method to utilize these resources while considering energy efficiency, and suggests lightweight mutual verification and key exchange methods based on a hash function that has no restrictions on operation quantity, velocity, and storage space. This study verifies the security and energy efficiency of this method through security analysis and function evaluation, comparing with existing approaches. The proposed method has great value in its applicability as a lightweight security technology for IoT environments.

Given that entropy-based IT technology has been applied in homes, office buildings and elsewhere for IT security systems, diverse kinds of intelligent services are currently provided. In particular, IT security systems have become more robust and varied. However, access control systems still depend on tags held by building entrants. Since tags can be obtained by intruders, an approach to counter the disadvantages of tags is required. For example, it is possible to track the movement of tags in intelligent buildings in order to detect intruders. Therefore, each tag owner can be judged by analyzing the movements of their tags. This paper proposes a security approach based on the received signal strength indicators (RSSIs) of beacon-based tags to detect intruders. The normal RSSI patterns of moving entrants are obtained and analyzed. Intruders can be detected when abnormal RSSIs are measured in comparison to normal RSSI patterns. In the experiments, one normal and one abnormal scenario are defined for collecting the RSSIs of a Bluetooth-based beacon in order to validate the proposed method. When the RSSIs of both scenarios are compared to pre-collected RSSIs, the RSSIs of the abnormal scenario are about 61% more different compared to the RSSIs of the normal scenario. Therefore, intruders in buildings can be detected by considering RSSI differences.