Search

Image captioning is the task of generating textual descriptions of a given image, requiring techniques of computer vision and natural language processing. Recent models have utilized deep learning techniques for this task to gain performance improvement. However, these models can neither fully use information included in a given image such as object and attribute, nor generate a domain-specific caption because existing methods use open dataset such as MSCOCO which include general images. To overcome these limitations, this paper proposes a domain-specific image caption generator, which generates a caption based on attention mechanism with object and attribute information, and reconstruct a generate caption using a semantic ontology to provide natural language description for given specific-domain. To show the effectiveness of the proposed model, we evaluate the image caption generator with a dataset, MSCOCO, quantitatively and qualitatively.

This paper compares the system-level parameters and output power of soft-material and hard-material piezoelectric energy harvesters taking into account the material softening nonlinear behavior. We validate the approach by verifying the compatibility with conclusions available in the literature at material level. Among the results, our system-level characterization confirms that soft-type materials behave intrinsically more nonlinearly and that neglecting their nonlinear behavior can lead to wrong conclusions about the system-level coupling coefficient. Then, we compare the generators in terms of power delivered to a load when actuated by a shaker. The power delivered by the hard-type transducer increases faster than the power delivered by the soft-type transducer when the acceleration amplitude increases which is also consistent with our characterization results and validates the system-level approach.

In this paper, We propose a novel video captioning model that utilizes context information of correlated clips. Unlike the ordinary “one clip - one caption” algorithms, we concatenate multiple neighboring clips as a chunk and train the network in “one chunk - multiple caption” manner. We train and evaluate our algorithm using M-VAD dataset and report the performance of caption and keyword generation. Our model is a foundation model for generating a video story using several captions. Therefore, in this paper, we focus on caption generation for several videos and trend analysis of the generated captions. In the experiments, we show the performance of intermediate results of our model in both qualitative and quantitative aspects.

Image captioning is the task of generating textual descriptions of a given image, requiring techniques of computer vision and natural language processing. Recent models have utilized deep learning techniques to this task to gain performance improvement. However, these models can neither distinguish more important objects than others in a given image, nor explain the reasons why specific words have been selected when generating captions. To overcome these limitations, this paper proposes an explainable image captioning model, which generates a caption by indicating specific objects in a given image and providing visual explanation using them. The model has been evaluated with datasets such as MSCOCO, Flickr8K, and Flickr30K, and some qualitative results are presented to show the effectiveness of the proposed model.

Image analogy is the process of creating an image filter that precisely reflects the characteristics contained in the training data. Recently, the image analogy problem was generally handled by deep neural network (DNN) with the development of a deep learning technology. Generally, DNN suffers from a fatal problem in that it requires large amounts of data for training. However, as pairs of images with the same relationship are needed for an image analogy, it is hard to collect sufficient data for image analogy using DNN. In order to solve this problem, we propose an image analogy method using a Gaussian process. In this method, a Gaussian process regression is used instead of DNN regression to adjust the feature vectors which will be used in creating filtered image. Additionally, in order to accelerate the training speed of Gaussian process, we also propose novel sampling methods that select salient instances from a given dataset. Our experiment result demonstrates that the proposed image analogy method using a Gaussian process with salient instance sampling performs significantly better than DNN in environments with small dataset size.

Navigation tasks for mobile robots have been widely studied over past several years. More recently, there have been many attempts to introduce the usage of machine learning algorithms. Deep learning techniques are of special importance because they have achieved excellent performance in various fields, including robot navigation. Deep learning methods, however, require considerable amount of data for training deep learning models and their results may be difficult to interpret for researchers. To address this issue, we propose a novel model for mobile robot navigation using deep reinforcement learning. In our navigation tasks, no information about the environment is given to the robot beforehand. Additionally, the positions of obstacles and goal change in every episode. In order to succeed under these conditions, we combine several Q-learning techniques that are considered to be state-of-the-art. We first provide a description of our model and then verify it through a series of experiments.

The piezoelectric energy harvesters for energy supplier have attracted attention due to its high energy density. These types of devices have features that they are resonated at specific frequency from external vibration and this frequency significantly impacts on the amount of harvested energy. However, most excitations in operating environment have broadband characteristics rather than one peak at certain frequency. Hence, the optimization of these properties is critical to maximize the generated power in the device. The focus of this study is to discuss the design and optimization of a piezoelectric device used to supply power to wireless temperature sensor network under wideband frequency operation. The performance of piezoelectric energy harvester was investigated based on the results of numerical calculation and harvesters were optimized to be operated under wideband frequency operation.

This paper proposes the rehabilitation treatment coach robot which will help at-home patients do their rehabilitation exercises at home without any professional trainers. The coach robot is designed to be cheap enough for patients to afford it. The robot suggests the rehabilitation program and corrects the posture of the patients during the exercise. The deep neural network is used for posture correction. Besides, the voice interface is applied for convenient interaction between robot and patients during the exercise. The emergency detection module is adopted which will inform doctors when emergency happens on patients. The emergency detection will be implemented using deep neural network on voice input and video input simultaneously. The detailed data collection plan for training deep neural network and performance evaluation plan are also provided in the paper.

It is still a long way to communicate humans and machines emotionally. There are some tries to provide sentimental conversations among humans and machines. Computational humor is one of research topics in computational linguistics and artificial intelligence. We introduce a new method to generate jokes in a sentence related temporal and spatial contexts for continuous conversations with images. We propose a novel model based on a recurrent neural network with natural language processing (NLP) and understanding (NLU) methods. The method generates jokes in a sentence considering temporal and spatial context. The method can joke to trend sensitive users according to different points of humor that vary from region to region. Through this, the user can feel the interest of the conversational service with humorous responses or contents. We apply the method to some applications such as psychiatric counseling and stress management to enhance the applicability of conversational service.

The rehabilitation treatment is important because it helps a patient restore physical sensory and mental capabilities. The patient whose symptoms are moderately relieved, or outpatient, usually rehabilitate the individual alone. Improper exercise or posture can slow the recovery of the patient or even worsen the patient's health status when doing rehabilitation exercise alone. The best way is to receive home visiting treatment from professional therapist until cured. However, such way is a burden on the patient in terms of cost. This paper proposes the novel model that corrects the improper postures of the patient when having rehabilitating exercise alone. We use Microsoft Kinect to recognize the posture of the patient by extracting the human skeleton. We will adopt deep neural network to analyze the extracted human skeleton, in order to determine whether the posture is correct or not. The data for training our model will be correct postures and incorrect postures and detailed data collection plan is provided in this paper. The implementation and experiment will be performed in the future work.

In this work, we propose a concept of emergency detection algorithm for healthcare robot which adopts discriminative restricted Boltzmann machine for anomaly detection. We will adopt anomaly detection rather than simple emergency case classification as it is hard to collect real emergency data to train the effective classifier. The conventional anomaly detection method uses decision tree to analyze the signals obtained from the sensors attached on the bodies of the patients to find out the emergency situations. We propose anomaly detection using video and audio signals as they are easy to be obtained by the healthcare robot, with equipping a camera and a microphone, and it is much more convenient for patients. The discriminative restricted Boltzmann machine which is specialized in learning probability distribution in an unsupervised manner will be applied for anomaly detection. This paper only provides the novel idea for emergency detection. The implementation and the experiments will be conducted in the future work.

In previous studies, we proposed and developed a CloudDMSS (Cloud-based distributed multimedia streaming service) system. To provide reliable streaming service in the CloudDMSS, a distribution streaming system was established using streaming task distribution algorithms, RR (Round Robin), and LC (Least Connection). However, as the RR and LC methods did not consider the utilization or streaming transmission rate of the streaming servers, it negatively affected the load on the servers. Therefore, to improve the capability of our CloudDMSS, this study proposes a streaming task distribution method using an SRC(Streaming Resource-based Connection) algorithm that considers the utilization of the streaming server.

In this paper, a framework of sporadic noise detection and cancellation is proposed. To improve the intelligence of sporadic noise detection, the difference residual detective signalbased detection algorithm is proposed. For performance evaluation, missing and false error rates, and speech recognition rates are measured. The results show remarkable performance improvement.

This paper reports on measurement results for the simultaneous use of multiple WLAN APs in a large area. We describe some problems of applying legacy IEEE 802.11 WLAN technologies to a freight container terminal as a special application example. We begin by describing the characteristics of IT infrastructures in a container terminal environment. Then we examine S/W application requirements and work processes in a container terminal. One of the key problems based on clients' complaints is frequent network disconnections. To find the cause of this problem, we observe the characteristics of service clients and collect the raw network data from actual measurements in a container terminal. We transform the collected raw data into meaningful secondary data and analyze the results after verifying our measurement approaches. The goal of this paper is providing a special measurement report and suggesting reasonable solutions for discovered problems.

This paper propose adaptive noisy signal reduction method on smart devices for voice communication service in pervasive home environment. In order to reduce noisy signals efficiently on smart devices we provide Wiener filtering technique based on input-SNR estimation method with preprocessing on smart home devices.

A modeling support system that provides an easy and fast FE-modeling for the static and durability analysis of the lower arm in an automobile suspension module is described. The system can exclude human interactions such as meshing using pre-defined shell elements and can impose complicated boundary and loading conditions via a multi-point constraints technique. The platform of this system is implemented based on MSC.PATRAN, and MSC.NASTRAN and MSC.FATIGUE are utilized as the solver. The parametric geometry modeling tool utilizing CATIA V5 is linked with the modeling support system, which enables changes in 3D models, that occur frequently in an early stage of the design process, to be taken into account. The integration techniques using XML-based wrapper enable to integrate the overall process. This system was applied to an actual work process to calculate the fatigue durability of the lower arm in an automobile suspension module, and time and cost reduction for this analysis were validated.